[prev in list] [next in list] [prev in thread] [next in thread] 

List:       opensolaris-lvm-discuss
Subject:    [lvm-discuss] Re: Coalescing large writes to RAID 5 volumes?
From:       Sanjay.Nadkarni () Sun ! COM (Sanjay G !  Nadkarni)
Date:       2005-07-18 23:42:53
Message-ID: 18874186.1121755234978.JavaMail.suncom () oss1
[Download RAW message or body]

The 64K chunk makes sense since the stripe width is 64K.  If you want to futher \
improve the performance, change the stripe width to 128K or 256K.  To ensure data \
consistency in RAID-5 configuration, writes are by default "logged".   This is called \
the prewrite region.  Look for the keyword prewrite in \
http://cvs.opensolaris.org/source/xref/usr/src/uts/common/io/lvm/raid/raid.c.

Ofcourse the real question is what lies on top of this RAID configuration ?  If \
there's a file system (UFS logging), the writes are typically in 8K page chunks.   \
Reduce the block size to match the stripe width and you should notice better \
performance. This message posted from opensolaris.org


[prev in list] [next in list] [prev in thread] [next in thread]