 doing it in production. We'll fix it in post and they start laughing. We'll fix it in post, but we're doing it live. That's a fun part of our work are all these movie companies that we get to work with. That. We get to, they actually do things in post. They are some of my favorite, honestly. They have cool stuff. If anyone doesn't know, movie companies from a tech side are cool to work with. Welcome to the blog. There's a number of 321 Rack questions, lab testings, errata, and the Rack questions and probably I should have titled it hardware reliability kind of go hand in hand because I knew when I did the Rack video, the armchair people would get upset of what wasn't there. And there's at least one user who just thinks everything should have redundant connections. And when they warrant it, but when they don't, why do I need redundant 10 gig links? I'm trying to figure that out still. If you're in a colo, I get it, where you're not, yeah, it's a four hour drive for me to go fix that. Okay, I understand that. Yeah, so the Rack video just invited so many comments related to that. So it's one of those things that I'm gonna pull up and bring it up as an image here so we can kind of walk people through it for those that didn't watch the Rack video. But hey, I do have the Rack video, so you can watch that video. Come back and play us a 2X and catch up to the conversation. But um, but um, but um, but this is the back of our Ryzen servers and you know, they don't have, we have separated our storage versus network. That part is done, but we have not done though is created any redundant links between them because it's just not that necessary for what we do. It comes down to what your use case is. So I will say it's not saying never, but I'm saying not very frequently is this needed. So this is the basic setup here. We have this 10 gig link going to storage, which is down here at the 45 drives. We have these 25 gig links on there. And oh, someone asked, and do you know the answer, Steve, to why I put QR codes on there? In the hopes that someone would scan them? Yes, cause they're Rick Rolls. Yeah. So it's not really an organization thing. It's more about just, I did it for the lulls when I put it on there. I wanted to label the one that was for storage, but then you know, when the orange cables came in, it became rather obvious which ones are storage, but yeah, I color coded everything when I designed this. Yeah. So the, I have to remember now, I want to say the green, the green is the IPMI. That's what I thought. Green's IPMI, gray is something. Yeah, all the management network for it. Well, no, IPMI is network management. And then the gray is the secure network, if you will, the separate with a different thing on there. So, I mean, I granted like the power supply, we didn't go through it on the power supplies. We cared more about power efficiency. And if you go read on the power supplies, you lose, you sacrifice a level of efficiency. We have two servers. So if a server goes down, I can start the VMs on another server. Plus we're technical people and it isn't going to take us more than, I mean, I would say maybe an hour to replace a power supply. If we're dragging our feed and I have to get off the phone and I got to unbox it. Like maybe there's an hour of from the time it died to the time it is completely down there. So it's just one of those things that people overthought it so much. It just comes down to, we got rid of, and we did because we have a use case for it. We had some R730s and we sold them to a client. They have levels of redundancy in them. And we sold them to a client who needed that level of redundancy because the power supply failure for them, even an hour is a bad thing to them. So it's actually just a spare server for some workloads. Like they needed an extra server. So there's definitely times to put HA in there, but in our office is not one of them. I just don't know why people think like I should have a power bill double what I have. Not to mention their consumer power supplies. It's not like we're talking a Dell proprietary slide in power supply that we're gonna be on the phone and you are a few days out on shipping. Yeah. We have them on the shelf. And in fact, we keep a spare one in a box in the rack. And I wanna point this out though, because this is the other side of it too. And we can tell you this because we manage a lot of systems. And Steve, how many tickets a day are for hardware failure? Almost none, what? Almost none for hardware failure that you could make fault tolerant. I can't solve, you know, your printer died. Yeah. I wanna answer this question for you here. Is it common practice to have a storage server and another virtualization that stores the VMs on the storage server OS and everything? Yes, that's a common setup in the enterprise. You generally have a separation between storage and compute. Even in your cloud compute, even though you may see it as a thing, there's not necessarily, there's a separation. There's the sand that you have at the large enterprise versus the compute nodes. But if you read this, and I can throw a link to this for people interested, this is Tom's hardware. This is a review from Puget Releases Hard Reliability Part. They're a high-end builder of systems. But one of the things they talk about here, and it's gonna be kind of small to read on the screen. So let me throw a link in there for all those following at home. Basically, too long, didn't read with a limited exception of this field failure rate of these, what do you call it, NVIDIA Quattro. That was the only thing that even had a higher failure rate among GPU cards. Their failure rate, though, out in the field for the more common RTX cards and things that was 0.26%. But in terms of server failures, most of these server failure problems they're having, like these are processor failure rates from 2019 to 2021, so modern processors. You're looking at, with one fluke being the Intel Core 11, or Intel Core 11th generation one with a 5%, most of these failure rates, I mean, look at this, 0.29%, 0.8%. It's just one of those things that they just don't have. Hardware does not have the failure rate it used to have. It's always a software problem. Like our tickets a month are rarely a hardware failure. There's one exception to that. And of course, this is why TrueNAS is so popular. I thought this was funny. Once you start breaking down all of the stuff, let me find the exact ones. There's another page I have here, right here. So among hardware failures, what fails? 80% of the failures, hard drives. Like hard drives are the failure rate. And I think those Quattro cards fail because when you look at the heatsink design on an RTX versus the Quattro, you'll see why those fail. Yeah. That was also, Puget Mode notes that these are, what do you call it? They're also systems that are the $4,000 ones. They're not as well mass produced. They're like the ones that we just put in for that 1AMD what we just did. They're not as mass produced. They're kind of a niche system. And most of their failures were in the lab. So they were shipped to them like that. They didn't fail after they left the build. They failed the build test. So that's why Puget has two numbers on all these. Like bad out of the box versus bad in production. The failed in the field is that other number. So 6.7% failed out of the box essentially, like in their lab, 0.7% in the field. So once, if they worked when they got there and like they set them up and started working them, they continued out in the field over the next few years working. So, but back to, it's pretty much hard drives though. That's one of the things like, and RAID saves you from that. That's the most common failure is and not too surprisingly is hard drives. Now, SSD failure rates are not that high, I think. Do they have the SSD failures in here? Yeah, here we go. They're on this page. The only thing that scares me about SSD failures is you don't know they're coming. They're catastrophic. Yeah, they're just, it doesn't work today. Yep, it works until it doesn't. There was no warning. There's no, the hard drive. You can hear it. It feels a little slower. You're like, something seems a little off. And in fact, my mom was just telling me about that at her work. The IT guy came out and she's like, yeah, I'm trying to do this stuff and it feels really slow. He starts going through processes and that looks at, he goes, your drive's dying. I'm just gonna take this and clone it now before it dies. But the failure rate on the SSDs, the Samsung evos were at 0%. That seems sus. The Samsung 860 Pros are at 0.1%. And the 980 Pros are at 0.11 at the highest. So I don't think it's that, it's just one of those things, like it's, they're just not that common. You know, it's just not that common for hardware to fail with hard drives being the limited exception in that. So yeah, it's just one of those things. And this is why, you know, for now, the rate arrays and TrueDance are incredibly popular because we still haven't got the, especially the spinning media, but generally speaking in media, it's just not that fault tolerant. So it's just one of the few things I want to share with the crowd here. And this is, you know, working in the industry gives you an insight into this because the problem you have is if you have something fail, it may be stuck in your head that's the favorite. How many of that's something that you have? With us, we manage IT for like 70 different companies we're doing all these tickets from all this consulting that comes in from YouTube and everything else we get going on. It's like we're interacting in it all the time and we're having people call us when we're having problems and most of those problems they have are not hardware. They're just misconfigured software. Two calls I did today were completely goofed up systems in virtualization. So that was like, and we didn't find a hardware problem yet with either one of these