|By Jeremy Geelan||
|September 8, 2009 06:00 AM EDT||
Cloud computing, both private and public, will unquestionably make life easier for IT. That is the firm opinion of Jason Liu, CEO of Univa UD, expressed in this Exclusive Q&A with Jeremy Geelan of Cloud Computing Journal and conference chair of SYS-CON's market-leading International Cloud Computing Conference & Expo series. Liu also discusses many other aspects of Virtualization and Cloud Computing.
Cloud Computing Journal: How exactly is High-Performance Computing converging with Cloud and Virtualization-focused usage models?
Jason Liu: Let’s talk about virtualization to begin with. Historically, HPC has taken the perspective that the infrastructure can never adopt virtualization because of performance degradation. Now, Univa is actually pioneering the use of virtualization in HPC – it’s viable right now and ready for prime time. We’ve already started moving customers to virtualized environments (at their demand), and benchmarking has shown less than 5% performance degradation. So, you get the huge benefits of virtualization with essentially no changes in performance. That’s why we’ll see the majority of HPC environments moving to adopt virtualization in the next 12 months.
Now let’s look at how cloud plays into this. Cloud environments have these 5 characteristics: Support for boundless applications, resource pooling, a virtualized infrastructure, a services approach to application delivery, and a metered usage-based payment model. Given this definition, it’s clear that HPC is actually an early-adopting cloud market because HPC environments already use the majority of these capabilities and technologies. What’s been missing has been the support for virtualization. But now, we’re able to turn HPC environments into efficient private clouds where virtualization is supported and performance – as I discussed before, remains high. In fact, I’d say based on our work with users that technical computing is actually one of the faster adopters of private cloud technology.
Cloud Computing Journal: And how then specifically does Univa UD’s cloud suite fit in, is it an evolution of your grid computing heritage?
Liu: Absolutely. Grid technology is a precursor to cloud, and Univa is of course one of the pioneers in grid computing. With our years of experience in grid and utility computing, our products are production-ready for internal, external and hybrid cloud scenarios. Unlike other vendors who are trying to retro-fit old technology into cloud or have developed brand-new tools from scratch, we’re unique in that we’ve been dealing for years with the same technology challenges that cloud brings. Historically, our grid technologies have supported 4 of the 5 key capabilities that define cloud (as I outlined previously). Univa technology was built to meet these challenges and has done so for years. And now, our infrastructure products offer support for virtualization, the final component in cloud enablement. It’s this heritage, along with our proven technology, that positions us well in the cloud enablement and management space.
You also asked about our product suite. Univa is the only company out there that is offering products to address all 3 of the top cloud use cases: public, private and hybrid. Not only that, but we address both infrastructure management and service governance – think of this as the muscles AND the brains of cloud. You can’t get that range of proven technology anywhere else.
Cloud Computing Journal: Have advancements in Virtualization and Cloud Computing made managing an organization’s IT resource inventory easier or harder?
Liu: Let’s keep in mind that virtualization is only 1 component of cloud (along with resource pooling, metered payment and other criteria). To speak to Virtualization in particular, I’d say that Virtualization will definitely make IT easier in the long run. But the challenge right now is that many companies have deployed Virtualization without a proactive strategy, and they haven’t instituted a manageable environment. As a result, they’re dealing with virtual server sprawl without a plan to manage this. Over time, with the proper adoption of dynamic resource management technology and planning, companies will begin to see the true fruits of their virtualization efforts.
Cloud, both private and public, will unquestionably make life easer for IT. Private cloud technology, which is general will be the first major adoption area, will dramatically reduce the costs of internal IT and more importantly allow IT to measure the cost of delivery. That hasn’t been possible in the past – and how can you evaluate or improve your performance if you can’t measure it? Cloud computing allows this. Also, cloud lets internal IT groups compare their performance to external benchmarks like managed service provider offerings to determine the real value of their services. Additionally, cloud can drive efficiencies through the pooling of resources which means lower costs but also reduced management overhead – again, making life easier for IT.
As for public cloud computing, of course IT’s life is made easier when computing can be effectively and securely outsourced. With today’s products, on ramping to a public cloud can be done in hours and companies can very quickly gain enormous benefits with no more effort than they’d exert in a static internal compute environment (or even less).
Cloud Computing Journal: What real-life use cases of cloud computing saving time and money would you point to right now as examples from which other companies can learn?
