Amazon has introduced EC2 Reserved Instances, which allow you to obtain a reduced hourly fee for an upfront payment that varies according to which level of EC2 server you use. So, for example, with an upfront payment of $325, the hourly fee of 10 cents is reduced to 3 cents for a year. With a payment of $500, the fee is reduced for three years. Amazon's CTO, Werner Vogel, explained that the initiative was launched to address applications that are available for long durations, rather than being used for short periods of time and then taken down.
In essence, this is aimed at organizations that want to use AWS as an external data center for standard applications that are used 24X7.
This offering clearly enables a lower TCO, depending upon how much time the application is actually up. Algebra not your strong suit? Fortunately, Geva Perry, a well-known cloud blogger, created a calculator to allow you to figure it out automatically. Filling out a spreadsheet too much work? Geva bottom lines it: if your app is up more than 12 hours per day for one year, it makes sense to use a reserved instance. Obviously, the blended pricing gets even more attractive over a three year period, where the yearly upfront payment works out to $166.
An interesting variation for companies is this scenario: an application that must always be available, but experiences intermittent large increases in use. Obvious examples are seasonal retail, monthly reporting, etc. For these apps, a mix of reserved instances for typical load with on-demand instances for short duration peak load will provide a good solution that leverages the flexibility of AWS while reducing minimum operations expenses.
Since one of the typical objections to cloud computing is that its TCO is higher than self-hosted systems, this new offering changes the equation — it may not affect the outcome of that discussion in every case, but it definitely tilts the balance toward the cloud. I believe the argument that cloud computing is, ipso facto, more expensive is more and more tenuous every day.
Microsoft also announced a change in its Azure cloud offering: they now offer standard SQLServer relational database support. The money quote: “a developer can take an existing application and just change the connection string to point it to the cloud and have it just work.”
Another common objection to cloud computing is that it isn't easy to migration existing applications to the cloud. In particular, the PaaS offerings don't support standard relational databases or SQL syntax. Cloud advocates argue that this isn't a big deal, because applications are increasingly scale-oriented and run too large for relational databases anyway, which is why the key-pair persistence offerings of the PaaS vendors is more appropriate. I think they're dead wrong in this assertion. While there is no doubt that what I refer to as “Internet-scale” applications and data stores are becoming more common, requiring scalable infrastructures beyond the capability of most data centers, the vast preponderance of applications fall into traditional architecture infrastructures, which rely on SQL databases as the persistence mechanism. Moreover, there is huge potential demand for IT organizations to migrate these apps into cloud environments if the economics are sufficiently attractive.
Microsoft has really increased the attractiveness of its offering, especially to less-sophisticated IT organizations, which comprise a significant portion of the overall Microsoft user base. For those customers, being able to take an existing app and move it into the cloud, or easily install a new app in Azure, makes Microsoft's cloud service extremely appealing. With this announcement, Microsoft has created a hybrid IaaS/PaaS service that can address both traditional architecture applications as well as the new breed of Internet-scale applications. I have to say I'm really impressed, and especially with how quickly Microsoft has created this capability, in vast contrast to the extended delivery cycles we've experienced with many of their other products.
Overall, cloud computing offerings continue to evolve quite rapidly, making the “buy vs. build” of cloud vs. internal data center even more interesting.
Bernard Golden is CEO of consulting firm HyperStratus, which specializes in virtualization, cloud computing and related issues. He is also the author of “Virtualization for Dummies,” the best-selling book on virtualization to date.