 Good afternoon to everyone here in the Boston area and welcome to the August 21st peer insight where we'll be discussing harnessing crowd wisdom for the future of virtualization. I'm very excited today to be joined by Noemi Graysdorf who is VP of strategy for Cambridge Computing Services, Computer Services. We also will have today Nathan Smith, Senior Citric Engineer at Centered Networks, Yaqum Hepner who is Senior Manager at Sanofi, as well as Duncan Epping and Frank Denerman from VMware, and John Blumenthal from Cloud Physics. So the title for today's peer insight, Harnessing Crowd Wisdom for the Future of Virtualization, is really all about bringing a big data approach to analyzing VMware environments and making them easier to manage. Just a couple of logistics before we get started. Remember that it's a pound, sorry, star six to mute the line. Is that right? Yeah, star six to mute and star six to unmute your line. If you're not speaking, please mute your line so that we don't have feedback on the line. And if we will be pausing throughout the peer insight to give folks on the line an opportunity to ask questions of our participants today. So with that, let me get started. I want to start with you Nathan. Nathan again, thanks for joining us, Senior Citrix Engineer at Centered Networks. Nathan, first tell us a little bit about your environment. So we provide hosted virtual desktops and hosted applications. Our environment is about 95% virtualized. And it's a blend of VMware and Zen server. And we host on NetApp Storage. Okay. And tell me a little bit about some of the problems that you've been trying to solve on the management side. I don't know if it's management necessarily, but where we've run into a lot of issues is trying to make sure our performance is good and we mix our SLAs. We've run into some issues with the root cause analysis of performance problems in general, but specifically when we get down to the VM storage level, we were finding we had a kind of blind spot there where we had a feeling that we were heading towards the right direction. But once we started looking for some data on VM storage, we had a bit of a blind spot. We could see from the Windows guest devices that things were fine. And the NetApp performance advisor was also telling us things were good, but we couldn't really understand how the hypervisor was viewing the storage. And that's where we started working with Cloud Physics to try and get some metrics around that and then also be able to interpret those metrics in a useful way. What were some of the metrics that you were able to capture using Cloud Physics that you couldn't capture before? I'd have to decide the problems from Cloud Physics to really get into the details, but there were these good statistics. And one of the advantages to others of working in this way kind of utilizing other people's knowledge and experience as well as other people's data is that we were able to get some relatively simple views of some very complex numbers. So the way that when we were working with Cloud Physics that it presented back to us kind of simplified a lot of that, which is why I don't have a very clear view of the specifics. I just know the way it was interpreted. I was able to get a feel for in more kind of layman's terms whether our storage was in a good place or a bad place. So what did you do as a result of the analysis? At this point we haven't done anything differently. We're still working through it's obviously in this kind of process and this approach is in its very early stages. So we're still working closely with Cloud Physics to try and drill in a little further to figure out what we can do to improve things. Yacum, can you describe your environment there as Sanofi? Yeah, so we are a VMware based environment with all VMware as far as virtualization goes with EMC storage. The environment today is roughly 70% virtualized globally. Okay. And what kinds of challenges were you having? Are you having in your environment? Well, I think again for me, in my opinion, just going forward in the future, we're going to have to start looking at things a little differently, especially when we start looking at Cloud based computing, all the appliances that are out there. Again, these are things that are going to change the way we manage the platform. Just as a reminder for those on the call today, if you could please mute your line if you're not speaking, it would be helpful. That's Star Six to Mute. Again, Star Six to Mute, thank you everyone. And Cam, I hope you get back in the office soon. So are you sending data up then to Cloud Physics? How does that work? Yeah, so we just started participating in a data program and we've been sending our data for about a month now. Okay. And one of the things that's always sensitive in corporations is sending data about your environment off site. I'm curious as to how you approach that, who had to be involved in your organization to sort of make decisions regarding sending your environment data to a third party? All right. So we've looked at what our current policies have and we've actually already participated in something similar. We were part of the V-benchmark program from VMware. So we've done something similar in the past. So again, looking at what Cloud Physics actually collects. They don't collect any local or sensitive