 Live from Las Vegas, Nevada. Extracting the signal from the noise. It's theCUBE covering IBM Edge 2015, brought to you by IBM. Hi, buddy, we're back. This is Dave Vellante, and this is theCUBE. We're live at IBM Edge 2015. This is our second day here at Edge. Jeff Barber's here. He's the DS8000 business line executive. Also has a responsibility for tape, which believe it or not folks, we're going to talk about tape and the SAM products. Jeff, welcome to theCUBE. Thanks for coming on. Dave, thank you for having me. I appreciate the time. So yeah, I'm really excited to be talking about this stuff because people, you know, it's kind of the, it's not the buzz of the conference, but it's certainly the workhorses of a lot of data centers. So give us the update. What's going on in DS8000 land? And then I'm going to talk about tape. Well, Dave, I think you hit the nail on the head. No two ways about it. You know, people think DS8000, they are not thinking sexy. But when you start track and how much transition, transaction work load runs through these things around the globe, it is a huge percentage of the world's data processing because all the banks and all the insurance companies live for it. And what we really are bringing to the market this quarter is some new innovation, particularly designed at ZOS. When I tell you what we're doing from a hardware standpoint, you're going to yawn. We're introducing 16 gigabit per second interfaces on Ficon, so that'll be supported in ZHPF and people kind of go, all right, great. We kind of knew that was coming. And that's just the underpinnings of the good stuff. I think you know that in the DS8000, we spend an awful lot of time making sure we are well integrated into ZOS and making well certain that we are integrated into the subsystems of ZOS, so DB2, those kinds of things. So what we've introduced in addition to the 16 gigabit cards, we spend a lot of time with the DB2 engineers trying to figure out how can we really accelerate the performance of DB2, not just in online, right? Because online is a big deal, but it really makes a huge difference in analytics. I think you know when people use analytics on the mainframe, a very, very large portion of that workload is done on DB2. So if we can accelerate the analysis, the queries, they can get insight much more rapidly in addition to just having faster online. So we've introduced a couple of new capabilities for DB2. The net of those capabilities are you're going to be able to get 60% improvement in megabytes per second transfer and you're going to get a 60% reduction in latency. So the OLTPs benefit and the analytic queries benefit a great deal. So it's a real, real benefit to our customers on the high end. Is that a compression play or? It's not a compression play, they're just some things that we've done with the channel command words in Z that just allow it to go that much faster when you get into 16 gigabit environments. And we don't need to go any further than the channel command words. You can talk CCW. So Mike, let's talk business questions. Are these exclusive to DS8000 or can other array manufacturers take advantage of them? Right now they're exclusive. We have licensing agreements with the other disc manufacturers. We provide them the interface. They have to then go and engineer how they're going to work with the interface and eventually they'll get there. Okay, so that's interesting. So you're opening up the innovation to your competitors. Yes, sir, we've always done that. As a rule, we've wanted to make sure that new capabilities that show up in Z are propagated into the Z ecosystem. You know, it's interesting. Customers would not be real thrilled if we said, hey, guess what, you like Z and you run all of those huge database oriented applications and you have to buy our disc. We wouldn't be real popular. So Oracle's doing. We're not Oracle. It might be a good strategy, but we're not Oracle. Well, that's interesting. That's good that you're doing that. I think you're right. Customers, they want choice and you got to compete not based on locking out. But there's lock in, but then there's lock out. Lock out is just as bad. Okay, and Flash in DS8000 is sort of extending the performance curve. Is that the correct way to think about it? Well, the thing about Flash is, and I said this yesterday in the great big storage review, Flash is a media, right? And people start to think of Flash as a product. Flash is a media just like a hard drives a media or like tapes media. Now, the thing that's interesting, if you think you need the performance of Flash, you then have to ask the question, how many people are just gonna take Flash cards and attach them to a server? You actually want management capability. You want to be able to take snaps of them. Some people that have operationalized workload that runs on Flash, wants the ability to do remote mirroring, two site, three site, four site. So the long story short is, we take Flash, we put it directly behind the DS8000, for those high performance environments, whether they be ZOS or Power or Systemi, they