I had the opportunity last week to sit in on a CloudSlam multi-vendor panel session on the future of storage in the cloud. It was a very interesting and revealing discussion. I’ll do my best to summarize it here and share my thoughts and conclusions.
What is CloudSlam?
CloudSlam is a virtual conference, an online gathering of Cloud computing professionals. Now in its third year, CloudSlam 2011 had thousands of attendees from over 80 countries.
The session was titled “EMC, HP, IBM, NetApp: where they are taking the Cloud”. It was intended as a discussion of where the vendors see the future of data storage for the cloud. To quote the session description:
This will be a technical discussion, featuring senior technologists from the world’s four largest providers of cloud infrastructure. Analyst Mike Karp will moderate a roundtable of senior strategists discussing where each of their companies is going to drive cloud computing over the next 4 to 6 quarters.
Each panelist will make a brief presentation, and will participate in what we expect to be a vigorous Q&A session.
This discussion is open to everyone, but will be particularly valuable to technologists and senior planners involved in setting their company’s cloud computing strategy. If your company’s strategic plan includes an investment in cloud computing, and if you play a part in that process as a decision-maker, influencer or implementer a, this roundtable discussion is for you.
(Yes, I see the typo in the session description. I quoted it as-is.)
The panel was moderated by Mike Karp, VP and Principal Analyst at Ptak/Noel.
The four panelists were:
- Chad Sakac, Office of the CTO at EMC
- Geoff Hough, Director, Strategy & Planning, Office of the CTO at HP
- Lauren States, Vice President, Cloud Technology and Client Innovation for IBM Corporate Strategy
- Val Bercovici, Cloud Czar, Office of the CTO at NetApp
[NOTE: In the interest of full disclosure, I have varying degrees of relationship with three of the five above-mentioned people. I work closely with Chad Sakac and his team on several Cloud- and virtualization-related initiatives. I met Mike Karp at VMworld 2010 and have engaged in some conversations with him via Twitter. Val Bercovici and I met over Twitter because of my live-tweeting of this panel session, and have exchanged a few messages since. I do not (yet) know either Geoff Hough or Lauren States, although I do now follow Lauren on Twitter.]
The panelists were each allowed eight minutes to present and a question and answer period. The panelists presented in alphabetical order by the vendor they represent.
Chad Sakac began by acknowledging that it is EMC’s goal to be the undisputed leader for Cloud infrastructure and the undisputed leader in partnering with Cloud Service Providers. Now, I don’t think it surprised anyone to hear that’s EMC’s goal, but personally I found it refreshing to just have it out in the open from the beginning.
Chad was clear throughout his presentation that EMC believes that in order for customers to realize the full potential benefits of Cloud that the move to Cloud needs to be about transformation of how customers think of IT (e.g.: workflow, process, and design models), not just about new or changed infrastructure. He talked about the different layers EMC suggests that customers think about in order to be successful when building their cloud.
Chad advised that customers should first think about who the Users of the Cloud will be. What are their needs? What kinds of work do they do? What types of access do they need?
Knowing the Users’ needs leads to the second Cloud layer to think about: Applications. What applications will you need to run in your Cloud to meet your users’ needs. What will let them access their data in the way(s) that they need? What are the storage requirements for theses application? The performance requirements? Uptime/availability requirements?
Only with the Users and Applications figured out are you ready to think about the third Cloud layer: Infrastructure. What infrastructure will support those applications’ requirements? How can you build the most cost-effective infrastructure to serve these applications and data in a way that meets the requirements?
Chad next talked about growth and scaling in the Cloud, as well as tools to help with migration of workloads to the Cloud. We’ve all grown accustomed to the obvious benefit of virtualization: The ability to have a groups of users and applications share a pool of resources. In order to grow, Chad said that we need to stop thinking about pools, and instead start thinking about “pools of pools”.
In order to take advantage of these pools of pools, organizations need to look to moving their applications — re-platforming them. Chad suggests looking at technology like SpringSource, vFabric, and CloudFoundry as enablers for this.
EMC believes that the way customers will get to this “pools of pools” point is through hybrid cloud and federation.