information. It's basically just performance stats coming out of each center, which is fine. Okay. And do you control access to that data or does it just... You've decided up front what data you're going to send and it just goes. Right. Based on the account we used for the privileges, we didn't give it admin rights. So again, it's using some basic privileges to collect the information. Just enough to collect the information it needs and send it up to Cloud Physics. Duncan, you've been acting as an advisor to Cloud Physics. So in the context of your role at VMware, how is VMware and how is Cloud Physics working together? Duncan, star six to Unmute. I was being unmuted at the moment. I'm indeed an advisor for Cloud Physics, but that's more from a community-slash-blogging perspective. Okay. And it has been my primary focus, at least from working together with Cloud Physics, to give early product feedback and product direction. And I guess that what I've been focusing on most is the community aspect of things, the usability of the product, and the potential of the data set itself. Hey, allora, Daniele. Prottini seti. Noemi, you've got some experience here with Cloud Physics as well. Can you talk a little bit about that? One of the reasons that I sort of started getting involved with Cloud Physics and having those conversations is a lot of our customers who are running relatively large VMware environments are struggling to identify performance bottlenecks to really get, as I think it was Nathan who mentioned, really getting down to the smallest increment in terms of... ...what the Automated Self-Service Directory... ...what the environment is doing in order to optimize it in order to get better performance, in order to support more applications in virtualized environment. And so we have had Cloud Physics introduced to a number of our customer accounts in order to help them sort of get a better understanding of what's happening. I have not yet talked to a single account where there aren't either challenges in managing storage in virtualized environment or who are not in need. So everybody needs something to either troubleshoot what they already have, to plan for the future, or to get a better understanding of where they are based on where they want to go. One of the things that we've been looking at is how introducing changes into the environment and impact the overall environment. So it's been a lot of discussion regarding SSDs and impact of SSDs in VMware environments. Can you talk a little bit about that from your perspective, what you've seen? So there is a sense to some extent in the marketplace that if you have SSD or SSD is the answer to all your problems. So if you put in SSD, it automatically solves your performance challenges, especially in the VMware environment. We're finding through conversations with customers that oftentimes adding SSD into the environment, you solve one problem, the bottleneck moves elsewhere. That's always the case, but when you're adding something that is so exponentially faster than what we have today, the hard disk drives, that bottleneck can be significant and cause significant issues. So you can't just blindly put in SSD to accelerate the performance and expect that the rest of the ecosystem is going to function perfectly well. You're going to either in the long term or in a very short term or likely run into bottlenecks along the data path, whether it's in primary storage, whether it's in data protection, replication, or something else. So I think one of the hopes here using a data analytics approach to managing VMware environments is that you can benefit from the experience of other users who've gone before you and tried similar things. So can you talk a little bit about what you see there? So it's a combination. In the presentations that I've sat through with Cloud Physics to the end-users, it definitely resonates the idea that somebody out there in the large VMware communities is either implementing SSDs or implementing a slew of storage systems that we have on the marketplace or doing something that is relevant to you. And so having that feedback loop is very, very valuable to them. Also being able to post questions to a large dataset is very valuable. What's really valuable about that is the fact that it's sort of anonymous. Anonymity is a big issue for many organizations, especially if you get into government, a lot of research labs. They're very conscious of the security issues. Even though they're only collecting operational data, there's still some issues around security. The other part of Cloud Physics that I think is valuable or is interesting is that there always has to be somebody who's first. If you're the follower, then there's plenty of people who have done certain things that you can take a lesson from it and implement in your environment or use it to troubleshoot your environment. But if you're the first one trying to do something like when vSphere 5 came out, if you were going to migrate to that, you don't necessarily have the experience of the crowd to leverage to see how to do it or what the best practice or what are going to be the gatchas. But the information that is collected by Cloud Physics inherently is also useful to model and to understand sort of what might be the gatchas in your environment as you try to implement certain