systems. The problem with SSD raiders, they all fail at the same time. There is a chance of that when H, who was it? Seagate had those SSDs that had the bad firmware. There was a timer. They only would work till a certain amount of time. Unfortunately, they were all put in at that same time all started at that same time. So predictably they would fail all at that same time. Yeah. So that's definitely an issue. Now I see someone here, I've seen more PSU failures and hard drive failures. If you have something in a hot room, yeah, that's a thing. There's definitely bad power, dust, dirt, like power supply failures are common under less than ideal conditions. Yes. I think I've seen more hard drive than PSU but usually when a PSU goes it's, there's so many variables like Tom said with the bad power in that but also what all are you running in it and who made it? Yeah. And as far as like the Dell PSUs and those, I can't think of a time I've had to replace a Dell PSU like for all the servers I don't think we ever have. And they're nice and we're in, we have the redundant hotswap ones in a lot of these pretty much we always, we even, we retired those. What were those the Dell 2960s that, I mean, those things are hugely inefficient, old servers way past the end of life but they're still spinning and warring and there's not, and there's just like dust bunnies hanging off the back of them and the thing just keeps working. Like it's so dirty, I can't believe it's on. We have that client who had the server that saw the Clinton administration. Yeah. And it was still going. Exactly. So, I mean, consumer ones, yeah, they're definitely gonna be there. Yeah, the Dell and HP consumer, they're notorious for putting the bare minimum wattage needed. So as the power supply ages and the wattage now drops, the effective output drops, they go. Yeah. Speaking of redundant arrays of things, we did get a tornado in. So I'll be doing a quick review of that. It's just, it's novel, it's a tornado. I just really like the whole 45 dry setup. The tornadoes are just one more. We haven't done any testing on one. They just, they don't top load, they front load, but they're SSD arrays. So that thing's all set up. That's, I think I got, yeah, I got the tab over here. So I haven't configured anything yet on it. It's just configured in terms of, I wanted to make sure it worked. I'll do some load testing on it and make sure it does what it's supposed to do. We created a test pool. But once again, why is there an array of dissonance thing? Because stuff goes wrong. We got some 3.49 TB SSDs in it. So I'm sure it will be fast. Stornado is badass. I'll go with that. The next topic, I'll let Steve comment on this one because people ask any question all the time and it was asked again today of Steve, but I did a whole video on this and Unify really stands out in the market alone. I know TP-Link is a thing, but for the most part, you can't get much better than Unify for the price point management. People ask about Mikrotik and for example, if you have 20 switches in your organization, the management plane for Mikrotik, not as good. Management plane for Netgear. Netgear has some cloud offering, it's okay. And Genius has a cloud offering. I did a review of it, not okay. It's ugly. And all of these, of course, don't allow you self-hosted controllers. So if you start with a prerequisite of I want to self-host a controller and a reasonably priced switch, you're gonna run into going Unify or TP-Link. And TP-Link, I don't have the biggest trust and security or longevity of the product and they're what, 20% less than Unify? They're not like half. I wouldn't, the TP-Link Omoda stuff I wouldn't be looking at because like you said, it's such a close price to the Unify, get the Unify, which has a more reputable controller. TP-Link is where I go when I want like a cheap one-off switch. Yeah. Like they don't make bad hardware, but... I don't trust them for security very well. Like I just don't feel like there is on it with security. It's, so some, and this may be a different market and this is something interesting that I don't have good insight into. Omoda is about 40% off in mid Europe. Now that's where I do know there's some big differences. Some of them hardware being American-centric here and I don't look up and understand the pricing models in Europe as much. Yeah, there may be a bigger price difference in Europe for Unify than there is here with import taxes and things like that. That I can't speak to as well. So, but I also don't feel just like Steve, we've talked about like, I remember Log4J was an easy example of a security update that took a while. Yeah, and I think some of it is with Europe and that look at with the client we have in Canada who buys Unify equipment from the US store and then has it shipped to us because he can't get it in the Canadian store. Yeah, some things are just not available. So once you start crossing markets, I know that muddies the waters a bit in terms of that. But yeah, it's just, there's still not a competitor but I think the reason there's not a competitor it's like Unify has got like a $13 billion market cap. They're not easy to, if I wanted to start up a company I would have to raise an absolute ton of capital then get people to believe in my product which is a long multi-year run by reaching out to a bunch of YouTubers and sending them hardware, which is exactly what Ubiquiti did. They reached out to lots of YouTubers and sent us hardware. I think Cody's here. Yeah, and Cody can, Cody just think them, they sent Cody something else. They don't send me anything else because I was too harsh under product I guess, I don't know. We've said mean things about the disappointment machine. Yeah, I keep calling it a Unify disappointment machine so that is that. So yeah, I think it's just kind of one of those things it's so hard to compete in the market against them and the fact that you only can compete by selling massive amounts of hardware because you're not, the moment you charge for your controller is the moment you lose out so to speak to, oh, let's go back to Ubiquiti. Cause there's a couple of companies that tried to compete with them. I had someone reach out to me and I asked them how and they didn't reply yet. They said they're building a Ubiquiti competitor. I said, how? Because ingenious told me they were too and how bad was the ingenious locally hosted controller Steve? Wait, I was responding to something I drifted out for the question. How bad was the ingenious when they tried to compete with them? Oh, so when I did that setup, that was a few years back. They had three different, I think two or three different controllers at the time depending on what ingenious product line you bought and the one the client had, you couldn't get, you could self host it but it wasn't an installer. It was an image for a Linux OS with, that it was designed to run on Raspberry Pi basically. So you got this ISO that you had to spin up on a Raspberry Pi or load in VMware and then configure the web ports for and then log into. Yeah, that's, it was just bad. They abandoned it. The ingenious has their own cloud now and it's just, yeah. I will answer this person cause I've seen some questions on it. Reolink versus Amcrest, Amcrest? So we bought a bunch of Reolink. We had a high failure rate, not like 50% or anything crazy, but high. 10% maybe. Maybe 10%. It was annoying. And their RMA process was so-so. Yeah, but that's actually why I use Western Digital instead of Seagate, the RMA process. So we moved to Amcrest and we've only had a couple of them like we're down to maybe one, two percent. And the one that I just replaced was a bad seal. That happens on anything. Yeah. So our feelings are still that they're pretty good. So the, I haven't tested Reolink recently to see if they make a better camera today than, I mean, the last time I bought one was I think 2019. So we actually haven't bought one in a couple of years. We've used a few GeoVisions. And to my understanding, those are all still working. And we have a client who had a GeoVision system from another company that, most of their problems are like the cheap PoE switch the other company used died. So if you can't get a, if you can't get an Amcrest, take a look at a GeoVision. Yeah. I also will comment. Someone said Cisco has a higher market cap than the Amiquiti, but Cisco's not. Cisco's like a completely different company. Like Amiquiti is a niche company. Cisco's a mega company doing so many different things. Like they don't just make switches. But so, but it's, and they, they have, as me and Steve were trying to point out last week and plug Thursday, Cisco's trying to compete in that market space. They're doing an awful job of it. I mean, sure they have the enterprise market. There's no doubt because people are used to buying them and they make some good products. And they certainly, when you get to the enterprise level switches, they have features that you do not get with ubiquity. And I'll agree with that. But do you need those features is the question. It's like, I mean, Freightliner makes a hell of a truck and you know, but I don't really need it to go to the grocery store, but it definitely can haul way more than my truck. There's no doubt like the, the half a million dollar Freightliner truck certainly is able to do more than Tom's Dodge truck, but it's not what I need to go to the grocery store or tow my motorcycle. So, you don't buy it, just go your rent it. Yeah, the license fees are a thing. Amcrest comparable quality. I think you meant to say real link quality, video quality on the real link was really not bad. It was good quality. All of our issues with it were more of a physical build quality. Like we had more of them that tend to have like bullet cameras that would get moisture in them. Yeah. And by the way, this is a, let me confirm it before I say it, this is before I share it, I should say, but I think this one's one of the, is