They have launched new high memory instances with 34 & 68 GB RAM with lots of CPU power thrown in.

They are lowering the barrier to use their DB service by allowing easy MySQL import/export.

I think the capex/opex argument is very luring for any CIO to atleast consider looking at SaaS/Cloud based applications.

However, fundamentally the business needs to answer the question as to “what value do I gain by owning an asset?” If the application/software/process under consideration does not translate into a strategic differentiator for a Company, it may make more sense to transfer the asset to someone else’s book.

In fact, besides salesforce.com there are other success stories in this space such as ADP (payroll), FieldGlass (labor procurement) etc. who have rolled out successful business models within their area of specialization.

@yogesh, I like the JIT analogy.

@manas, Other obvious issues with Cloud Computing in addition to what you already outlined are, data security, uptime and SLA compliance, governance and change management.

The essence is same: You have less and less inhouse and more and more being delivered over the network (in case of JIT or outsourced manufacturing, the ‘network’ was roads, seaways and airways).

So the experience says that cloud computing is here to stay just like JIT did. Not to infer too much from this parallel, but the apps that will get the most return from the cloud are the apps that will benefit from large global economies of scale, that are common to literally all the business world, and need some customization, but not a lot (just like cars!). So financial apps, expense report apps, recruitment apps, and similar stuff should already be moving to the cloud to realize major economies of scale across large as well as small customers. Somehow I see salesforce, IBM, and Oracle doing this better than amazon or Google. Think captive customer relationships and sweet transition deals.

The parallel also says that if network transport becomes more expensive, end user prices will go up. e.g. If power and bandwidth cost continues to go up, we can expect higher prices on the cloud and flexible contracts which allow vendors to increase prices quickly. However this may happen in the long term, not now.

The above discussion on Cloud pros/cons is heavily centered on databases. Has anyone run compute-heavy apps on the cloud? What is your experience? (I am expecting somewhat different weightages regarding IO speed, machine CPU speed, storage, RAM etc.)

Regards,
-Chaitanya

Horizontal scaling is non-trivial in the RDBMS world. What I meant to say was 15GB is still a lot of RAM. I think there would be very few applications which would require more than 15 GB and less than 64 RAM upfront.

SimpleDB or Simple Queue are different. I won’t put my data on SimpleDB or Simple Queue. I see them more as SaaS delivered from the cloud rather than a fundamental attribute of “the cloud”.

But again, S3 is not so bad.

I don’t have any hands-on experience with horizontal scaling, but I think, it’s not trivial. And you don’t want to spend time at the beginning working on a future problem.

The vendor lock-in kicks in if you are using platform specific tools. For example, if your application relies on Amazon SimpleDB or Queueing Service, the switch is not going to be as simple.