things. Duncan and Frank, I know you have both been involved in looking at high availability. You've written extensively on the subject of high availability. And often when we make changes in an environment, it impacts high availability. Can you talk a little bit about what you are hoping to see out of this sort of big data approach to analyze environments and improving high availability as there are changes? Well, one of the things of big data is that we can actually validate if our best practices or the industry best practices are actually applied. And we can also see that if other settings are more common in most environments. So essentially identifying the best practices across a large community of environments and understanding which of those are able to maintain high availability. Yeah, let me add to that. Another thing that we're working on with Cloud Physics is specific simulation models for HA and DRS clusters. And those will basically allow you to try out changes in your environment without actually introducing them in your production environment. So that's a very effective way of seeing how a specific change to your environment can actually improve the resiliency or, for instance, improve the availability of workloads but also the availability of resources. All right. And I know for most IT shops they're relatively thinly staffed and they don't have a lot of time to do proof of concept so anything that will shorten that cycle is a huge benefit. Identifying what might improve the environment before you actually implement it in a what can be very expensive POC. Expensive not only for the customer but expensive for the suppliers. There's also a significant value in comparing different offerings in the marketplace more comprehensively. You could identify and create a workload set and then you can give that workload set specific to your environment to the various vendors and have them even run it in their lab because it's, you know, there's the conversation around, you know, we handle sequential loads well or we handle random workloads well, small file, large block, small block. But every environment has a different mix of those different parameters. So actually capturing a workload and making it possible for a vendor to test that workload against their system gives the customer some kind of, it empowers them because when the vendor then gives you the configuration it's the configuration that they should have tested and it says exactly it's going to work for your workload and this is what you're going to get out of it versus what a lot of times happens today. There are some basic calculations that may happen in terms of how many IOPS or how much latency you're going to have on a system and then when you actually implement the system it may or may not give you that because, again, the variability of workloads are significant across environments. John Blumenthal, I'll give you a second to unmute your line but I'm interested to hear from you. Again, John is the CEO of Cloud Physics. I'm interested to hear from you how much data you're actually going to collect. I'm sorry, I hope you guys can hear me. Right now we have 50 working data sets that we're using for aggregation. The number this week is climbing incredibly given the launch of our website. Yeah, you just came out of stealth mode yesterday, right? We did. We were working with larger data sets mainly to prime the back end of our system to start understanding how to load balance it. Now we have, we're in the process of onboarding nearly a hundred of what we call observers, useful appliances that are the data collection mechanisms that users download as an OVF and simply installed and from there anonymization occurs before the data is chipped up to our cluster that we run as our back end. So the total data set has grown into the hundreds of terabytes already. We're actually using aggregations of data that we've been collecting over the course of the last year while we've been operating in stealth mode to test a lot of the analytics that we've been writing and some of the cards or small applications that you can see on our site currently especially things related to high availability that Duncan and Frank have also taken. Again, star six if you can mute your lines if you're not speaking and appreciate it. I've seen a number of mentions of cards. Can you talk a little bit about what cards are and what that means to the community? Yeah, so cards are our visual metaphor that represents the user experience and orientation of our product, our service approach. Our service is a task oriented approach where users can derive very specific benefits from applying a card to a very specific problem and our service is not meant to manage or necessarily provide real time monitoring of a system. There are plenty of solutions like that in the marketplace and nearly all the solutions in the marketplace are focused on kind of an inventory oriented approach where you are presented with the environment and the environment is navigated by the end user to find an object on which to operate or execute a task. Our entry point is more task oriented where the user comes to cloud physics with a specific problem. The problem is rendered or represented as a card as a narrowly focused tool and you use that card to apply a specific solution to that problem. In this