Database in the Clouds?

By cwardell • May 27th, 2009

I am struggling with this one. For VLDB’s, I can’t say I subscribe to the cloud databases like EC2, SimpleDB, etc.. I love the idea of throttling up or down the capacity on demand but I am not sure it is practical in the query and reporting sense.

If I had a transactional system like Facebook, MySpace, Travelocity, Expedia, it would make much more sense for me. The data requirements are large but the data is accumulated over time, transaction by single transaction. Ad hoc queries, reporting, and analytics that happen on VLDB’s, usually, or at least, should not be performed against operational transaction based systems. There is usually an extract of the transactional data and it is housed in either a more suitable schema or better yet or more suitable VLDB technology.

Here are a few questions and challenges you may face:

- • How do I get terabytes of data from my operational system to the commodity clouds effective
- • I am not sure I would want to say to my customers: “Just send your data to Amazon.com and I will take it from there.” Imagine how many hands it needs to touch before it gets to the floor of the data center.
- • Are you in an industry (financial, pharmaceutical) where regulations are such that a multi-tenant grid would be frowned upon?

I definitely see the positive side if the data was already at the cloud, but typically, we receive data via ftp in many different formats at different frequencies. They all need to be standardized and purified before it gets into the Cloud DB. So from my point of view, data warehousing is not a real good candidate for the Cloud.

I love the idea of the cloud and would welcome differing opinions. Am I missing something?

Comments(0) Tech Talk