this a, what model is this? Yeah. So I have, we had that real link at the office laying around because we didn't buy any more. So I've got the real link at my house, it works. It's the real link, RLC 422. I've had it now in my house for a while. We had it as a demo unit at the office, but it even has the, this is a PTZ model. So I can actually zoom in and out with it. And it works great. I haven't any problems at all with this one. It's also indoors. It's also, yeah, it's in my, it's inside my garage. It's not exposed to the weather. And that I think is where most of our failures were were on exposed ones. Yeah, mostly, but oddly, I know we have real link at that one outdoor park and it works. That's been where we had to replace one. Oh, did we? Yeah, we did have to replace one there. We have them out at that moving company. We've had to, I think, replace company ones because we had to get a lift to go replace them. That's the part that I said, all right, we're putting up a different one. So because I had to replace these. And it was same thing. It was moisture. And I think the one at the park was a bad seal on a dome and then the moving company was a bad seal on a bullet cam which is usually unheard of. Yeah. I have a video all about securing your network. I don't even care if they want to contact China. You don't need to give your cameras internet access. Don't do it. Usually what we do when we do a Synology setup, only the Synology gets internet access. So we usually create the firewall rule. The IP address the Synology is at, you get internet. Everything else, nothing. You don't need to give internet. So it becomes a complete non-issue. That's the important part is mitigate the problem by not allowing it to happen. You know, I maybe do all videos sometime of where do these camera reach out to? All they reach out to that I've ever seen though is like the Amcrest ones, they're just paying some NTP servers. They're not sending any data. Not enough data, I should say. They're certainly not sending a camera stream based on how little data they send. So in the likelihood of these guys, I think they're more inept than they are in nefarious. Like they're built in China, they're set to use China time. Yeah. They'll look, what is the server they look up? So maybe I'll do this a video. Someone will be curious about it. I'll just show all the logs of where they go, which is really nowhere. They're not sending much data. Have I looked at frigate? No, I haven't had time, I'm in busy. It's, for those of you wondering, frigate is an open source NVR. It still looks like just a hobby project has been low on my list of, like it looks neat, like, hey, I wanna play with that, but I'm busy doing lots of other things. But it's novel because it's an open source project for that has some cool object tracking. It's not anything I would ever, it's not ready for commercial usage at all. That's not that I could tell. But hey, it looks like a fun project for homeland people to play with. So definitely play with it. Do you wanna pull up the... Which one? One from Z-Racer, Z400 Racer, the Synology and Cameras on their own. Yeah, that one, I was gonna respond to it. I was gonna respond to it here. So we put the Synologies and the cameras, the NVR Synology and the cameras go on the same network. If you need a data Synology, yes, that is a separate Synology. We always separate those out, especially because you don't wanna mix the drive-wearing of your NVR Synology and the fact that you're gonna fill it up recording all that data with, hey, I need to put data somewhere. Yep. Someone else says, really not bad, have 25 outside cameras, had one broken. So what in 25? 4%. Yeah. A little bit better than us, I think. I don't think Yubi-Kee has, maybe I'm wrong. Is there a Yubi-Kee support for RDP? I don't think there is, natively. I don't know. I've never tried to use it, so. Yeah, I don't know that it would pass that. Yeah, I don't think there's a, I could be wrong. I've not used it for, I don't use Windows RDP and we don't have, I mean, most of our users, if they're using RDP, they VPN in first. So I would not expose RDP. No. Even if it's, cause the problem with RDP isn't two-factor, it's Microsoft's poor code and flaws in it. So I trust the people who write VPN software to be more robust, to be internet-facing than Microsoft's problems they have with writing RDP software, which they've had several flaws that allowed you to break their RDP. And you can keep it up to date all you want. There's always some new zero-day for RDP. Yep. Oh, let's see. Do you guys know of any good endpoint manager besides Mass Central? We don't, so my problem I have with things like Mass Central, I don't know if they've gone through any security auditing. I don't keep up with that software. Maybe they've gotten around to having a security audit. But when it comes to software, does you want to trust for managing things? There's a reason we use some of the commercial software. I would love, in Bitwarden's example, this, and by the way, you can dig around through old vlogs. You guys, maybe it was some of you, maybe it was someone else. I was always asked about things like, Tom, what about Bitwarden? I'm like, have they gone through a security audit? But they're the open source password manager. And I'm like, yeah, have they gone through a security audit? And when Bitwarden went through a security audit, we moved to Bitwarden. So I'm not against using an open source program that's gone through security auditing. I just haven't had a chance to look at Mass Central. And I certainly am not qualified myself to validate their security. And I'm certainly not about to rely on the fact that they wrote something that they think is secure, not until it goes through some rigorous testing. There's an RDB should we get off, but it doesn't actually work. If DoD funding doesn't get them to make you fix it, I don't know what will. Well, Microsoft's such a monopoly. Like Microsoft, they just can't make rules in the government that would break Microsoft because they can't, there's not an alternative. They've got such a monopoly on the market. So 3389 was never supposed to be internet-facing. Oh yeah, but you can go on show Dan and know that. Yeah, I responded to that one. Neither was SMB, but people still do it. We just did work for a client who had AFP exposed. He's like, yeah, we're gonna set up a VPN and I'm gonna retrain everyone. So we're gonna fix that. Yeah. He just took a moment. My thoughts on Veeam, Veeam's not bad. I have friends that use it. It looks good. They seem to like it. I've heard it's got some, it's not cheap, but it's supposed to work really well. Has nice integrations. I like the fact that you can roll your own storage on it so that part's nice. And someone says, yes, RDP can be configured to use Ubiqui 502. Okay. I don't know how, but I imagine it's a Google search to see how they integrate. Can't fix stupid, but you can bill for it. It's not possible to fix it, but we can send you a bill. Took over local business and not that they have QNAP exposed to the internet, they have RDP. Of course, I mean, why not go all in on this? It wouldn't be the first time. I went to a client's once and I'm like, how do you access your server? Cause they said they had a VPN and we were looking at replacing the firewall. I go, can you show me this VPN? Yeah, we just clicked the icon on our desktop. Hey, Brack, can you go to this IP address and RDP? Yeah, it's just open to the world. You don't have a VPN. Yep. Do you have plans to use Elastic or AMP monitoring? Not on my to-do list. Elastic's neat, but yeah. I sell to fix stupid every day, sometimes. Yep, yep, yep. I should make a shirt that says that. Yeah, maybe. Can you provide examples as to be exposed to the internet? Is that like having your NAS accessible for cloud access? Yeah, basically, the client, we did some work for, they're a mostly Apple shop and that's how they were writing to the file shares. They were just mounting them to their Macs when out in the field and they just had AFP open to the world. Yeah, people do a lot of things like that. We sorted out a medical place that was using public IPs for the sharing between the scanning system, remember? It was like mind-numbing that they were setting them up this way. Now, what saves them in that case is a lot of cable models block port 135 and things like that. It actually wasn't sharing it with the world and they didn't have a password on it, but still, it's an SMB password, yeah. Getting public IPs to do your SMB sharing and assigning public IPs to medical equipment, scanning imaging devices so I could share the stuff. It was just like, when VPN equals virtually private. My job exists because of stupid engineers. We unravel a lot of mysteries. People create things. I was on a consulting call this morning. They have new IT that took over and they contracted us and I have some future plans. So this is gonna be a sounding board for, we don't know why the guy did this. I wanna make sure he wasn't playing 40 chess when he set this up, but it looks dumb. I looked at him like, yeah, I don't know why he did that either. Okay, cool. We're gonna start fixing it. Any board medic, ARC alone, automate, regular encrypted backups, SMB shares to go directly to TrueNAS using second time machine required. Not really. I've not used Borg or RClone for that. I mean, I'm sure tools exist. Borg backup's popular. People seem to like it. It's a good open source. I'm saying I've heard it's a good open source, but it's not on my to-do list to use it. We don't use it commercially and don't plan to. So you're kinda, I mean, there's not people using it. You could probably find some documentation on it, but I don't, it's not something we engineer for people or would, too low on the list. Usually you want some commercial product and the biggest reason for it is you want some monitoring you can do of the backups because untested backups or unwatched