challenge that we have hosted this week on the FireCupium World you can see hundreds of cards that have been suggested in ways that are actually quite remarkable and representative of the power of the community. Literally there are hundreds of these things already just in the 48 hours that we've been live that describe very specific problems that products today just don't address or render and it's our ambition to continuously provide end users with these on-going solutions and as a consequence of this on-going community involvement in the development and delivery of these solutions this very massive data set will arise in these places. Again if you're not speaking please press star six to mute your line. Which one is my... Excuse me. Anyone on the call who's not speaking please mute your line star six. So I want to pause here for a moment and see if there are any questions from the community. Okay so... Hi John. Yeah. This is Mike Alvarado. I didn't have one question. Hi Mike. If there's a view into aggregating resources or making available a some kind of shared model where resources are actually tapped into. John can you answer that question? Yes it's a great question. We have built and are in the process of implementing you'll see this very soon a secure data sharing model that looks a lot like what you may have experienced in social networking already where for example with Google circles you can light list people to look at different portions of what you've created as your profile and effectively your portion of the file system tree. Similarly what we are working on and will have already very soon is the ability for you to share a key with the third party and that key can be revoked at any time. It can be fired and the idea is that the third party can then render and view your environment in an unmasked fashion and they could if you wanted to look at the association of who you are with that unmasked view but you as the end user have entire control of who views and who can access your actual unmasked state of your data and the technique is actually very well known in the way that social media works with privacy controls so we've borrowed from that very heavily to implement a very similar type of data sharing mechanism. We have a storage partner already who is using this process for the purposes of pre-sales and someone on the call today mentioned this already where a vendor in cooperation with the end user can request the key and allow for an analysis of an environment prior to sitting down with the end user to talk about their environment and the actual value proposition that a vendor's product might bring to that environment. In fact the call today we hope in the very near future to work with Wikibon to deliver this type of data sharing mechanism such that participants in the call like Nathan and Joachim could simply provide you the participants in the call a temporary access to look at their environment in a kind of schematic way for you to securely review and analyze what your environment might look like as a trusted third party. It's a core piece to what we're doing. It enhances and I think congeals a technical community and it's a core focus of what we're delivering. Mike did that answer your question? Any follow-up? Hi this is David Slyer. I have a follow-up question for that. Obviously the more people that you have in your database would seem logical to me but the more people that you have in the database the greater the advice and information you can give out to your users. Is your business model to allow users to value the data and then a subset of those users to purchase your services from that data? That's a great question. The power is really in the statistically meaningful size of the database as you indicate. Our business model is really twofold. The first is to deliver directly to end users value through a subscription service that is comprised of an ever-increasingly rich set of cards like the ones you see on our site as well as what has been suggested by the community. In fact, the idea here is that the community itself will be able to develop cards and off of a platform access that will be exposing such that third parties can develop their own cards somewhat like an app store in which these focused solutions could be distributed on an ongoing basis for a variety of cloud type infrastructures. And then the second part of our business model is focused on the product vendors who are constructing and delivering products to those same end users consuming our services. And so these product vendors consume really three types of modules that we're putting online for the vendor community. One is a sales module that increases the effectiveness of the sales force and customer satisfaction in the sales process. The other is a support module that addresses the multi-vendor environment you typically find in a virtualized environment in a way that allows for coordinated reviews of the user's environment in a secure fashion using this data sharing technique I described earlier so that the time to resolution of a given problem is accelerated and the VMware admin is not left as an administrative assistant trying to schedule vendor coordination and vendor calls. So that's our kind of two prong approach to the fundamentals of our business. And is the plan to expand this beyond the VMware