June 07, 2010
Last week Microsoft announced that around 10,000 customers were using their relatively new Azure platform. The company's general manager for Azure, Doug Hauger, was careful to address the fact that while that is the approximate tally of customers, many are running services for a large set of users, thus the number of users likely far surpasses this figure. Still, this is an impressive number for a platform that only went out of its year-long beta in February of this past year.
While this is still a far cry from the anticipated many millions of technical computing users that exist in the much-discussed missing middle of HPC who could be ushered in by the promise of its developer-friendly cloud, the news here does not reside in the numbers necessarily, but how they are translated into practical use.
What Microsoft is noticing about these 10,000 customers and their associated use patterns is worth noting for the sake of everyone from potential end users in the technical and enterprise computing space to ISVs to, quite frankly, anyone who wishes to compete for the same broad base of users. HPC and enterprise are converging and accordingly, what's relevant news for one is equally so for the other as far as cloud platforms and application development challenges and successes are concerned.
Let the Revelations Begin
In an address at the Cowan and Company Technology, Media and Telecom Conference in New York last week, Hauger (Microsoft's General Manager for Windows Azure) provided a critical breakdown of who their customers were, at least in the general sense, and how they seemed to be deploying its cloud offering. Predictably, end user details were somewhat vague in terms of application-specific details, but actually offered more in the way of end user adoption uses than other companies in the space are willing to divulge. The company was able to shed some light on the current state of cloud adoption models and general use patterns by providing these details, which actually makes it far easier to comprehend real cloud deployment outside of the rhetoric about adoption that is usually offered when companies with similar offerings discuss their cloud products.
The following is a description of what Microsoft has been noticing with Azure, which is categorized into five key use patterns: on-and-off batch job computing; resource gains for small start-up operations; predictable bursting; unpredictable bursting; and perhaps most surprising, steady state operations.
Azure for On and Off Batch Job Computing
To put this in appropriate context, Microsoft uses the example of risk analysis and management and the deployment of a Monte Carlo simulation on trading data. One unnamed company was previously running this exclusively on Windows Server High Performance Computing Edition and now instead of that exclusivity, it still runs on Windows sometimes but on occasion it is batched out to the Azure platform where they can then scale out to thousands of instances at a time for a sort of "burst" mode but only for this one specific application.
Startups and Azure
It's always a pleasure when cloud vendors name names in the case of startups being enabled by the cloud, and Microsoft mentions French retail analytics company Lokad to highlight Azure's capability for the small business that has the capacity to scale. This firm started off very small, not wishing to invest in its own datacenter, and has now grown exponentially, thus making the scalability coupled with the lack of up-front costs to invest in its own cluster a perfect scenario. This appears to be one of the key areas for cloud in general outside of what we think of as traditional HPC, especially as the volumes of data continue to mount.
Unpredictable and Predictable Bursting in Azure
Although Hauger lists these two separately, they have so much in common that coupling them together allows these two possible uses to play off one another rather nicely. In terms of unpredictable bursting and the advantage it grants, Microsoft cites Kelly Blue Book which is a vast database of prices that is dynamic in nature. During the U.S. Cash for Clunkers program, suddenly everyone wanted to check on the real-time value of new cars. The way the company structured its Azure platform was rather interesting because instead of handling the entire traffic flow in the cloud or only utilizing it for certain workloads, one in four visitors was directed to Azure and if demand increased dramatically, the potential to scale out their front end was granted seamlessly. This case highlights perfectly the strategy for many Web 2.0 businesses who are looking to the cloud to handle peak workflow and unexpected hikes in visitors or customers.
Conversely, some businesses are able to anticipate surges like Domino's pizza, for instance. While yes, pizza is a bit of out of the scope for HPC the example -- whether we're talking about food delivery based on online ordering systems or engineering-related simulations delivered as a service during expected times of the year when demand is higher -- this is all important. Having the ability to provision for known events like audits, for example is one of the more widely-discussed advantages to being able to scale out to a platform like Azure, among others.
The Big Surprise
Hauger says Microsoft estimates that close to half of all current usage of Azure is "steady run-rate usage, where companies are moving their applications onto the platform, and just having them run in steady state." In other words, their customers are simply shutting down their on-site machinery and migrating to Azure, presumably to remain running something in steady state for the long haul.