backups, you don't know if they're happening and Synology Act to backup as much as we like it, we also can't push it commercially because we can't be the babysitter for it because it doesn't have a central dashboard for multi-tenancy. I don't know why it doesn't, not because I haven't asked the engineers, it's only because the engineers haven't gotten around to writing it. So it's just one of those things. Like there's all kinds of cool backup solutions out there, but without a way to centrally manage them, there's no way it doesn't scale for us to manage backups for many, many clients. It becomes labor intensive if I can't bring all that data in one place. Yeah. And we get asked about the Synology one all the time and it's just, I told one client, I'm putting a Synology, I put a Synology in so just to host the local copy of their backup and one of the employees is like, hey, I use Synologies at home and I know how to do active backup, can you set that up for me? I go, as long as you're the one monitoring it, I'll make you a user and sure. Yeah. And so we have one employee at a client, he manages his own backup and I told him, because he even asked that. He goes, why don't you guys just do that? I go, there's no central dashboard. We can't scale this to, if we were a one-off like within a single business, it makes perfect sense. But when you got 30, 40 different businesses that you're doing this on, there is no way to scale that. Yeah. And I don't doubt that Borg backup is a cool product and everything else and it's neat, but without a nice central dashboard, it's hard to scale it across clients. I mean, for your project, it may work and if you're working in internal IT and your job is in internal IT is to build the dashboard and manage it or a home lab that you manage, great. You can do it. But when your job is, I need a multi-tenancy system where from my office, we can have one central dashboard that sees into 70 different companies, now you have a different challenge and you have to meet that challenge to be able to do it. So, so Borg does have some type of central management based on another open source project. So, I mean, it can be pieced together. It sounds like right under green text in there, always check your backups. Probably. Do they support Windows too? Here's all the stuff for setting up here from standalone binary, do they have a Windows binary? Linux, Mac, FreeBSD, that kind of, we back, lover hate it, we get the backup Windows machines. So, those are the ones you should be backing up. Yeah. So, I mean, cool that it's got Linux support in there, but I don't know, I mean, I like the backups for XCPNG. So, the backups work good. So, let's go ahead and run this backup here for XCPNG. I like the backups for that, but it's the, it comes back to scale. We're managing our XCPNG solution. If, you know, we have to manage it centrally versus if you start managing everybody's, it becomes a different dynamic. Now, there's actually ways to do this XCPNG to do some of the management, but, you know, it's just all the things you have to consider is how scalable is it? And not everybody runs XCPNG, so that's an entire different challenge that pops up for this one. Oh, I gotta recompile this so this bug gets fixed because it's gonna be fixed in 53,000 years. They found the bug and I tweeted at them and they said, yes, they are completely aware and they fixed it already. Was it last week that we pointed this out? I don't know, I don't know. I can't remember. But what is the bug? Is it like telling us number of seconds and then telling us years? I don't know. I just know it's funny. Okay. In 103,000 years, 94,000 years ago, this is the estimated end time. Okay. Yeah. Years ago, no less. I know, it's got, it's really wrong. But it's not always wrong. I can't figure out why it's wrong sometimes and run. It's a fun bug. It's just great to watch. I just wanted to stay kind of. I like it. It's amazing. I know, I like these numbers. It's way more fun than watching how many seconds it takes for a backup to run. Yeah. And someone said there's another one called Copia Backup. Do use Marker Reflects. I do not. We have used it, did we use it before? We used to use one of those for Client Lord. What's that? Our landlord uses it. Yeah, our landlord uses it. One client, the one that went through a few, they're all kind of IT related, but they went through a few people and the one guy left and then came back later and I think left again. I'm trying to remember you, but without saying the name. Oh yeah. They used it, I believe. And then that's when they had that raid that died and I had to do backup restores from a USB drive and that was the best two weeks of my life. Yeah. I think it took less than a number of years instead of it was gonna take to run this backup. I was just curious as I left that up on the screen. 90%, 91, yeah. Yeah, I don't think Macrium's a bad product. It seems to work every time we've used it. Yeah. It's like a Cronus and all those other ones, they work. They generally make a pretty good product. I would call it better. I don't know, does Macrium have a dashboard option or is it still just a one off like the basic Acronus? I don't know. Yeah, I'm not really sure. Might be just basic Acronus, there we go. Yeah, if it's still the basic local only it's not bad. Like I said, it's just, it's lacking. It's lacking, but that's life. I should reset. Again, the scalability issue of, I gotta do this on a bunch of computers at a bunch of locations and I wanna be able to see it quick and not have to spend the entire day figuring out what didn't and did not backup. Yep. Unified switches can be hard to find especially the higher MPOE ones. Yeah, out of stock, I mean, things are getting better but out of stock was definitely a problem for a while with just everything. And their new model for the switches with the POE that the wattage is so low like the 16 light which is there now really the one they want you to use is only I think 45 or 55 Watts of POE power. Like let me pull that up. Yeah, or send me a link and I'll pull it up on her site. Yeah, I can actually. Yeah, I think you can share stuff. I can, let me find it really quick. Let me at least get the tab up ahead of time. Oh, and I'm working on this while C's doing that we'll throw this, okay, we'll throw a C's on a screen. Okay, I shared it but it didn't come up. Yeah, yeah, 45 Watts of total POE availability. And we found that out when we plugged in about four WiFi 6LRs and then the fifth one wouldn't light up. It didn't have the power for a fifth one despite having eight ports of POE. Yep, that's a challenge. And here's the wild thing is if you look at the switching because I just had to look at these. Now these have a little asterisk, 52 Watts of total POE. This is the eight port with four ports of POE. The asterisk is with an external 60 watt power adapter which it does come with. So not if you power it off POE. Yes, if you use the actual brick it comes with. Yep. And even their, this is a problem with all their Gen Two equipment. If we look at the Switch 24 POE, same similar problem, 95 Watts. So you have 18 ports, 16 ports of POE. But if you're trying to plug in a lot of APs it's gonna be a problem because I wanna say a U6LR pulls close to 10 Watts at times. And I think they're even their 16 big boy, Gen Two. Yeah, yeah, yeah. Yeah, look at the Rackmount 16, 42 Watts of POE availability. So this is finding the high wattage stuff. You have to buy all the pros and it gets expensive. And I think that's where some of the complaint comes into on the Unify pricing. It's not like the Gen One equipment where you had two different 24 ports. Here's your, all the ports are POE and you have a 250 watt option and a 500 watt option. Yeah, they're changing them up a bit. Yeah, I'm less than impressed. Yeah, they're still good. It is just something you have to be aware of on a larger deployment. And when you're something to really note, you have to look at the full camera wattage of what the camera uses in night mode because that IR bumps it up. So you can actually have switches that get overloaded at night, but don't during the day. Yeah, they used to do 150 watt one. They don't do that anymore. I don't think they sell it. They used to have the 128 port and that's what a client had. And I had to get two of the new eight port lights because he was using all eight ports of POE but he had the old 8150. Yeah. So I wanna do a video on this topic soon so we can log into one of these VMs. We're on H top so we can see this. So you can see that this has 16 cores attached to it. So there's the 16 cores running here. It's assigned 16. What not everybody knows though, without restarting the VM, you can actually remove the cores. You can take them out and you can also put them back. But that's not what I was gonna do the video about. I was gonna do the video about, hey, look, they're all back. What happens when you over provision CPUs, the good, the bad, the ugly, because that was my consulting gig this morning. We had a discussion about the over provisioning of things. When to do it, when not to do it. And the consulting gig was an example of absolutely don't do it this way. They actually were able to, the new person coming in was able to just reprovision the CPUs in a rational manner and get better performance out of existing hardware. So there's a lot of little nuances to it. So it's definitely fun being able to do it, but it's actually kind of cool the fact that you can scale CPUs dynamically just like you can memory. So like right now this has six gigs assigned to it. If we wanted to only assign two gigs to it, we could just do this and reassign the memory to it. There's actually a lot of, I did a whole video about how memory management works in XCPNG. There's a little bit of trickiness doing what it's doing now of, what they refer to as evacuating the memory and bringing us down to like a two gig system versus bringing it back