take great advantage of Flash. Now, that impacts the old TPs very, very positively. The other thing that we've got to make sure that we do is we make sure that we spread the word that if you want to run analytics on any of those platforms, you ought to consider having Flash on the back end of your DS8000 because God knows, DS8000 is probably the preeminent tier one controller in the industry. So, it's the same concept of putting it behind any other controller, a la the Flash System v9000. Okay, let's talk about tape. I know you're super busy and you got to roll. So, we're going to try to get you out of here as soon as possible. But I want to talk about tape. Take your time, take your time. All right, good. So, Flap is a term coined, it's a funny term when I first heard it, it was coined by David Floyer, who's our CTO, Flash and Tape. IBM is one of a few tape manufacturers left. And obviously, you position tape not as the high performance device, you position it as sort of the data retention, long-term data retention, right? It's far cheaper than anything else, right? So, why don't you give me the bumper sticker on tape and then I really want to get into it and brainstorm a little bit. Well, Dave, I'll tell you, I think you hit the nail on the head. You know, they're controllers and then they're media behind controllers. And we know what the price point of a flash card is and we know what the price point of a SAS drive is and the near-line SAS drive and there's a hierarchy and you keep going down that hierarchy and you get the tape. And the one thing I can guarantee you today, it's the lowest price media that is really available for performance and high volume data transacting. So, it's got a place as many times as people have said it's dead, it keeps chucking along and people keep buying more of it. So, we think it's got a very real place in the ILM hierarchy and there are a lot of people that are, you know, CSPs that have lots and lots of disks that are realizing this is a way to save on costs and provide their customers lower cost options for storage. All right, I want to brainstorm with you a little bit and I want to put forth the premise that tape is actually for the use case of long-term retention can be a higher performance, higher performance solution and lower cost than spinning disk. And let me tell you why we think that. So, the caveat here is got to be for large. Am I going to have to pay you for this? No, it's all, we're open source here in the queue. So, the caveat is got to be for large object files but you can take small, you know, files and concatenate them into large objects. So, here's the premise. Would you agree with me that the notion of high performance disk is kind of an oxymoron, right? This is really with flash coming in, flash is the high performance. So, as a result, manufacturers aren't spending money, investing money on making disk heads more performant. There's no business case to do that. You can't spin the disk any faster and you can't move the actuator any faster. So, the effective bandwidth, the internal bandwidth of a spinning disk as capacity doubles every 18 months declines. It maybe grows, the performance maybe grows single digits, six, seven percent, but capacity doubles. So, performance goes to hell in a handbasket. So, the internal bandwidth is like sucking data out of a straw. With tape, that's not the case. You could, tape heads aren't hermetically sealed. You can double them up, right? You can, so the bandwidth as capacity, the time it takes to scan a five, what's the capacity of a tape cartridge today? Oh, God, well, we're going to be pushing up above 10. Okay. So, let's take a five, we're talking terabytes here. Five terabyte tape cartridges. However long it takes to scan a five terabyte tape cartridge today, it's going to be the same 10 years from now. Would you agree with that? Roughly comparable. Five years from now. Roughly. How long is it going to take? What's disk drive capacities today? Are what, six terabytes? The nearline sass, yep. How long is it going to take to rebuild a 30 terabyte disk drive? Forever. Forever. Behind a subsystem forever. Because the internal bandwidth of that, so we just showed that tape, as capacity increases, performance stays the same. Right, disk, it's the opposite. Okay. Now, here's the rest, this is kind of a long-winded theory. I am going to have to pay you for this. I get this feeling you're reaching on my pocket. Listen, this is like gold for you guys. And if you get this, you're going to sell a lot of tape. Now, the metadata that exists on where to locate tape data, where is it? It's stuck inside the cartridge, which is a sequential medium. So that's part of the reason why tape is so slow. But what we've seen some people do, some clients, is they're elevating the metadata to a flash layer. Right. And then, as we all know, time to first byte on tape is very slower than the random access medium, but a time to last byte, because of the internal bandwidth of disk can be faster. So what smart people are doing is they're writing algorithms to say, okay, if