1) DO NOT use Memcached as a datastore. It is a cache. You can not reliably store data in it.
2) Your solution with a load balancer and repcached is EXTREMELY WEIRD. Best practice is to just use the servers straight up and letting the clients choose which server to store their data on. With your current solution, adding a memcached server will affect the performance of the app negatively, thereby completely eliminating the linear scalability of memcached, which is the whole purpose of the technology in the first place. 3) How on earth do you suppose that people on a mailing list would be able to accurately guess the optimal parameters for your connection pool, without knowing anything about the traffic patterns of your app? BENCHMARK it yourself, simulate your 700 users, test different connection pool sizes, and pick the one that's performing the best. 4) Try 17. That's a cute number. It should work just as well as any other number for your connection pool size for your app. /Henrik On Sun, Nov 11, 2012 at 6:51 AM, Ravi <ravikiranva...@gmail.com> wrote: > Hi , > > We are using Memcache as a dataStore for our high traffic web site .Under > which there may be 700 concurrent users at a time . > > Our set is that , there are six web servers in front , controlled by a > Load Balancer hitting two Mmecache servers which are in cluster . > * > * > *String location = server1:11211 server2:11211 * > *MemcachedClientBuilder builder = new > XMemcachedClientBuilder(AddrUtil.getAddresses(location));* > *builder.setConnectionPoolSize(5);* > > > Till now i am using the setConnectionPoolSize to 5 . > > My questions are > > 1. Is setting ConnectionPoolSize to 5 is correct in my case ?? > 2. Is setting ConnectionPoolSize factor impacts the application > performance anyway . > > > > > > > >