up to be a six gig system actually goes faster. It's easier to balloon up than balloon down, but you can balloon these back and forth while they're running to be able to do all that. Do you guys use manager on manage switch security network manage switches? Pretty much we use a, manage switches used to be years ago a lot more expensive. So you could, you may not always have a managed switch. They're so cheap now from all brands, like the price of that of enterprise networking gear more specifically is managed switches has come down substantially over the last like 10 years. So pretty much they, we have managed switches in most clients, most all clients. I mean, a small office that just connects to the internet through a PF sense. Do they really need a managed switch? It's not, maybe not always the case, but only really managed switches. And if you plan to have VLANs and advanced configs in your networking, yeah. And we like to separate some things out on there. Does the VM pause during the Ram resize actually? No. So here's the council running H top and we see 16 processors available. So we're going to go here and we're going to make it have four gigs across center and you're going to watch it go from four, three, let me zoom in a little bit more here so you can see it. Enhance, enhance. And you'll watch it shrink the memory down. It takes a little bit of time to shrinking it, but the VM doesn't stop running or anything like that. Let's bring it down to two, enter. Enhance. There we, now we, you actually seen some CPU load because it actually kicks off a process to evacuate the memory. But that process is faster going the other way. So actually, what's the max I can set this one to? Oh, this can have up to 16 gigs. You have min max as you can set. So this can have a max of 16 gigs. So we'll balloon it up there. Enhance. Yeah. And now we're at 16. There we go. It goes up fast. Yeah, you can scale up fast, but it scales down slow. Need a video tutorial on ACL and gateway features for my new melanox. Nope, not going to do it. But if you feel like doing it there, Chris, awesome, go for it. You're going to find there's, there comes a point of diminishing returns. Like when you do some of the videos on something that's super technical like that, that's also niche, you won't get any views for one. And most of the people doing those prefer a technical write up. So they don't care about a video as much. That's one of the reasons I don't, like certain things like I don't cover BGP with PF sense. I don't see the demand on it. The couple of people who have covered it and made a video, there's no one really getting views on it because the demand, it's such a niche item. And usually people want a more technical write up than a video on it. That's one of the reasons I just don't spend a lot of time on it. Does the Windows driver support that? Windows will let you change memory. Windows won't let you change the... CPU. Right. Not that I know of. I don't know. I don't think you can do that. I think Windows doesn't support it. Probably not. So, can I just close that? I've lost the screen I was on. Whoops. Whoops. Mm-hmm. We actually see we got that new, so we have a new Windows server set up. Okay. Or the lab. Like we have one that is for some, because we separated the lab stuff and now we have the Windows server, so we can, where is it at? Guess I gotta restore it. I don't see it anymore. Oh, it's right. It was on the other AMD machine. Uh-oh. I backed it up before we took the AMD machine away. I did back it up. So, let me share the screen. We'll do a restore. We're gonna do it live. So, here's the backup restore. OTS lab server. March 9th, I backed it up today. And we'll put it on local storage. Well, no, that's kind of a waste to put it here. All right. Restore. Whoops. Incompatible. Oh, I did a thing that's dumb. Okay, I'm an idiot. I loaded it on Zen 8.3, not 8.2. So, I'll have to convert it. Well, I'll just have to upgrade our lab to 8.3 or something. I don't know, I'll figure that out. There's probably some other windows I can restore. I don't know, we have a new server, I don't. I haven't had to be in there looking at anything. This one's old. Oh, this one's not too old. March 2nd of 2022. That's gonna need some updates. Hey, look, it's gonna be done in 53,000 years. Yay, invalid date. Wow. That's how you know the bug is even better. We've gone beyond time. Heat death of the universe. And the heat death of the universe, this will be ready. Does Windows support, I know Windows supports memory management, but does Windows support processor changes? I just don't know. It may support it, I don't have a Windows system handy to test it, not, well, I don't have a lab one. I certainly have production systems I could do this on, but that seems terrible if it doesn't work. You cannot do VLANs if you don't have a switch that recognizes VLANs. There's not a way around that. Well, and that's the tricky part. So some unmanaged switches will still pass the VLAN tag and that's a gamble in itself on whether or not it strips the VLAN tag. Alternatively, you could get a small switch that does VLANs and then use that to just delegate VLANs out to your dumb switches. Yeah. We may be doing something like that. Yeah, with just our front counter area. Oh yeah, because it passes through that one, doesn't it? Yeah, we have, instead of buying a smart switch to manage a bunch of ports that were just meant for customer stuff to be plugged into, we have it going into a dumb switch, which then the uplink to the Unify switch from that dumb switch is tagged. So everything that goes in the dumb switch ends up on a VLAN. Got it. There are, their version aren't compatible really. I never tried this. So I was running the alpha version of XCB and G because I was testing that system and I backed it up while it was on the alpha version. I bet I can just lie and tell it it's the old version. I don't think there's any difference in there. But my, I'm running production 8.2 and I was testing on alpha 8.3. So this is something I did. This is not something you're likely to run into. It was me testing bleeding edge stuff. So if I have an old backup, I can't restore it to a system running 8.3. No, you have it backwards. I backed up from 8.3 and tried to restore it to 8.2. You can take an 8.2 backup and restore it to 8.3. That's how everything got on 8.3. I restored all the backups to 8.3. So you're in reverse. You can always go from an old version to a newer version, but I'm trying to go from a alpha newer version to the stable current version. So that's why I got that error. Windows Server can support full CPU memory hot plugging if you enable NUMA, not just balloon but virtual limited resizing. Well, it's a matter of does Windows support it not? It's, I know that if the hypervisor supports it, which it does dynamic CPU allocation, it's whether or not Windows supports it. Now the warning though is gonna come in the important warning is don't, if you over provision CPUs, this is one of the things that was happening at this. I think this is one of the problems they were running into at the other place. If you cross the NUMA boundary, so let's say I have two processors and both processors have 24 cores. And I would like to assign more cores than a single CPU has because why not? There's two. Well, when you cross NUMA boundaries, you now create a different problem because you now lose some of the benefits you have of the caching because now you're inside of two different processors and conflating things. So you don't wanna do that. VMware uses NUMA and there's, you can use NUMA in here too because let me go back to pharaonics. You control that under here, you say topology. So I can say it's one socket with that or two sockets. So I can change it to where it's using NUMA versus just one socket with 16 cores. I can say two sockets with eight cores per socket for or 16 sockets with one core per socket. You can do that too. I don't know why I would do this, but let's do it. So now we've changed the profile this dramatically. I wonder if it's gonna complain about 16 sockets. You got a 16 socket CPU here. Why not? Oh, did you set it to 16 cores one per socket? Yeah, that seems dumb, but why not? I don't, does it affect anything in reality though other than how the VM recognizes it? Yeah, probably not much. Wonder, I mean, does it even tell me that? I don't know. I mean, it would probably break some windows licensing. Yeah, that's the bigger one. Yeah, 16 CPU. So hold on, I wanna make sure that I'm thinking right. So let's shut it down. Cause I don't think I can switch the topology while it's live. I could probably try it. I don't know. We'll see what a default behavior topology. Well, we'll forcibly set it one socket 16. There we go. By the way, if you set this and then you try to change the CPUs, you'll notice that you're forcing a limit and you'll cause a different problem. Like when you try to change it, it'll say you can't change it. And the reason why is cause you've set different limits over here. So be mindful of that. People who toy with XCPNG or any of these things. I don't think 16 sockets, in other words, existed in a while. I don't know. I haven't seen anything past a four. I know, you remember those old TYAN, tie-in, how do you pronounce it? They made four core boards back in the day. And I don't even think, I think that's since went away. It displays the same. The CPU still says 16, so. Yeah, but now it's one CPU with 16 cores. Yeah. So whatever. Your Windows licensing is good again. Yeah. Now, Windows has a problem with sockets and cores. There's some limits in Windows. I figure what they are, Windows has way different limitations in Linux and not just licensing limitations. Yeah. And it's like dependent on what version you run. If you run home, you can't go past X cores, but Pro can go up to Y. Yep. You were saying the core threads per CPU because it's one big feature missing from Poxmods that's hyper-threading. Oh yeah, you can define all that in XCPNG. They give you the ability to define that. They give you the CPU mask, weight, cap, like you can, what you can do is you can create weighted averages so you can favor VMs, like if there's a resource conflict who wins by default, if you have them all the same weight, they all win simultaneously. And you end up with a CPU stealing is what it's called. Stealing is the term when you over provision CPUs where there's a conflict. So like I've got how many cores are available on our system? 