April 21, 2010
When we're talking about strict hardware-related HPC, defining high-performance computing is usually straightforward. However, when we extend the concept of HPC into the cloud and then to even further complicate the matter by adding in discussion about how HPC and cloud are being utilized in commercial and large-scale enterprise class settings, the roots of those concrete hardware definitions start to peel away.
Every business wants supercomputer capacity on-demand. And who could really blame them? It seems that most enterprises need vast, scalable capacity to remain competitive. For smaller businesses, getting competitive out of the gate is finally an option since the law of "he with the most start-up capital for tech infrastructure wins" is on the wane.
After skimming some HPC and cloud-related news I chose, at first anyway, to ignore, I started to think about these things a little more in-depth. With a big group of engineers coupled with some general HPC backing, cloud power, and an Ultra Marketing Bot 5000 (not a real product but thought a Nexus 6 Publicity Model would have been too vague), could just anyone compete with a company like, say, Netflix?
If every enterprise's capacity suddenly becomes unlimited, then does it all just boil down to who has the best architects and the finest sales force to convince the world it's better than what already exists in droves?
Netflix Cloud Adoption in the News
When I first saw this New York Times news story about Netflix's shift to the cloud I was reticent to draw everyone's attention to it by sticking into the "This Just In" section here on HPC in the Cloud because it didn't seem…relevant. After all, this is the mail-order movie business -- the post office is involved, for crying out loud. Where's the gritty HPC in that?
But you know, I didn't think my omission of Netflix through, so I decided to go back and revisit it in this blog.
So rewind and let's retroactively pop this into the April 18th This Just In…
When Netflix announced that it had moved into the public cloud space and was housing some of its operations with Amazon Web Services, the burning question was to what extent their operations had gone to the cloud already and how much -- what percentage, that is -- would be heading cloud-ward over the two-year implementation.
The New York Times and others touted this as a use case of a large-scale data-intensive operation going full-blown into the cloud, but following an interview this morning with Steve Swasey, Vice President of Corporate Communications at Netflix, this shift into the cloud isn't quite as comprehensive as it seems -- at least not yet.
The company has moved some of its major power-gobbling processes to Amazon Web Services' public cloud, but the really good and juicy stuff -- that's all hoarded away on Netflix's own internal servers. And that's probably not going to change, according to Swasey.
From what it sounds like, Netflix took one particular type of process (the encoding of new film into the system to make it available as a stream on demand) and since it would have been foolhardy from a resource, time (and therefore cost) perspective not to do so, they plunked it down on AWS. While Swasey didn't go into detail about other major operations in the public cloud, he did suggest that anything remotely sensitive was stored in-house. Other search tools, customer queues, and more customer-facing (versus internal) will be hosted in the cloud -- but the buck stops there.
Netflix isn't just shipping and receiving movies using snail mail and some scattered company PCs on an internal network, after all. There are multiple arms of this business that require vastly different resources and that also require immense scalability since there are most likely times and days of peak demand for instantly-available streaming movies on a PC. When coupled with the other side of the Netflix operation -- the shipping, receiving and storing of films in over 50 centers throughout the United States that send automatic messages in vast quantities to its huge database of users alone is complex. When we factor in encoding video and then turning around to deliver it on demand, there are whole new levels of resource and scheduling issues.
According to Swaysey, the adoption of AWS took place quite rapidly; Netflix began testing Amazon's public cloud at the beginning of 2010 and analyzed test results to gauge progress. The company found that it significantly reduced dependency on its own data centers as well as cut back on time and engineering time, especially for one of the critical functions of the Netflix service -- taking the raw film from production houses and handling all of the encoding in-house so that the films could be streamed to customers on-demand at the touch of a button.
Until comfort in the cloud grows -- something that we will monitor closely here -- companies with wide-ranging, large-scale, data-intensive needs will likely experiment with the cloud to outsource resource-heavy operations but in the end, security and protection of sensitive data seems to be the biggest hurdle. Companies like Netflix are willing to bear the much higher costs of IT resources as they keep their special data close to home.
Does that mean that the quickest way for new enterprises who start with the cloud (versus having to be talked into migration) to get a head start is by taking the plunge and sending all (even the juicy, private, confidential, secure stuff) into a public cloud?
While Swayse was tight-lipped about anything in the way of specifics in terms of the compute environment, it just took some thinking-through for me to see that they are managing, scheduling, and balancing the same large-scale, data-intensive, mission-critical workloads "real" enterprises in science and research are -- the difference is, well, it's "just" movies.
If your startup costs were minimal and you relied on pure ground-up cloud architecture and a gaggle of really, really smart friends, would it be possible to compete with a Netflix? Will it all just boil down to who has the most innovate marketing versus the best capacity if everyone has unlimited capacity?
If only I had more time.
Posted by Nicole Hemsoth - April 21, 2010 @ 5:40 PM, Pacific Daylight Time
Nicole Hemsoth is the managing editor of HPC in the Cloud and will discuss a range of overarching issues related to HPC-specific cloud topics in posts.
No Recent Blog Comments
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?