2 Replies Latest reply on Aug 14, 2014 1:43 AM by 2650877

    bdb efficiency in centos


      I use bdb to store big data in centos. each key/value is almost 10M.and a db file is usually more than 100G.

      1. when i use db::put to insert an key/value , it usually cost 2-3 seconds.however,windows, it cost less than 1 second.

      The same problem where i read from db

      2. in some machine(with centos),  i set cache_size = 128M, db::put cost more  than 30s .  how  cause this?

      i create db environment with DB_INIT_CDB, no other parameter was set.

        • 1. Re: bdb efficiency in centos

          from your description is looks like there is an issue with the machine.     We do not run with CentOS in our testing environment so this isnt going to be something that we can provide more details on   The BDB code base is the same on all those machines and if you have one machine where the performance is different, then our recommendation is to look at the machines are figure out how they are different.  File system behavior can play a large part in IO performance.




          • 2. Re: bdb efficiency in centos

            I try to find out the different part, but it does not work. So i try to find out how many parameters to affect IO performance.


            1、BDB parameters: I just set cache_size, how does this to effect  db::put performance? any other parameters you suggest to set to improve the performance?


            2、OS parameters : can you list some parameters that may affect the performance? Specilally in linux.

            I has two machine, both in CentOS, but os in one of them has been customized. Before customize, application works well, but bad after customize.

            pagesize, parameters with command ulimit -a,  most parameters with sysctl are the same, and both ext3, blocksize is 4K.


            I have no idea which parameters should i fouce on. any suggestion is welcome.