24. I have 24 cores available. If I assign two VMs, 24 cores, and they both all ask for 24 cores simultaneously, you end up with a CPU stealing going on. You can find some articles and if you type in what is CPU stealing, you'll find some write-ups on there. It's fine, it's not bad, but at some point it's where your performance bottleneck is for doing that. You basically have both VMs fighting over the same processing time. Yeah. Compute time, whatever you'd call that. Yeah, the Dell R9, we don't see many of them, but you write the Dell R9 30s or four. We just don't see many of them. I meant more like, because I was sure there were probably some Dell or Lenovo servers out there that had it, but I don't see much in like, let me just buy this board and put it in a server with four cores like we used to. I'm actually, I'm going to new egg now, because I'm curious, because they had the one years ago. And right here, man, here's the... Never mind. Super Michael. Why is it a weird shape? I don't like that. Well, where are you going to stick four processors? No, I mean like make it a square. Oh, you want to make it square? Yeah. Hey, guess what? Guess what company's still making them? There it is. Oh, these are Optoron. Yeah, those are old. How, what Optoron? Like the original Optoron? What socket is that? It Optoron's awful, so no. But what socket Optoron? Cause they kept the Optoron line for a while. Like is that the old quad 127F, 1207F? Okay, that's the, those were the newer Optorons. I'm like, did we just find the old board that was like the old socket 940 from a decade ago? But this one should make you happy, Steve, it's square. It is square. Is that an option and new way to search for four, four sockets? Yeah. CPU socket, quad. Quad 940, that's the old, old Optoron from like 2005. Yeah. I don't even think that's a quad board pictured there for 300 bucks. Yeah, there's not too many quad ones available. No. I mean, you could put four of those Optoron, that generation of Optorons, you could get your house. Yeah, you can definitely, those were so energy efficient. Yeah, they were. Hyper-V, we don't really use Hyper-V, but yeah, there's, I mean, Hyper-V has some dynamic memory options and things like that too. So that's definitely something you can do with Hyper-V. But back to, what was I trying to do over here? I already forgot. Oh, some of CPU stealing. Yeah, created in 1969. Started in 53,133 years. Yeah. I gotta recompile it. They do have an update to the master branch. So we can recompile the software and do that. Oh, I was restoring windows, that finished. I thought, yeah, I was gonna say you did that a little while ago. Yeah. Let me find windows and add the tag YouTube to it, which shows up my list properly. There we go. Actually, I like that I call it windows. That was my tag on this old one, this old lab base. This lab base is old, but it's good enough to do what we wanna do and play with the CPU sockets on it. Eight socket boards in China, that I believe, but they have a pile of old CPUs or something. That's it, what CPUs? Like, it's cool that an eight socket board exists, but if I'm slapping eight ARM processors on it, I'm not impressed. Yeah, so we can pull up the task manager. There is a chance that if you change processors, task manager will not display it until a reboot. Well, what I'm gonna do here, I'm gonna open another window to change the memory. That way, I don't have to switch back and forth. I can leave that pulled up and then I'm gonna go to the lab, go to YouTube, and we are going to change the memory in it to four gigs. And it's doing a thing. Yeah, available memory is going down. It worked, so memory's no problem in there. I don't like how that worked. Yeah, so now this machine, and you guys seen it in real time, now only has four gigs, so we can bring it back to eight. But all it did was just fill up that memory space in the... Yeah. I don't like that. I don't know how good this works. Now it's back up. I don't know. I think, I feel like Windows is so fragile that we didn't do this, but what happens if we do this? Itch. The MLAX feature. Hold on. No. I have to first set this. You have to set CP, you have to set the limits. So we'll shut it down. Did you have it at an eight, eight limit or? I might have had it at an eight, eight limit. So let's change the limits to like... One, one, eight? Yeah, we'll do four, 16. Sure. Single core, man. Come on. Oops. Also, crap, I gotta shut this down to the topology. I gotta change it. I'll let it boot. I don't wanna crash Windows. Windows is so fragile. I always feel that way about it, like... Task Manager sucks, blame Dave the plumber. Yes. I like Dave's garage. That's a great YouTube channel. I've talked to him. Shutting down, fixing topologies. Who knows why Tom was playing with the topologies? Probably for this exact reason. There we go. Now we leave it default behavior, so now it should just dynamically assign it and not force a certain behavior. Go, Windows, go. Risk servers, the problem with risk servers is they're pretty new, pretty different. So it's gonna take a little while for support to ramp up for it. I think they're gonna be... Risk offers a lot of advantages and I think they're gonna be popular. I think it's just gonna be a little while before they're popular. But it's cool that it's out there. All right, now we can go. Big core VMLux feature. It might be limited to server OS. Yeah, maybe. Because does this have... Yeah, they've got the drivers installed. Maybe it's gotta... Maybe it has to have that extra things installed for that to work. What if you try going... What if you try going down? Maybe we can go down but not up. Maybe. Well, what's interesting is it set it, so what we have to do is reboot it. And when it reboots, it'll have eight. But yeah, we'll try that. We'll reboot it and we'll see if it has eight and then we'll try and put it down. We're all learning together here, folks. Yeah, I've seen the management agent not detected and that's what I'm wondering if we have to load the management agent on there. The problem is I have the Windows update Citrix one. If I load the other one, it will puke. So I can't load... You can't have the Citrix one and the other one. No, no. And it's hard to fix it and change it. Well, I could. So now we have eight there. Go back to the council. I'm pretty sure it's gonna show eight in here. Maybe. Maybe. Task manager's weird. Yeah. Now it'll work. So... Eight. All right. So we have to just load the management agent. So let's see. What we have to do to load the management agent is stop, stop. We gotta go to the advanced here. Windows makes this hard. Can anyone tell him a Linux guy? There we go. Turn this off. Start it. Remove, then load. There we go. So then we'll load the management agent and we'll see if we can swap CPUs on the fly with Windows. Maybe. Maybe. Where do I put the disk? Yeah, it's fast enough there. LTS, welcome. But now we gotta do... Oh, Windows, go. God, I hate how slow that is. There we go. Doesn't it... Aren't the Citrix drivers here? You can tell how often I do this. I don't... All right, well. All right, I don't see him here. So let's... Can't update. I don't care. We're doing this live. We're letting it... Oh, no internet. What? Well, I'm not... I'm an idiot. You did not assign it a network adapter? I did not assign a network adapter. So I don't have to go any further. First, I'm gonna type Linux commands in Windows and be an idiot. But now we can just... It's at this way. There we go. Missing PV drivers. That's why. Derp. All right. Force shut down the VM? Yes, I would. Screw you, VM. There we go. Grab. Now, once I turned off the other Windows update thing, I think it uninstalled him for me. I think it does that. Sometimes, then sometimes it doesn't and they're just stuck in there and that's your life now. I always leave it by default. I haven't really had to change the video RAM. I've had to turn it to 16 sometimes at high res for screen connector. Oh, yeah. Yeah, I guess if you bump it to a higher res, that's a thing. And even then, I only go to 16. I actually had responded to that one. Sometimes I move it to 16 because screen connect uses video RAM and when I want 1080 video because I don't wanna be looking at a little 800 by 600 stretched. Yeah. So see, I'm gonna get that. I don't care about this. Citrix, hopefully this will get me where I wanna be. Did Citrix paywall all this? I thought these were old. Oh no, these aren't newer. Yeah. I remember trying to do this before and I couldn't find anything on the newer ones. Where's the, just download link. ZenBus, ZenViv. Yeah, I think you're gonna run into the problem I was running into. Version history, where's click here to download? I don't think it's there. I don't know if it's paywall or behind a login or something, but I remember running into this. You can't get past like a certain version. Yeah. Well, we can load the Windows XP and G. So Windows Management Agent. You can load this one too. There's the links. Oh, they don't have the link here. Back. There they are. There's a GitHub for it. Cool. Before installing. Oh, trusted vendor. Ah, I don't care about trust funds. Pre-release. Where's just the download? Oh. It comes hard sometimes. Yeah. I'm missing where it is. Oh, it's under assets probably. Hey, look, there it is. That's the one from 2018. Do we have one newer? Oh, wow. Hold on. See if there's anything newer in the forum post here. Assets. That's one's from 2019. I think I want that set up. So we'll extract this here. Where's extract all? There we go. Install. Let's see what explodes. Everything. Copying new files. Oh, Citrix tool uninstall herself. Leads various things behind. Whoops. Well, let's find out how wrong this is if we do it this way. I went through it. Next thing, yes, my way to success or disaster. Does it reboot? Does it reboot? Sorry, how works task and Synology? I don't understand. You can use tasks. Like I have it set up to do a web hook to activate Home Assistant. So Home Assistant does a thing to turn my lights on. So, yeah, 8.22 is the latest. Yeah, he doesn't. And I don't recommend loading it. I use, by the way, I tell people use the windows. Just turn on the windows update. Use the Citrix drivers. Like that's my answer as to what to do. Hey, look, it didn't die. Yeah, but can we change things now? Well, let's go to the device manager and see what drivers are loaded. So display, change processor. It's all we're really here to do. So it says not BMWax feature. So, oh well. Yeah, I wonder if it is something that's just missing in the OS. Could be. I got nothing. Maybe the Citrix ones to do it, but I don't feel like logging in in Citrix to set that up. Like I'm not gonna go through the trouble. I don't care enough, honestly. So there's a management agent that says it's there. 