environment? You're covering Citrix as well, Hyper-V, non-virtualized environments. What's the plan? Yeah, because of the domain expertise of the team coming out of VMware. But the other half of our team is actually drawn from Google and a lot of the big data community in the Bay Area. We are first and foremost focused on vSphere. We will be constructing data collection and analytics on other hypervisors as well as working with several cloud, public cloud providers that we're already engaged with. So the ability then to model your environment, understand your environment as you extend it out into cloud service providers? That's right. So the public cloud service providers have a variety of performance APIs that they expose for the purposes of monitoring the performance of instances you instantiate in EC2, for example. What's critical though is the ability to also understand what's happening behind the virtualized veil that the public cloud providers provide access to. So we are in the process of engaging several of these larger-scale public service providers to help with the modeling of how their back-end physical infrastructure responds to the demand coming from the logical side, from the virtualized side of their business. This is David again, just coming back on your business model. Again, this seems to emphasize the more potential users you have, like members and support people, that the more people you have in your system the better. Again, I was asking specifically whether you were planning to pay people for their information even though they don't actually want your services. Oh, I see. We haven't approached that type. We haven't approached that business model as a way for generating a revenue stream. Instead, the approach that we have taken is to really create a very simple API into the expanding data set so that by contributing to this independent platform, you could also build potentially your own business or your own product in many cases off of the data that we're caretakers for as an enablement platform. So it's a little bit the inverse of what you're asking. Instead of paying people for the data stream, we instead take the data stream and place it in a way that would allow you to build a fundamental application or report that would be of high value for others in the community to consume either free or under a revenue sharing model that might arise from that. So Yacum and Nathan, I'm interested to hear from you if you feel like either centered networks or Sanofi will go down the path of trying to generate revenue from applications or if we're really targeting a different sort of customer here. This is Nathan. I don't think for us we'd be looking to add revenue by using cloud physics. I think we would sit on the other side of the equation that John's talking about, where we look for value by using the subscription service and getting these very simplified views of a great complex environment with the added benefits that we're also able to leverage other people's data to get a more accurate view of how we're performing or how our environment is set up compared to other environments or best practices. And I suspect Yacum that in a pharmaceutical company the focus is going to be on improving your development of new medicines and not new apps. Correct. So again for us it's going to be more about maybe peer comparisons. Are we, from an IT point of view, are we at the same cost model as let's say one of the other competitors? Something along those lines. I could see it more useful on that category in the future. It's interesting if I roll it back 20 years ago or so when I was on the customer side we looked at different benchmarking tools to see how we were compared to similar kinds of companies of similar size. But you couldn't do a direct comparison between State Street Bank and Fidelity for example because neither company would want the other to have that kind of visibility. What kinds of comparisons to other people's environments are you going to be able to provide? And what kind of protection are you going to be able to provide relative to, you know, I'm target and I don't really want Walmart to know what my environment looks like. John, that's for you. Yeah, so this is Yacum. Are you asking us? Actually, I'll ask you, Yacum. What about a Sanofi? I mean you probably don't want Mark knowing a lot about what you're doing. Sure, but at a high level that information is public, right? So in the information week 500 and all those, you know, other similar magazines they do publish annual IT budgets. So you can kind of calculate your number based on the number of employees, which we do today internally, right? We always look at that. Okay. So it would be interesting to see maybe some other information. Again, you know, everybody's different, everybody uses it a different way. I don't know if you can gain one is more efficient than the other. I think from anything else it might be more of a, okay, that's interesting. Now we understand where we fit in this. Not so much, oh, we really got to get to that. Right. So it is more important to sort of look at, to satisfy your management that you're operating a very efficient environment relative to the competition. There's a lot of variability in that from organization to organization in terms of their, how risk averse they are, how conservative