Chad then offered some hints on the future from EMC.
[Disclaimer: I work for EMC. I’m in a position to know things about EMC’s product roadmap that I’ve agreed not to discuss in public. Therefore, while I may indulge in speculation on other vendors’ futures, I will simply report what Chad said, and leave you to speculate on it as you will, unassisted by me.]
Chad said that EMC is embracing the move toward storage software and functionality being provided on commodity hardware, and that this is important as value now moving in different ways.
Chad also invited listeners to speculate on the future of the EMC family of VPLEX products. He asked, “What if data could be local without being local?” and invited listeners to think about moving even beyond VPLEX.
HP’s Geoff Hough was up next. He started off by immediately stating that HP also believes that the cloud of the future is the hybrid cloud.
Geoff went further and stated that HP feels that a converged infrastructure all from a single vendor is the only way for customers to go. (Speculation as to which vendor Geoff felt should be your single vendor is left as an exercise for the reader…)
Geoff said that the end-goal of Cloud is to provide a way for customers and end-users to consume resources and data remotely. I was struck by what a great definition this was.
Geoff closed by stating that a resilient infrastructure important to meet Service Level Agreements (SLAs) and mentioned what HP products he felt fit those requirements.
Speculation and Commentary
It will come as no surprise to anyone who’s been following the storage industry to note that when Geoff spoke about HP storage products suited for the Cloud, it was not HP’s former flagship storage product, the EVA, he talked about (apparently even HP agrees that the EVA is ill-suited to meet the needs of the Cloud…). Instead the only storage product he mentioned was 3PAR, HP’s relatively recent acquisition. From this, I feel it’s safe to conclude that HP really does see the EVA at a dead-end and views the 3PAR line as their way out of that particular cul-de-sac. It will be interesting to see what sort of “upgrade to 3PAR” deals HP offers their current EVA customers.
Nowhere was HP’s other storage acquisition, LeftHand Networks even mentioned. At the time of the acquisition, we all speculated that HP saw LeftHand as their entry to the Cloud. Its grid-storage model seemed well-suited for the cloud, if they could only overcome its problems with scaling and lack of storage efficiency. Personally, I’m going to take the 3PAR acquisition as the sign that they were not able to overcome these limitations.
IBM’s Lauren States was up next. She started off talking about how the explosive growth of Big Data is a natural driver for the transition to and evolution of cloud. Lauren said that the move to Cloud needs to be about enabling business and making use of Big Data.
Lauren then stated that IBM believes that the future of cloud will be, by necessity, hybrid clouds. She then talked about how customers should build their clouds for success. (Her statement of IBM’s position on this struck me as similar to Chad Sakac’s statement of EMC’s position, although they both used different language and approached it from slightly different directions.) IBM believes that customers should determine what their specific needs are from the cloud (i.e.: levels of security, performance, SLA, etc.) and choose their cloud platforms/providers appropriately.
Lauren talked about ways customers might use the Cloud. She spoke about collaborative models that already exist (e.g.: online games and the communities of gamers that grow up around them, medical research, etc.) that help to make the mental/social transition to cloud easier.
Lauren closed by speaking about how Cloud enables better analytics. She said that IBM is acquiring commerce-focused companies and making more of a move into analytics and providing services in that area.
Speculation and Commentary
I thought it was interesting that IBM never actually spoke about products at all, but focused instead on service offerings. Now, on the one hand, that isn’t too surprising. IBM’s had a very strong business for years with their Global Services division, so their offering Cloud-related services is an obvious move, and taking advantage of the Big Data trend just seems smart.
On the other hand, a lack of mention of any product at all seems, well, a little off. My take on this is that it’s a further indication that IBM’s acquisition of XIV, the array that was going to make all other storage obsolete, is not living up to its promise. The line has been plagued by delays in the release of new features from the start and hasn’t been shown to scale beyond a single rack. Without scaling, either within a product or by federation, a storage system is ill-suited for Cloud use.
Val Bercovici of NetApp was the last speaker of the day. Val started off by saying that NetApp has made a conscious decision to focus on storage infrastructure and not enter cloud services market themselves and that NetApp is focused on lowering the price, and therefore ultimate cost, of cloud infrastructure.
