Comments on: TokyoTyrant – MemcacheDB, but without the BDB? http://fewbar.com/2009/06/tokyotyrant-memcachedb-but-without-the-bdb/ Technology, life, and mischief, not in that order Wed, 09 May 2012 03:57:55 +0000 hourly 1 http://wordpress.org/?v=3.0.5 By: Shane Hallo http://fewbar.com/2009/06/tokyotyrant-memcachedb-but-without-the-bdb/comment-page-1/#comment-2898 Shane Hallo Wed, 09 May 2012 03:57:55 +0000 http://fewbar.com/?p=85#comment-2898 I stumbled uppon your blog, I must say it is great. Thanks for sharing all the info. I stumbled uppon your blog, I must say it is great. Thanks for sharing all the info.

]]> By: altex http://fewbar.com/2009/06/tokyotyrant-memcachedb-but-without-the-bdb/comment-page-1/#comment-48 altex Tue, 18 Aug 2009 15:21:01 +0000 http://fewbar.com/?p=85#comment-48 Hi. I made all the tests from your zip file (i used latest brutis, it's realy buggy, i made some patches for it), and two libraries libmemcached and libmemcache. 1) -------------------------------- libmemcache sometimes made sochet read errors when 20 clients was connected, so libmemcached is the only way to use to realy test hiload access to memcache protocol. 2) ----------------------------------- >> Of course, in nosync mode, memcachedb >> is just throwing data at the disk. It is not correct. In nosync mode it just operates data in memory throwing data to disk from time to time. 3) ----------------------------------- You realy should test not just cpu usage, but also disk usage (iowait). So tokyotyrant use disk much more than memcachedb-nosync. I've played with various tests for about a week, and i think memcachedb-nosync is the best solution for me (i need hignload solution for a big dating site). Thanks Hi.

I made all the tests from your zip file (i used latest brutis, it’s realy buggy, i made some patches for it), and two libraries libmemcached and libmemcache.

1) ——————————–
libmemcache sometimes made sochet read errors when 20 clients was connected, so libmemcached is the only way to use to realy test hiload access to memcache protocol.

2) ———————————–
>> Of course, in nosync mode, memcachedb
>> is just throwing data at the disk.

It is not correct. In nosync mode it just operates data in memory throwing data to disk from time to time.

3) ———————————–
You realy should test not just cpu usage, but also disk usage (iowait). So tokyotyrant use disk much more than memcachedb-nosync.

I’ve played with various tests for about a week, and i think memcachedb-nosync is the best solution for me (i need hignload solution for a big dating site).

Thanks

]]>