the fifth byte in the sequence can be serviced before that, I'll service it now and grab it, because I got the metadata, it's going to tell me that, and then I'll reorganize it at the back end. Exactly. Okay. So in that instance, we believe that flake is not only far less expensive, but can be higher performance than spinning disk for long-term archival. The key is metadata. So if you get your engineers working on that metadata layer, you can then position tape as not only cheaper but higher performance than disk for long-term retention. Have you heard of Spectrum Archive? No, tell me about it. I mean, I've heard of it, but I don't know much about it. Eric Herzog was talking about it. Well, you hit the nail on the head. That whole notion of metadata, right? With LTFS, that was its prior name, linear tape file system. I think that's still within the marketing vernacular. I don't think it's been sponge- LTFS is the key enabler here. Exactly. So what we do is at the head of every tape, we basically have all the metadata. When you open the cartridge, you basically read the metadata into electronic memory. And from there on out, every search, you know exactly where the file is. You start the tape and go exactly to a point. You're not winding back and forth trying to locate something. You've got, I'm going to date myself. If you recognize this acronym, you'll date yourself. We kind of have a V-talk for tape. Oh, a volume table of contents for tape, yes. There you go, that's what we've got. We've got a V-talk on every tape. Exactly. You pop the thing in, you start running, you find exactly where the file starts, you do high speed out to that spot, and you start reading. And then all of the dynamics that you brought in, kick in, where you start streaming it, you can save it in buffer, because I think once you start doing sequential tape, it's about the same sometimes, faster than spinning sequential reads. Take advantage of it, it's there. That's a hundred million dollar idea for IBM. Get your engineers on it, get your marketing people on it. And I think it just drives a lot of value for clients. When I've talked to customers about this at first, they're like, that guy's crazy, doesn't know what he's talking about. And then afterwards they come up and they give me the business card, they go, you know, that's really interesting. And I've got examples of people in the field actually doing that, so let's talk afterwards. Yes, sir. I'll be honest, if you need more, if you need more spottings in the field and you want to talk to customers, we've got plenty of them. It makes a ton of sense. It saves a boatload of cash and I keep pushing this and we have a couple of CSPs that have taken us up on it. Use this as your back end to your great big scale out cloud scale storage farm. Sooner or later, that price of near line SaaS is still going to be too much and tape will always undercut it. Start thinking about how you can move the data off the spinning disc on a tape and you'll undercut your competitors. And there are plenty of them out there. The winners are going to get tape, the losers won't. One-way ticket to the Bitbucket. We didn't even talk about retention times, tape lasts way, way longer than disc. You've got to cycle through. Lot of interesting advantages and discussions. I'd love to get a card. I'll send you the piece that we wrote on this. You got to go, meet a customer. I want to talk some more about DS8000. Okay, give us the, I'll give you the last word on DS8000. You guys, look, I'm going to make this really simple. If you've got ZOS, if you've got high end power and you really have, really have high, high availability needs, you got to go with the DS8000. When I say ultra high availability, let me define it real simple. And you, I'm sure, run into this. If you're in a position where your company is invested in the second data center and perhaps a third, and if you really stretch it a fourth and there are plenty of companies that have, you need to have the DS8000 because in that environment, nothing works as well as it does. We've added recently the 16 gig, as I've said, and all the performance advantages that DB2 now can deliver as a result of that 16 gig. I think it's an overlooked platform, and frankly, it's the most robust in terms of RAS. Number one, it's the highest performing. I have to admit, out in the distributed space, not everybody needs what it can deliver, but there are an awful lot of applications both in Wintel and UNIX that still need that kind of performance. If you're having troubles with what you've got installed, bring in a DS8, you won't have troubles. Whether it's spinning disc, if you have high cash hit ratios and flash, if you don't, you're going to get more performance out of a more beefy controller that's what the DS8 delivers. All right, Jeff Barber, thanks so much for coming on theCUBE. It was really a pleasure meeting you. It was great getting caught up. Thanks for the time. Thank you for the time. We'll be back with our next guest right after. This is theCUBE. We're live from IBM Edge 2015. Right back.