Consider the problem though. Random access to trillions of records with no guarantee any one will be fetched twice in a short time frame nullifies the effectiveness of a cache unless the cache is enormous. If such a cache were that big, 100's of TB's, I wouldn't be looking at on-disk storage options. :)
-Greg
On Mon, Sep 26, 2016 at 6:54 AM, Yves Dorfsman <yves@zioup.com> wrote:
Something that is not talked about at all in this thread is caching. A bunch of memcache servers in front of the DB should be able to help with the 30ms constraint (doesn't have to be memcache, some caching technology). -- http://yves.zioup.com gpg: 4096R/32B0F416