Val talked about importance of open standards and building for them. He referred to the upcoming Storage Networking Industry Association (SNIA) standard for a Cloud Data Management Interface (CDMI) as a specific example.
Val pointed out that in the cloud, physical assets are shared, so it’s important to remember security. He then stated a position that echoed EMC’s and IBM’s that the future of cloud storage will be driven by application requirements not infrastructure.
He went on to state that NetApp sees future of cloud storage as “going beyond the concept of LUNs and filesystems” and that NetApp believes that the cloud storage of the future will be object-based, requiring rich meta-data.
During the question and answer period, Val talked about some (unnamed) applications that cannot move to cloud. Chad of EMC interjected saying that he believes the only barrier is in the thinking — that replatforming is possible for almost everything.
Speculation and Commentary
I found it noteworthy that NetApp was the only vendor on the panel to not explicitly call out hybrid cloud as the way of the future. It’s possible that this was intentional, as part of a desire to maintain differentiation from other vendors, or perhaps even intentional as NetApp sees another route to the future of cloud.
What other route? Well, again, I can only speculate here, but two possibilities come to mind.
The first would be that NetApp might be close to actually incorporating the intellectual property from the Spinnaker acquisition and may finally be on the verge of have a true converged OS, and will be capable of offering Cluster Mode across all storage protocols.
The second possibility, and in my mind, the more likely one, is that NetApp is counting on the additional feature support for NFS that VMware has promised in the next version of vSphere will allow them to use global namespaces for NFS filesystems spanning multiple sites. Remember always that NetApp’s FAS architecture is tied to their Write-Anywhere File Layout (WAFL) filesystem, and that their origins are as a company that provided NFS-only storage. (And did it very well. Back in 1993, I personally installed and supported several of the first Network Appliance “file toasters” here in New England. This was back when the term “toaster” was intended as a compliment rather than disparagingly.)
[Further Disclaimer, if it’s needed: Yes, I’ve installed and worked on NetApp gear at different points throughout my career. I haven’t been employed directly by NetApp. The work was done either when I was a contractor working for resellers, or running data centers when I was a NetApp customer.]
I also found the discussion of the future of Cloud storage being object-oriented very interesting. In the abstract, it’s very logical. In a “true” cloud model, customers wouldn’t think about VMDK files, they’d think about a “virtual server object” having certain characteristics (e.g.: service level requirements, backup and recovery requirements, performance requirements, etc.). Customers wouldn’t think about database files, they’d think about a “database object” that has certain characteristics (e.g.: the type of data stored, lists of who has what level of access to the data, security requirements, etc.). And so on.
Take the discussion out of the abstract, though, and on the surface it’s an odd statement coming from NetApp. NetApp doesn’t actually offer an object storage platform. Sure, they’ve got StorageGRID, but at its heart it looks to b nothing more than a software layer built on top of WAFL.
With no object-based storage platform of their own, and a stated belief that object-based storage is the future, well, I can’t help but wonder if this was the reason behind NetApp’s acquisition of Engenio from LSI. I’ll be watching the future of that one with great interest.
(Yes, I’ve spent more time speculating about NetApp than I did for either HP or IBM. What should you read into that? Only that Val offered me more to speculate on then Geoff and Lauren did.)
However ‘s going to turn out, the future of the Cloud is going to be exciting. There are several certainties that you can count on:
- Cloud and virtualization are coming. It’s no longer a matter of “if”, it’s “when” organizations will adopt it.
- Vendors are starting to “get it” that it’s the users and their applications, not infrastructure, that drive data and cloud growth.
- Vendors and customers will have to meet the challenges of explosive data growth and Big Data.
- The ability to provide proven data protection for the Big Data in the Cloud will likely become a big differentiator.
- Data and workload mobility are becoming more and more critical.
- With the combination of the importance of the data i the Cloud, use of shared resources, and increased user mobility, the ability to provided security — trusted Clouds — will be hugely important.
What are your thoughts on all of this? I encourage you to engage in your own speculation on the future of the Cloud and cloud storage in the comments section of this post.