You mention that some of these solution could be used in the Financial industry. I would be cautious of using these, especially since some are eventually consistent. If you are just tracking data these may be fine though.
Most of the financial sector is eventually consistent. I'm not just talking about traditional banks, I'm also talking about the markets as well. Most revolve around batch settlement processes where consistency essentially occurs on a schedule. While at the micro level, the components of these systems are somewhat built on ACID databases, that becomes irrelevant when as a system, it functions as eventually consistent. Dynamo itself is based on the fully ACID BerkeleyDB at the node level.
But facebook recently dumped it for inbox feature and began using Hbase instead. Not sure if Sequoia backing Riptano is supposed to be a bug or a feature?
Comments
You mention that some of these solution could be used in the Financial industry. I would be cautious of using these, especially since some are eventually consistent. If you are just tracking data these may be fine though.
Most of the financial sector is eventually consistent. I'm not just talking about traditional banks, I'm also talking about the markets as well. Most revolve around batch settlement processes where consistency essentially occurs on a schedule. While at the micro level, the components of these systems are somewhat built on ACID databases, that becomes irrelevant when as a system, it functions as eventually consistent. Dynamo itself is based on the fully ACID BerkeleyDB at the node level.
I agree with that, but you know, it's very hard to find good examples. :) I'm open to better suggestions :)
Some points on Cassandra:
- Facebook designed it for inbox feature - - SOLR/Lucene is being integrated - recently Sequoia backed Riptano - see http://www.riptano.com/
But facebook recently dumped it for inbox feature and began using Hbase instead. Not sure if Sequoia backing Riptano is supposed to be a bug or a feature?