Re: [squid-users] Tweaking Squid for speed, not max requests

From: Brett Lymn <blymn@dont-contact.us>
Date: Fri, 31 Aug 2001 10:03:13 +0930 (CST)

According to Brian:
>
>The major performance issue with ufs is that it blocks until the
>read/write completes, so it can't handle other tasks while it waits for
>the disk.

Errrrmmmmm.... not what I understood UFS did. The _metadata_ (file
create, directory update) is a blocking wait for disk operation sort
of call but normal reads and writes are done via the buffer cache. So
if you have enough memory the writes will just go to memory and get
flushed to disk later, the read cache has read-ahead which is
controllable so you can bring in disk blocks to fulfill future read
requests. You can decouple the metadata updates by either using a
logging file system (which does have downsides - especially on large
deletes) or log the metadata to another disk which you can do with
Sun's Solstice Disk Suite just to name one product.

> With a RAM disk, there's no wait involved.
>

Unfortunately, it is ephemeral - you reboot the machine and you lose.

-- 
===============================================================================
Brett Lymn, Computer Systems Administrator, BAE SYSTEMS
===============================================================================
Received on Thu Aug 30 2001 - 18:30:36 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:01:57 MST