March 28, 2011
Amazon announced today that it would take one step past its recent announcement of their enhanced Virtual Private Cloud (VPC) offering and provide the option of dedicated hardware for customer applications.
According to Amazon’s Jeff Barr, the company’s VPC services weren’t enough for some customers who required network isolation. Some users expressed worry about the fact that other companies could be running on the same host--one of the most frequently-cited cloud concerns.
Certainly, as with any other on-demand instances that aren’t operating in a centralized datacenter that belongs to one's organization, this announcement still does not alleviate all concerns about using cloud-based (read as 'remote' resources). There's still no way to have complete control over data, but this can be something of a balm for both perception purposes and more concretely, for compliance.
For those who need computational resources and don’t have the wherewithal to purchase and maintain servers or a cluster, this is an ideal offering, in part because it speaks to the perceived security concerns that arise in a multitenant environment even if users might still contend with worries about loss of control or resource centralization. The need for resources versus the inability to invest in hardware have been the central attractors to the public cloud but for those who were hesitant before due to the multitenancy argument, this might finally make the clouds less...shady.
With the current VPC option, users will be able to select whether they want to spin up a private cloud that combines dedicated and normal AWS instances or if they want to just run their application on dedicated hardware.
Barr notes that when it comes to virtualization and the availability of virtual private clouds, even though the company already uses a sophisticated version of the Xen hypervisor to ensure that users are completely isolated from one another, customers have been hesitant.
In essence, this type of service is roughly the same thing that HPC users could get from an HPC on-demand service providers. R Systems, Cycle Computing (which pulls its bread and butter out of the Amazon empire), Sabalcore, SGI’s Cyclone---the list goes on—all have faced stiff competition from users who need resources without the hardware investment but now instead of standing on the security and regulatory grounds to show that their services are more appropriate, these HPC on-demand providers will need to beef up other claims as well as provide clear signs that their pricing models are more attractive.
Comparing pricing for Amazon’s Dedicated Instances running for HPC workloads will take some application-dependent math and guesswork but taking it a step further, using that information to compare the Amazon Dedicated Instance to the host of other HPC on-demand pricing models is going to be a challenge due to the completely different ways each provider prices their offerings.
The pricing for this new option is somewhat different than the way Amazon bills for some of its other cloud services as it is split between two fee structures. On the one hand, users will incur a fee for each instance and then will also pay what they call a “dedicated per region fee” which is a flat $10 per hour although isn’t contingent on how many instances are running in any particular region. This regional fee varies by where the instance is running and also changes according to the OS and performance level.
As Barr states, “When you launch a dedicated instance, we can’t use the remaining ‘slots’ on the hardware to run instances for other AWS users. Therefore, we incur an opportunity cost when you launch a single dedicated instance. Put another way, if you run one Dedicated Instance on a machine that can support 10 instances, 9/10ths of the potential revenue is lost to us.”
An additional option is one-time payments per instance based on a contract with Amazon in which the user would get a discount on the hourly use fees.
What’s worth noting here, however, is that while it might be difficult to understand exactly how the pricing models compare with another, one thing that users will need to factor into any decision or comparison is the support angle.
While the Dedicated Instances might be attractive from a cost perspective after the heavy lifting process behind price comparing is finished, remember that HPC on-demand providers factor in a very important element that Amazon doesn’t provide—support. And lots of it.
Put another way, Amazon simply provides you with the hardware and counts on your own internal expertise but for some applications, bare hardware and basic instructions aren’t going to cut it.
While it might be tempting to think that on-demand providers are going to be further spurred to beef up their offerings and take another look at their pricing, it’s probably more realistic to see the critical support factor becoming central to their messaging and packaging—and rightly so.
Posted by Nicole Hemsoth - March 28, 2011 @ 10:55 AM, 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?