or not, where their focus is in terms of IT, you know, how they manage their information. There are some things that are very standard, but there are some things that are very specific. Again, you know, even if you take two research labs and their HPC environments, they could be all in life sciences, but they could be completely different workloads and how the data is stored and how the scientists are using the data is going to be very different. So the value of the aggregates is really more around the infrastructure and how do you optimize the infrastructure against the kind of workload. So it might actually, for Sonofi, it might be more valuable for them to see or to compare themselves against somebody else who's running a similar infrastructure. It could be in a different industry, but they might be doing something that could help Sonofi improve either VM density or performance. Or whatever it is that they're trying to achieve. So it's not just benchmarking against your industry peers, but against the larger peer community. It's good. So let me throw it back to you, John. What kinds of organizations do you imagine having an interest in developing the apps for the community that they might sell? I think this begins, as we're already seeing, as a grassroots initiative within the very active community that Duncan and Frank are particularly strong in their leadership. I think it begins literally down at the people on the front lines of managing these systems. So it's really at the individual level. You can imagine some individual deciding to, hey, I've got an idea here and I've got a way to make some money. Yeah, I think, I mean, in a way, first of all, I think the community, and I deferred to Duncan to speak, and Frank to speak authoritatively about this. But I believe there's a heritage here that goes back to the UNIX utility days where you have very focused utilities that were the core philosophy of how you manage UNIX back in the day. And that if you wanted to create something of utilitarian value, you did that in a very narrow focus way. And then people invented ways to type these things together to build increasingly rich solutions. I've drawn from that experience in, I think, that trend that led to a lot of success in managing open systems. And I see that happening in the community today. I think Duncan and Frank could comment, I think, about the kind of demand in the nature of the community in terms of either looking at their work as a means of deriving revenue, or whether there's more of the spirit of sharing an openness that is driving some of the things that we're already seeing at Cloud Physics. No, good. Frank and Duncan, why don't you flesh that out for us and give us some of your perspective? How do you think it's going to play out? Is it going to become a charitable organization, a charitable approach, a big open source group hug, or is it going to become everybody trying to build their own portfolio for fee apps? Let's hope that it's going to be a big open source group hug. That's what I would prefer. And by the looks of it, if I look at the current backend of Cloud Physics, there's been many cards submitted already by the community. And I'm guessing that it's going to be the same in the upcoming months. A lot of people actually see the value of this platform. And we had some early blogger calls, and they also saw the value of this. And they also see where they could actually benefit from getting access to a data set like this. So they're all happy to contribute to the platform by feeding in desk practices or some sort of monitoring, logging reports. They're all really happy about doing this. And I think Frank and I started doing that with the HA scorecard, which is already in there. And Frank is currently actually working on a DRS scorecard. And that's also going to be a community effort. So maybe Frank can explain the part that he's working on. Yeah, Frank. So like Duncan mentioned, the DRS scorecard. So one of the most brilliant things of this is that we can use the analysis of the program to actually understand that if you change a certain setting, for instance, a VM reservation, what impact does it have on the performance of the virtual machine? And if I change it, if I increase it, what will happen? Not only to the performance of that virtual machine, but also on the rest of the ecosystem. And like Duncan said, it should be a big group hug. And I think it's easily comparable to the PowerSheelite community in the environment. Everybody's just working on creating cool scripts. And I think the similar thing will happen, creating cool cards and cards which no benefit from. Yeah, and an awful lot of VM where, especially into smaller organizations, gets implemented, gets shipped by sort of one, two-person shops that are, you know, selling into a particular community. It seems like this would be a great resource of best practices for them as they're trying to not only build out their practice, but improve the customer experience. I think it actually goes into the larger shops as well. We've talked to a number of, you know, a number of our clients who are even larger environments and what would be considered a relatively large VMware shops are struggling with getting the information and being able to work with the settings. DRS being a great example. I remember one meeting