On Tue, May 1, 2012 at 9:06 PM, Edward Capriolo <edlinuxguru@gmail.com> wrote:
Also there are some tickets in JIRA to impose a max sstable size and
some other related optimizations that I think got stuck behind levelDB
in coolness factor. Not every use case is good for leveled so adding
more tools and optimizations of the Size Tiered tables would be

Agree. We trieved leveled compaction, but that didn't work well for our dataset and traffic. But something inbetween would probably be a nice fit for us, a compaction model where you specify a max SSTable size, and then your data would be split across (total size)/(max size) files. And extra bonus points for applying generational thinking to it so that your oldest data goes in the first file, next oldest in the next file and so on. That way you'll have a nice separation of "dead" data and "active" data. Of course I don't know if this is a better model than the current one, I'm just guessing here. :-)