Liu: We’ve got several users I can point to right away who are having great success. In the private cloud arena, I’d reference Corus, a leading global manufacturer of steel products and services who is optimizing their computational workload and making HPC simpler and more flexible for users via SaaS delivery of applications over the web. With our private cloud technology, Corus has created an internal environment for more easily submitting, executing, tracking, and monitoring HPC workload, providing their users all the necessary functionality to run their HPC applications over the web in a secure internal environment.
As a public cloud use case, I’d cite Pathwork Diagnostics as an early adopter who is doing cloud clustering in Amazon EC2. Pathwork was looking to avoid a large hardware investment, so they decided to implement cloud and chose Amazon EC2 for pay-per-use computing and Univa for building the HPC clusters in the cloud. It’s been a fantastic solution for Pathwork – they’re really the model user for a service like EC2 because the cloud is serving their peak needs and helping them avoid purchasing hardware that would go underutilized a majority of the time.
Cloud Computing Journal: So in your view Cloud Computing is definitely recession–relevant?
Liu: Oh, very. That goes for both private and public cloud computing. Private cloud is all about measuring usage in a service based metered approach, and also more efficient use of hardware resources. So yes, the recession is forcing IT accountability and thereby requiring proper measurement and reduction in costs thru better utilization of what you’ve already paid for. With a mature virtualized environment, we’ve seen customers go from a 10% server utilization mark to a ceiling of 50-60%. Now with cloud, we can take that to 70-80% while reducing their OPEX and CAPEX (operational and capital expenditures). This is a very compelling value proposition in today’s economic climate.
With public cloud, the price points are dropping as we are increasingly seeing cost analyses demonstrate that you can save money by using public cloud. At the very least, a public cloud option gives businesses the flexibility to dynamical adjust their cost structure based on IT usage. It’s not about cool technology – it’s about the bottom line.
Cloud Computing Journal: And how about service governance, how does that help a company’s IT organization optimize the resources they are utilizing in the cloud?
Liu: How would the body operate without a brain? That’s the best analogy for what service governance does for a cloud environment. You may have tools in place for moving resources around and sending out workload, but unless there’s some logic and intelligence to what’s happening, there’s no way to optimize utilization and make sure the right decisions are being made (and actions taken) to protect performance and SLAs. That’s what products like Univa’s Reliance do.
Cloud Computing Journal: So this can be managed on the fly and still meet user SLAs?
Liu: It certainly can be. Right now in most data centers, SLAs are managed manually – that is, a human has to notice that an SLA is being violated and proactively initiate the actions (let’s hope it’s the best action) to rectify the problem as soon as possible. I think it goes without saying that an automated approach is faster and smarter. Univa offers the most mature SLA automation product in Reliance, which comprehends both application and infrastructure SLAs to ensure application performance is optimal, and also provides content management to ensure high-value users or workloads get priority.
Cloud Computing Journal: So your contention then is that if you run your applications on precisely the right cloud resources, you’ll not pay more than absolutely needed?
Liu: Part of the argument to manage toward application service levels is cost efficiency – users can reduce cost and increase utilization this way. But the biggest benefit is actually managing the business more efficiently. I’ll give you two 2 examples: First, a company running a web application with no concept of queuing. This is like McDonalds operating a drive-thru with absolutely no sense of how long people are waiting or whether orders are being filled correctly. Managing SLAs lets businesses prioritize high-value customers ahead of low and make sure those customers don’t wait too long for their work (where too long = whatever the customer defines). It takes 8 times more effort to gain a new customer than to retain an existing one. So shouldn't smart businesses make sure that happens, especially for their high value customers?
My other example is a Service Provider who can manage applications on a service level basis – so the provider can charge for services based on the SLA. This lets him prioritize customers and offer higher pricing (and service levels) to customers who will pay it, much like an airline does. It also lets the provider ensue that these customer with demanding SLAs actually get a higher level of service – automatically, guaranteed.
Cloud Computing Journal: How would you characterize the overall state of service governance technology today. Is Univa the sole pioneer of application-centric resource provisioning or are there competitors out there?
Liu: Well, there is certainly lots of confusion and misinformation out there. The reality is that no company is truly doing application service level management except Univa. The majority of others who claim this are either (a) only monitoring an environment that they can’t automate – so they require manual processes to remediate issues, or (b) are offering infrastructure but not application SLAs – so actions are taken based on the needs of the infrastructure but not the application. This means for example that if a server goes down, it gets replaced right away (some IT guy gets a call to go fix it) – even if no application SLA is in jeopardy. Is that smart use of resources, both human and machine? Not if you can avoid this by automating the whole scenarios based on previously negotiated application SLAs. The bottom line is that users want to define good service based on how quickly they get the results they want, not on the % that some server is being utilized. Tools like our Reliance make that possible.
