Up in the clouds of SOA

Posted May 7, 2008 at 12:15 pm by Eelco van Beek in Cloud Computing, Infrastructure.

In the past everything used to be better, my dad keeps reminding me. In the olden days: when CPU-cycles still needed to be applied for and bugs were real insects, things were a lot more manageable.

The mainframe was an entity with computing power, available – on demand – to run certain (business) processes. After a short period where servers and desktops handled a more individual approach, we seem to have partially returned to the old ways of thinking where the internet takes over more service and CPU roles. Google, Microsoft, Yahoo, and many others, focus primarily on this change under the phrase ‘cloud computing’. Besides supplying the more basic IT requirements (such as Office, CRM, email, etc.), not only an application is made available, but also (parts of the) the interface, to use this application as a service and to link it to internal or external processes. And that’s where it becomes really interesting.

Costs get reduced drastically (you no longer have to develop a complete system yourself, and by sharing, you also share costs), environment availability services are usually better (no need for maintenance and redundancy of your own infrastructure), and due to more knowledge and experience matters such as security are also better taken care of. Naturally, many lawyers will get nervous at the idea that functionality is realized from a wide range of linked services. Especially when they learn these services will not be delivered by in-house computer systems.

Where is the line of responsibility, and therefore liability, drawn? It can even be taken a step further. Nowadays, Amazon offers ‘storage-as-a-service’  through S3 and through EC2 ‘server-as-a-service’.  Google joins in and offers (for now) Google App Engine, which amounts to an application platform based on Python-as-a-service’.

Long term, this allows a much more abstract work level. Development of functionality (or application) will – apart from some small components – be carried out by people who use different generic services to link functions.  When an EC2 or GAE platform gets used to create specific application components, these same services can simply be offered again to other companies. Huge IT cost savings lie in waiting.

But obviously there are some risks as well. Clouds tend to merge together; smaller cloud computing providers will be swallowed by a small group of large companies, offering a wide range of functionalities. This creates a position of power.

But anyway, such dependencies already exist; because each company also needs electricity to deliver its services.

In addition, the principle of the weakest link causes all services depending on each other need to be arranged on the same service and response level.

The quality of the final application, which consists almost completely of linked services, shifts nearly entirely from development, programming and maintenance of software to writing and monitoring service level agreements.

Everything was better in the past, but the past isn’t today’s yesterday for nothing.