Is Amazon’s strategy a winner?

Amazon's EC2 and S3 services have received a lot of attention and have been hyped as a winning strategy and innovative business model. Let's take a look at the
two most discussed technologies – EC2 (Elastic Computing Cloud) and S3 (Simple Storage Service).

EC2

Essentially, EC2 partitions the server – using Xen's hypervisor – into as many partitions as there are CPU cores, each given a fixed (not elastic) amount of memory and local storage.

Let's compare EC2 with offerings from lots of service providers available for a long time called VPS (Virtual Private Server) or VDS (Virtual Dedicated Server). VPS/VDS offerings use our server virtualization technology SWsoft Virtuozzo that actually created this whole market. The ability to oversubscribe memory and disk resources, as well as sophisticated resource management capabilities (including boost limits), result in a really "elastic" service that consumes no resources while idle but can utilize everything available when needed.

The major difference between a regular VPS/VDS service and EC2 is the ability of the latter to create new instances via API calls, which is also provided by some of our hosting partners. As for the price, I looked at the web sites of several of our service provider partners, and compared to the numbers I get from AWS Simple Monthly Calculator, Virtuozzo-based VPSes are an order of magnitude cheaper.

S3

S3 is a redundant reliable storage, optimized for working with BLOB objects. Many Bubble 2.0 startups hailed it as a great tool in building their mega-scale web sites. S3, indeed, seems to be a great storage-in-the-cloud service, with a few caveats:

• An application needs to be re-written to use S3 as primary storage – S3 does not support regular file system APIs
• Re-writing an application is not straightforward. S3 has a very different programming model with radically different performance assumptions.
• Most applications use relational databases, and neither MySql nor SQL Server work on top of S3 – application storage layer has to be completely rewritten.

Overall, S3 will only work well for new applications written by highly-skilled developers who understand the tradeoffs of using local disks versus stream-oriented high-latency network storage. Existing applications cannot take advantage of S3 without major effort. S3 makes a great storage technology for disk-to-disk backups, though. Albeit, a little expensive, at $0.15/GB/month – not including the traffic. Hard disks cost $0.20/GB in retail, double that for redundancy, appreciate over 3 years and you get a little over $0.01/GB/month.

Hardware-as-a-Service is exactly what our partners - hosting providers – have
been offering for more than a decade. And unless I'm missing something here, Amazon is charging a very high premium for the services long available from majority of the hosting providers that just didn't happen to market them in quite the same way. Can Amazon's market reach and brand spin an old idea into a winning strategy? I don't know. What I know is that in most cases it is cheaper and less risky to rely on the hosting providers for storage and computational needs.

Shameless plug – check out my Server Virtualization Blog.

Comments

[Chris]

you didnt got the main idea of EC2 and S3.
virtually no limits in resources, in a grid that runs one of the most successful ecommerce sites in the world.
together with the automisation (hey thats your biz, ain't it?)
it means one can scale rapidly in a moment - without paying for the availability for the resources...