What Early Analysis of Azure Use Indicates for HPC
In summarizing the use patterns for the first 10,000 customers (and who knows how many users) Hauger indicated that in his opinion adoption has been very good and surprising, "in that I expected the adoption at the lower-end with the startups and with sort of individual developers and small business ISVs as they get in to the VAS business. Where we're also seeing adoption is at the high end where we're seeing enterprise companies coming in and actually moving into cloud computing." He attributes this surprising finding to the economic situation, especially as it existed at the launch of Azure suggesting that this force prompted enterprise to look carefully at the challenges, benefits and risks of a multitenant environment. With this powerful economic incentive they were more willing to look closer at what the regulatory and compliance environment was in practical context versus overlooking cloud alternatives out of fear and misinformation.
Why Emphasis on HPC and Azure is Critical to Microsoft
Hauger notes that in terms of the better-than-expected adoption at the high end, Microsoft is seeing Azure as a suitable place for HPC applications. "It's actually net additive for Microsoft because we're selling our highest end server SKU on premises and we're able to leverage our economies of scale and we have the the benefits of us running this highly efficient platform out in the cloud. And so we get the benefit of both."
One of the questions that naturally emerges is how to move existing workloads into this cloud, which is a particularly relevant one in the HPC space. Hauger waffles for a moment, giving the standard answer to this question for any cloud vendor—that the ease of transition is dependent on too many variables to address in generalities—but then does make a few succinct points, claiming that "if your application is a big, huge hairball and very expensive to run, it's still going to be a big, huge hairball over here in this IaaS space and be difficult to run and expensive. And that's where if your application is architected and then sort of specific to Azure, if you have a stateless well-architected service-oriented application that's written in .NET on premises, moving that over to the Windows Azure platform is a simple thing."
Well, okay..."if" is the key word here, but this is an expected barrier for cloud adoption more generally, which is why Microsoft does seem to be doing a decent job through its Visual Studio offerings, for example, of making these things a bit more palatable. In terms of this issue, Hauger states that despite some of these challenges, the advantage they have with their Azure offering is based distinctly on the developer community. "We've come at this as being truly committed to building a platform, a cloud platform, for developers. For Microsoft, I think, more than anything else, our DNA is developers. And we've built a platform that really addresses their needs, making their lives easier to get into the cloud computing space. And there very few, I would actually say probably no other companies out there, that have built a cloud platform that are truly developer companies."
If Microsoft is looking to differentiate itself by the value it provides to the developer, it will be interesting to see how their competition steps up to the plate to address similar concerns. Without a doubt, the sticky issue of getting applications into the cloud -- which may sound basic now that there's already a platform for them to run on -- seems like a top topic to tackle as we continue to evaluate how the cloud is being deployed in more application-specific contexts.
Note: The full webcast of the interview with Doug Hauger can be found here.
Jun 19, 2013 |
Ruan Pethiyagoda, Cameron Boehmer, John S. Dvorak, and Tim Sze, trained at San Francisco’s Hack Reactor, an institute designed for intense fast paced learning of programming, put together a program based on the N-Queens algorithm designed by the University of Cambridge’s Martin Richards, and modified it to run in parallel across multiple machines.
Jun 17, 2013 |
With that in mind, Datapipe hopes to establish themselves as a green-savvy HPC cloud provider with their recently announced Stratosphere platform. Datapipe markets Stratosphere as a green HPC cloud service and in doing so partnering with Verne Global and their Icelandic datacenter, which is known for its propensity in green computing.
Jun 12, 2013 |
Cloud computing is gaining ground in utilization by mid-sized institutions who are looking to expand their experimental high performance computing resources. As such, IBM released what they call Redbooks, in part to assist institutions’ movement of high performance computing applications to the cloud.
Jun 06, 2013 |
The San Diego Supercomputer Center launched a public cloud system for universities in the area designed specifically to run on commodity hardware with high performance solid-state drives. The center, which currently holds 5.5 PB of raw storage, is open to educational and research users in the University of California.
05/10/2013 | Cleversafe, Cray, DDN, NetApp, & Panasas | From Wall Street to Hollywood, drug discovery to homeland security, companies and organizations of all sizes and stripes are coming face to face with the challenges – and opportunities – afforded by Big Data. Before anyone can utilize these extraordinary data repositories, however, they must first harness and manage their data stores, and do so utilizing technologies that underscore affordability, security, and scalability.
04/02/2012 | AMD | Developers today are just beginning to explore the potential of heterogeneous computing, but the potential for this new paradigm is huge. This brief article reviews how the technology might impact a range of application development areas, including client experiences and cloud-based data management. As platforms like OpenCL continue to evolve, the benefits of heterogeneous computing will become even more accessible. Use this quick article to jump-start your own thinking on heterogeneous computing.