How To: Check Couchbase Speed

How To: Check Couchbase Speed
Memcached is at the heart of Couchbase and its the worlds fastest and most reliable cache solution. Like any process or application on a server there are many varibles that go into determining speed. PRO TIP – When testing or evaluating Couchbase remember to install it on a vanilla OEM operating system image. I have seen many times Couchbase performing sub-par because a “standard corporate” image have other applications or setting that interfere with Couchbase operations. How do we measure the performance of Couchbase? The tool to use for performance information is CBSTATS. CBSTATS is a CLI tool ussually located in Couchbase server at /opt/couchbase/bin/cbstats This tools only relates to performance stats to only a particular node and bucket. The stats are for only performance on the couchbase server level side and does not have any data of client side stats. #./cbstats <IP>:11210 <command> -b <bucket_name> -p <bucket_password> <command> all allocator checkpoint [vbid] dcp dcpagg dispatcher [logs] failovers hash [detail] items kvstore kvtimings raw argument reset slabs tap [username password] tapagg timings vkey keyname vbid CBSTATS is on a per node per bucket bases. NOTE – This only measures speed at the Couchbase Server Application it self and not to the SDK. The specific command you want is timings #./cbstats 127.0.0.1:11210 timings -b <bucket_name> -p <bucket_password> This will give you a histogram of all the specific events. You will not be able to track down a single and specific GET() or SET() but you can see a history.Many times you will want to test and clear the history and test again. The best method for that is to use the reset command to clear the histogram. #./cbstats 127.0.0.1:11210 reset -b <bucket_name> -p <bucket_password> For more details about CBSTATS click here on the official documentation: http://docs.couchbase.com/admin/admin/CLI/cbstats-intro.html The source code for CBSTATS is here on GITHUB: https://github.com/couchbase/couchbase-cliArray Array

Fast Machine Learning with PB-BLAS & Michael Payne in HPCC Systems

Fast Machine Learning with PB-BLAS & Michael Payne in HPCC Systems
Watch Michael Payne , PHD student at Clemson University, at the 2014 HPCC Systems Summit talk about PB-BLAS (Parallel Block Basic Linear Algebra Subprogram). With PB-BLAS machine learning algorithms can be more efficient and faster by 6X or more. HPCC Systems Machine Learning Library has just been updated. https://github.com/hpcc-systems/ecl-ml     Wikipedia: https://en.wikipedia.org/wiki/PBLAS Detailed Paper: http://www.netlib.org/utk/people/JackDongarra/PAPERS/079_1996_pb-blas-a-set-of-parallel-block-basic-linear-algebra-subroutines.pdf Array Array