8.2. So it's installed, but whatever. Not too worried about it. Just not that worried about it. We came, we saw, we poked. This was completely, this machine's so old it's gonna have a ton of windows updates that I don't feel like doing. This is such an old lab instance. Just run Linux. We only run windows because we have to. Matter of fact, there was an article about how the Steam Deck has really pushed gaming into the Linux world even more because people are sick of windows and the problems that associate it. Yeah, I don't know if I'd like, I don't seem to have windows problems with my games. I'm running games from 20 years ago, but I will say the ability to play them portably and have those switch-like functions of, let me just hit that power button and pause my game. I love that. I'm using Cat6 Slim cables. I left links. This is a common question on each of those videos. There is a kit.co link in the description of every video that links to a store where we list all the parts we buy and then we have subcategories and specifically in, I'm assuming you're talking about the rack video, there is a parts list for the rack video that has the rack, the cases, including the exact, a link to the cables we're using inside of there. So there's this Slim Run, is it a monoprice? Monoprice, yep. There's a link into the exact ones we're using. I actually think, go ahead and answer that while I find it. The task for the camera of the service. It's the task you run inside the DVA model. So if you want the DVA to do face recognition, you're limited into how many tasks you can have for face recognition, for example. So I'll pull up mine real quick. So here is the deep video analytics. So I have one called Who's on My Porch and I have one called Watch the Shed, that's it. You can only have two. If I hit add, it won't let me add more of them. If there are any, actually I take that back. I'll let you add them. It will only let you enable a couple tasks at a time. I like the names of your tasks. Yeah. Who's on my porch, Watch the Shed. Who's on my porch, Watch the Shed. So you can add more tasks but once you start enabling tasks, that's it. And with face recognition, you can create task and we'll say like garage for face. And then if I try to add another task that I've reached a limit, you can either have one face task or two deep video analytics tasks. That's it. I covered that in more detail. I have a dedicated video about the... Per NVR. Yeah. It's the DVA. I did a review it on my channel and it's a nice system but there's definitely limitations. You have to be fine with those limitations if you want to use it. Or you get the bigger model. Yeah. Or buy the bigger model which has different limitations. So I found my multi-color Slim Runs and these actually do say cat 6A on them. So they should logically do 10 gig. Yep. And this one's loose. Quick question. Home network, symmetric gig to a net gig, 6100 and a 24 port edge switch worth upgrading to switch to 10 gigs router for uplink. You all firewall rolls to the net gate. I mean, if you have a symmetric gig, 10 gig doesn't get you more than symmetric gig. So I don't completely understand the question. Like if you, one gig routing is definitely capable on the 6100. So I don't know what you have now but 6100 is a good one for that. I wouldn't worry, do you have anything on the LAN like some kind of file server or things where multiple people connect and that could exceed one gig? So that way no one person could saturate it. That would be the biggest benefit of it. Yeah. Model price Slim Run are best. Color code, KVM management, the data ones. Yeah, I really like those Slim Run tables. They're just nice. Yeah. That's why I did most of the, all the patch cables in the rack are the actual runs to the machines. I just got a bunch of colored ones from QVS and cut them up to make the runs down. Yeah. But like these are those Slim Runs and they just have a nice look to them. They got the nice little press the button unlock them. I like the, I like the snag guard on them other than sometimes it gets stuck under the clip. Yeah. You have to flip them over, but they work great. Yeah. I think my favorite snag guard still, but they're less pretty cause they're not slim. My favorite snag guard are still these ones from the ones we used to get from QVS. Oh yeah. Yeah, those work. Cause they don't ever get caught under the clip. And if you've ever been at an odd angle reaching behind something and why can't I get this out? And then you find out that, oh yeah, the snag guard is caught under the clip. Good luck. The, yeah. Dak, the Dak and QSAPP cables are thicker, especially these like, these are 25 gig cables over here at this end, but this bad boy, once you go to the longer cable, this is just a 10 gig cable, but it's a longer one that wraps around the case a little bit further. You can see how thick they are. So once you go to those, I think it's once they go over like, what's that? What is that one? Yeah. That goes down to the, I feel like it goes down to the 45 drive server. Why is that so thick? That's not even 25 gig. Cause it's one of the thick boys because we had those ones already. They're the ones that can carry for greater distances. I guess, like it shouldn't be that long. It's only going to the bottom of the rack and I thought I bought all new ones are roughly the same size. Nope, it's the thick boy for there. Let me, I think that one was just thick to be thick. Yeah. Refresh task manager, open it back up, we'll show memory difference. Okay. Interesting. So pull up task performance. So we currently have, there you go. Memories available, five gigal butts of Ram. Watch it go down. Or do we have to close and open it? Hold on. I think you got to close and open it. Yeah. So we're at four gig now. Back to task manager. No, still says eight gig. Huh. Yeah. Someone said it's because it basically fills it up, gives it back to the hypervisor, but then lets you come back and like pull it back later. Modified. I don't know if they're right, but it made sense. Yeah. Like, I don't know that you're right, but that's sounds good. Yeah. Can windows survive with two gigs of Ram? Then what happens? I don't know. Still shows it. Well, what if we reboot it? Then what does it show? So currently we're showing eight gigs to reboot it for fun and science. Thick with two C's, man. Yeah. I just responded to that. Like most three feet or less cable runs are patch cables. So I mean, you're not really bundling them together in a way that I would imagine you would get much crosstalk. This is an old myth. I actually think this is addressed me and one of the people who's on the IEEE committee, we did a video together, Dan Barraris, they're on there. And there's this weird myth that got persisted for a long time because of someone confidently posting a forum post that you can't use a short cable because it's so short the wires don't cross over each other enough times to eliminate crosstalk. And I'm like, that's not true. And it's not how it works. This is not how it works. Just because someone confidently said this in a forum, but it was a really old forum post. And apparently this person had enough authority that this myth was perpetrated for a while. Seeing now what I want to know from a physical perspective, how are you getting that many three foot runs bundled together like that? They're not that long. Right. Like I have someone put her- It wasn't those run with two gigs of RAM? What? Yeah, Windows will run with two gigs of RAM. Not great, but it will. Okay. It'll walk with two gigs of RAM. I don't know. I think the minimum for Windows 10 is like 512. Okay. It's booting. So there's that. Ram, one gigabyte for 32 bit or two gigabytes for 64 bit is the minimum system requirement per Microsoft. It's shut down. So we'll pump it to four. I don't know why it's shut down. Maybe related to that. Let's find out. Or do we just break Windows? I don't know. We could have broke it with the other stuff we did. But yeah, per Windows supposedly it's two gigabytes. But not this install. We did something. Now we've got to please wait. I'm gonna go with, we may have just broke this because I don't have the patience to please wait. Oh, that's all right. We're reaching the end of the vlogs. I got a few things to go do. Your swap file will be really active at two gig. That seems likely. Yeah, your drive is gonna be banging away and I don't think it would have changed the page file. So we would have enough page file to still do stuff. It would be wild if you spun it up with two gigs from the beginning because