we had where the admin is going to try to modify the default settings of the DRS. But there are so many variables that it's very difficult for one individual to optimize the whole environment and, as Frank mentioned, consider all the implications of what they do on the ecosystem. So it's not just I'm dealing with this one single VM because of the abstraction and the shared infrastructure. You really have to understand the issues that are going to cascade out of what you're doing. And for even a larger shop, it's actually even probably more valuable to have the ability to leverage a community and to get that kind of information as they try to optimize and use these tools. Yeah, I keep coming back to the notion of SSD, which has been so, it's gotten so much visibility. I mean, IBM just bought the oldest solid state disk company, I think it exists in Texas memory. In recognition, they had, and Texas memory has a variety of different formats, different offerings. But there are so many different places where you can interject solid state disk or flash into the environment, and each one has its own sort of implications, right? And each one has, and may have downstream benefits or may have downstream, create downstream problems. Absolutely. And if you could marry that up against sort of pricing data, it starts to, customers should be able to make much more informed decisions. I think this is all about making more informed decisions. Yeah. Let me open it up again for questions to the community. Anyone else have a question for Yacum or for Nathan? This is Scott. Well, I've got pretty much my standard question these days is how are organizational CIOs that you talked to, how are they looking at this product and what kinds of activities in their environment are they hoping that will be streamlined or negated? And basically, what's their goal for an ultimate outcome? So this is John. What we're finding is a reflection, I think, of Noemi's statement, which is by having access to more of a global set of data and data scientists who are trying to continuously interpret the data as it relates specifically to your environment, it leads to just better decision making where the notion of relying on market-wide operational insights and how that might affect your environment in terms of cost, risk, and waste, I think that's the ambition of cloud physics and I think it bubbles up to the CIO by seeing greater efficiencies and greater effectiveness in the overall operation of the infrastructure. And by operation, I mean that comprehensively from dealing with vendors and how you efficiently understand and procure the right products and then how you implement those products in the best possible way and do that in a rapid cycle driven by a kind of collective intelligence that doesn't mean you have a reliance just on the one overloaded individual who simply doesn't have the time and is operating under intense pressures. The CIO, I think, just sees just greater effectiveness and speed out of what he commands. So this is a more efficient use of IT resources, but it's faster time to market things like that. It's essentially faster time to market. Okay. Any other questions? So John, a couple of times you've mentioned the cards and the submission of cards and the uploading of massive amounts of data. If someone wants to share their data with you or wants to suggest a card, what do they do? So today you can register on our site. You can use a limited set of cards that we have developed in conjunction with the community. So these are community big hug cards that anybody can have access to? Yes. Okay. So we have this online. We can see that there are hundreds of cards being suggested that we're in the process of sorting through with the assistance of the community as to prioritization and code development. We have not announced any of our pricing or packaging around any of this. We're going to be doing that in the course of next week with VMworld. Which is where we're really publicly presenting the company and the work we're doing. The data sharing mechanism I mentioned has not been implemented at this stage. And instead what happens is your data is anonymized so that neither cloud physics nor trusted third parties at this time actually share data with one another in its unmasked state without explicitly engaging cloud physics as an intermediary. That is the process of doing that we are working on and will be implementing so that cloud physics is removed as that intermediary for the purposes of sharing your data with another third party. And for what we're doing with that primarily is working with a small number of storage vendors to allow for the pre-sales process to operate far more effectively. So in this case, let's say an entity vendor is wanting to engage an end user of cloud physics. That storage vendor has to talk directly to and call up their prospective customer who grants access to their unmasked view in which case cloud physics then receives the key from that end user and allows the third party the storage vendor to then look at that customer's data set in its unmasked form. We're in the process of removing that manual step in between to allow access to the currently anonymized data. And so where is all this data kept, John? So physically it all runs out of EC2 and it sits in a massive HBase cluster that runs