Cloud Computing Journal: There seems to be little doubt that lightweight infrastructure is The Shape of Things to Come, for Enterprise IT? When, though? Are we talking 2010 or more like 2020, for major enterprises?
Liu: Let’s use the electricity market as a cloud analogy. Large enterprises went to private generators for 3-4 years before utilizing public services. Likewise, cloud will start internally (with private clouds) and enterprises will also start to experiment with public services. Then, after year 4, price points will be low enough and security concerns will abate enough so workloads can move to a public cloud, reducing the size and weight of the internal infrastructure.
Cloud Computing Journal: How big an issue is security for enterprises who wish to migrate toward Cloud-based infrastructure wholly or in part?
Liu: In private cloud environments, it’s no different an issue than it is for their current data center. The bottom line is that large enterprises must move to this type of setup within the next few years. They will be getting calls from service providers who will be able to quote pricing in the single digit per hour range, and they simply must be able to react by either implementing a private cloud or going with a public service. This means understanding what their costs really are, having that data figured out, and then also being able to deliver more efficiently.
Private cloud is really the only hope for companies who want to optimize internally before computing externally . In public clouds, we’re really dealing with 2 types of scenarios: (1) the consumer focused, Amazon EC2 type of cloud. Here, security concerns are being reduced due to requirements from the users, Amazon has a VPN tunnel to alleviate security concerns, for example. But our sense is that the majority of enterprises will go with (2) the long-term relationships the already have with service provider who can offer a semi-dedicated cloud, where they are fully aware of security, infrastructure and the like. This provides control and peace of mind while still delivering the benefits of a cloud environment.
Cloud Computing Journal: How big a part are standards going to play in the success of the Cloud?
Liu: In private cloud, and also in semi-dedicated clouds where I think the majority of enterprises will go, this is less relevant. For consumer-focused public clouds, there are benefits to standardization especially if you are an ISV who wants to port your applications to the cloud. The biggest problem right now is that applications must be written in a 3 tiered way to enable scaling in a cloud. That has to become standard. This offers complete portability of applications which will become a requirement in the public cloud.
Cloud Computing Journal: Which do you see as being the fastest-growing sector, Private Clouds or Public Clouds?
Liu: For smaller enterprises, I’d definitely say public. The value proposition is incredibly compelling, the technology is ready, and the barriers to adoption are being driven down. For larger enterprises, private cloud is where the growth is happening. This is being driven by competition from public clouds – meaning, large enterprises must go public or figure out a way to operate as efficiently as public clouds do.
Cloud Computing Journal: Finally, what's you top tip, as a seasoned software executive, to those other CEOs out there right now with respect to migrating (or not!) to the Cloud?
Liu: My top tip is: Know the landscape and pick your best approach based on who you are. Look at the cloud world as either private and public. For public, view the options as semi-dedicated public or fully public (consumer-focused) offerings. If you’re an SMB, you should be looking at public cloud computing now. If you’re a large company you should be experimenting with public offerings but at the same time moving rapidly toward private cloud transition within your own environment.
My other tip is: Beware! There are plenty of amateurs out there. Align yourself with seasoned experts who know what they’re doing and have been working these types of environments for years. There are 2 types of misinformation out there: from startups who are just entering the space with new technology, and from client-server infrastructure software companies trying to convince you they can take their tools and apply them to cloud. Be sure you choose vendors to work with who are neither brand-new to the business or so entrenched in one area that they can’t be agile enough to adjust to what cloud users need.
- The Top 150 Players in Cloud Computing
- Dolphin Announces Open API With Over 50 Add-ons Including Dropbox and Wikipedia
- i-Technology 2008 Predictions: Where's RIAs, AJAX, SOA and Virtualization Headed in 2008?
- The Top 250 Players in the Cloud Computing Ecosystem
- Success, Arrogance, Rise and Fall
- Cloud People: A Who's Who of Cloud Computing
- Cloud Computing Expo 2009 West: Call for Papers Now Closed
- Cloud Expo New York Call for Papers Now Open
- Cloud Expo Europe 2009 in Prague: Themes & Topics
- The Top 100 Bloggers on Cloud Computing