I don't know if Windows still has this, but the old default used to be to create a page file equal to RAM. Yeah. It's doing something. I just don't know what. The BloomDrive and other services are both trying to gobble up RAM and none of them can complete this attempt. Probably. I don't know if you're right, but like Steve said earlier, it sounds right. Yeah. And having no other answer, we're going with your answer. How's it like from, I think that was a comment and it's always sunny about like burning trash and then it goes up and it makes stars and they're like, that doesn't sound right, but I don't know enough to argue that. Yep. Did we not have any questions that were emailed in? Oh, thank you for reminding me. I thought about it at the beginning and then you talked about a couple of things and then I was like, actually, I thought we had zero. We had zero questions emailed in, so. Convenient. Yeah. So we did check. And I'm shocked there were no questions. Send more questions. Yeah, send more questions. They will throw the banner up here because we're going to wind this down here, but it's email questions to vlog Thursday at loresystems.com. Yeah. As I made my show notes at the beginning and I was like, huh, I checked that because we do this on the Home Lab show too. And they actually go, well, there's a sub folder for feedback. I got like a feedback and I got two folders. One's for Home Lab, one's for Vlog Thursday. And Home Lab actually has more, but Vlog Thursday does not. So no one about it emailing me. Email, what's next, voicemail? Yes, please fax us your questions. Please fax us your questions. Nail mail it. Carrier pigeons. Yep. I think we just flat out broke windows so I'm going to just. You might just let it ride for a bit. I know. Or doing that may fix. Sometimes doing that fixes it. We're just going to remove it. That's. You don't need stamps if you use a carrier pigeon. Yes. All right. Any more final questions here? Because I'm going to go watch Last of Us because I haven't watched it yet. Not the last episode at least. Oh, I love the story of the character that, oh, what's his name from Parks and, Nick Offerman, they said that they had to get on him. There's a story going around that they have to get on him because he keeps using all the tools to fix things around the set. Yes, yes. Nick Offerman brought Nick Offerman to the set. That was great. Yeah. Nick Offerman is awesome. Windows is drunk, yes. We're going to have Veronica probably on again for the homelab show. I don't know if she's available in the afternoons to do Vlog Thursday, so. And we mostly ramble on about stuff here. We don't have a real structure to this. We come on here and we kind of babble on and answer a lot of people's questions and things like that. Break some Windows devices and. We broke Windows today. It's not a deaf vlog without a bug. Yup, so. All right, well, I'll wind it down here. I'm going to go do some stuff. As always, email questions to Vlog Thursday. We like answering these questions on the vlog. So, oh, someone has one more question here. My last question is cabling. Why use 25 gig cat eight copper when you have fiber that can scale to a heart gate? Because it's easier to find people who can terminate cat eight copper. Yeah, I mean, we don't really use cat eight that often. You don't see it with the exception of weird people who say, oh, I recommend using it in my homelab. What have fun, dude, the stuff's expensive. We've seen some people who, yeah, I just built a new house, so I wired it for cat seven. And I'm like, I guess I get it if you're going to be there in the next 20 years. Cool. Yeah, for the most part, like what we use in ours is these, these are 25 gig DAC. So the 25 gig is just all DAC in here. DAC is better than fiber for things like this. You know, why use a bunch of fiber connectors when you have... It costs less than fiber connectors because the cost of that, that cable is only slightly more than the fiber that would go between them, but I didn't have to buy two transceivers. Right. The two transceivers well exceed the cost of, I think one transceiver in many cases starts exceeding the cost of a DAC cable. Yeah, and I have a whole video called put some DAC in your rack, and it just sounded great. Yeah, I like that. The only, we have three fiber runs in the office. So if you go up to the picture at the top. Oh yeah, yeah, let me find that one real quick. I know I highlighted it in a video, which I don't know if it's just clear here, but yeah, here it is, I mean, I can read it. Wait, why the... Yeah, we have three. Yeah, one goes to our front lab area. One goes to the back switch that supplies the offices where the, if you guys remember videos from long ago where the rack used to be, I couldn't move, could back up. I didn't want to, I could have done anything. I didn't want to move all those connections. So we just had a fiber that ran back there already and then made that go to a switch there. And then the green one is actually OM5 fiber that is running 25 gig to the lab rack. Right. And I went OM5 because, well, it's gonna be cool when we get 40 gig stuff later. Yes. Well, no, we're gonna go to 100 gig. So we're gonna go 25 and you get the four by 24. I don't know that it'll do 100 gig. I think when I looked at the OM5, I would have to look at the OM5 standard again. Hmm. Yeah, we'll figure it out. Yeah. We'll just start bonding more of these together. Quick question is NFS turned over to site to site tunnel. No, don't do that. I mean, there's probably somebody doing it, but I don't recommend it. Do not run NFS over site to site. I mean, it's just, there's better ways to do backups. NFS is over a site to site tunnel. NFS and SMB aren't favorable to loss, like if there's a drop, so I wouldn't do it. One of those, probably not the best idea. Like if you're transferring over a VPN, definitely you wanna be doing that with a different protocol that's more internet friendly than NFS. Transfer things over SSH or, I don't know, there's better ways. SSH is good. S3, bucket transfers. So that's one of them too that can work. Oh, I think I've slowed down my stream. What'd you do? I may be downloading something. Whoops. Is it season eight? You're not supposed to admit that. No, I legitimately, yeah, whatever. It's too late for that. I bought it on iTunes. I bought it on iTunes. Whatever. All right, we'll almost stop this here. Thanks everyone for joining. Is my voice choppy? Or just a little bit. Your voice is actually all right. Your movements look like a five FPS security cam. Don't click the button while you're doing, I guess I legitimately pay for HBO because they're a wonderful company. They gave us John Oliver. Well, they host John Oliver. Yeah. He's entertaining at least. And they put Nick Offerman on there. So there's that. And doesn't it have Adam Driver as well? Not in last of us, no. Who else am I thinking of? I don't know. I don't know. We'll think about that later. So, oh, last question. What ports open up for clients connecting Windows servers? None. None. You don't open up ports on Windows servers to the internet. None. Not unless you want to have a bad day. Yeah, unless you want to have a bad day. You do not open ports to these things. That's just. No, you shouldn't. Put it behind a VPN. Everything goes behind a VPN. I have to admit, I love the clients who are just 100% on board with that now. What ports do you have open? VPN. Not internet internal, like internal. I mean, only what's needed, nothing more. Principles of least privilege. Like if they have something they need open, if not, why open it? Like workstations generally don't need anything open. And not even all workstations need RDP open, so. Yeah. I guess I don't understand the question, I guess, Fooley. I mean, whatever ports are needed on the server to do the thing you need to do. For serenology surveillance, that's usually we try to do it if you're cool with it. Most people are cool with it being a little slow. They're not watching the live stream all day. So we usually try to keep them on Quick Connect. That way we don't have to open ports. Cause you can do Quick Connect with not opening ports. If they want a faster connection, and this is why you segment your serenology to a separate network, you can do an open ports on there. So if you open a port on your serenology, you can. 5001. Yeah, not 5000. Yeah. 5000 is HTTP, 5000 is HTTPS. Yeah. Or I'm sorry, 5000 is HTTP, 5001 is HTTPS. Yeah, it's very clear. So if I have an AD setup, so do clients need to contact ADS? Yeah, I feel like three. But yeah, you don't need to, like your server has, like if you set up a Windows server and you set it up, it has the right ports open for you. It does that automatically. So you don't have to do anything special. And almost all AD stuff is mostly DNS. Right. So you shouldn't have to tamper with the ports. SMB automatically gets opened when you set up a file share. Right. So you set up Windows, and Windows has certain ports that are actually. When you set up AD, SMB gets turned on because of the net logon file. Therefore, SMB is now automatically open anyway. Right. So yeah, there's nothing special. You have to have net logon in SysBall. Yep. Yep, to get all the AD stuff working, so. But all right, hopefully that cleared any of that up. Send questions to vlogthirstailhornsist.com and we're more than happy to bring those questions next week. I think next week, I gotta see if I'm going to Chicago for a thing. I think I'm coming back on Thursday. So I don't know what time I'm doing vlog Thursday. I'll have to figure that out. One, two, three, four. Yeah, and then I won't be here next month. Like on the sixth, so four weeks from now. Four weeks, four weeks. Fun stuff. All right, a week after that, I could do it from like the woods in Florida. I don't know. That's true too. We'll figure it out. Thanks everyone for joining. Have a great time. See you next week. Or whenever time next week is.