out of EC2 at this stage. Okay. And you're keeping multiple copies of the data? Yeah, by HBase standards you can define within HBase how many copies will redundancy are kept. There's actually a running debate that we're engaging the community with as to how many copies are necessary as well as how much history people want to deeply access because the back end storage system is in and of itself a major systems challenge that we've worked on that involves a tiering mechanism such that the user experience for analyzing more recent data can be delivered in rapid form with very low latencies. And data that goes back in history needs to be moved off that more performant tier onto a not an archive but a slower less expensive tier. We're challenged by many of the same challenges that our end users are. Well, I appreciate the perspective. Just as a reminder, we're here talking about harnessing crowd wisdom for the future of virtualization. And we have Duncan Epping, Frank Denerman from VMware on the phone with us. We have Nathan Smith from Centered Networks and Jakim Hepner from Sanofi as well as John Blumenthal from Cloud Physics. And Noemi Graysdorf VP of Strategy and Alliances at Cambridge Computer Services. So we've got quite a group here, but I'm going to give sort of the last word or last comments to Jakim and Nathan because they're the practitioners there in the field. What do you hope to see from the likes of Cloud Physics or other organizations in terms of helping you improve your operating environments? What are the next steps? This is Nathan. Sorry, go ahead. Go ahead, Nathan. Then we'll go to Jakim. I was going to say, really, I'd probably echo the things that have been said already, which is, for me, it's really three things. One is being able to stimulate those changes so we can look at the outcome of the change before we actually have to touch on any production. The second one would be, you know, to carry out proof of concept without actually doing a proof of concept. I think, as Noemi mentioned earlier, even if you can run a proof of concept, trying to get that to match your real-world data is impossible. You don't know how many hours you spend trying to simulate that workload, but you just can't simulate it. So being able to take a real workload and run it that way is huge. And then for us, particularly, this idea of simplifying everything down to the cards that we keep talking about, where we can very easily, not just someone who's technical, but even someone who isn't technical, can log into a portal and see a kind of almost traffic light system of green and red. You know, there's your HA, so probably here are the 10 things you should be doing with your HA. And you can just check through, or rather, you know, the card will look for you, that's the way you mention it. So they would be the three things that we're most excited about. Great. And Yakom? Yeah, I agree with Nathan. You know, again, some of the big things in the future as the virtualized environment gets larger and more complicated, it's going to be much easier to go to an external resource like this and try to simulate changes we might make in the environment, or, you know, if we want to grow it even further. Also, from a troubleshooting point of view, you know, I see that being helpful again as, you know, as other people run into this issue. Are we unique with this problem, you know, before we go down the road of opening a case? Again, the whole notion of self-service and looking at the community for help with some of the more complicated items is a good thing. Great. And just as a reminder, so Cloud Physics website did open up yesterday to the world. Cloud Physics is going to be at VM world. And you're exhibiting, what's your booth number there, John? We'll actually be exhibiting from the Fusion IO booth. Fusion IO is one of our strong partners. So as a startup, they have been gracious enough to allow us to exhibit out of their booth. Okay. And so submit your cards, submit your data, and help influence the community and make this a giant group hug. So, again, I'm John MacArthur. I'm here with Noemi Grey's Dwarf. Peer in sight in advance of VM world, harnessing crowd wisdom for the future of virtualization. I want to thank Scott Lowe, Mike Alvarado, David Fleuer for their questions. Also, Nathan Smith and Yacum Hepner, thank you very much for your time today. Noemi Grey's Dwarf, thanks for making the trek into the studio with me. Thanks. Nice to see you again. Duncan Hepping and Frank Denerman. You get the late night award. Well, I guess it's not that bad. It's six o'clock. So, there's seven o'clock there. So, thanks for dying in and being part of today's Peer in sight. With that, if you can, at VM world, please watch the live streaming coverage of VM world on Silicon Angle TV. The Cube will be live there at VM world and have a number of Wikibon analysts there on site doing lots of executive interviews. So, please join them on Silicon Angle TV. We will have research notes up by the end of the week on today's call and you can replay the audio on the Wikibon site or watch a restream of the video on Silicon Angle TV forward slash Wikibon. Sorry, on Silicon Angle TV, my apologies. We keep moving things. So, again, thanks very much for joining us today. I'm John MacArthur at Peer Insights here at Wikibon. Have a great day.