11 messages in com.mysql.lists.mysqlRe: mysql speed - raw i/o under linux...| From | Sent On | Attachments |
|---|---|---|
| David Haring | 02 Dec 1999 03:12 | |
| Ed Carp | 02 Dec 1999 03:24 | |
| srf...@mistik.net | 02 Dec 1999 10:42 | |
| Michael Widenius | 02 Dec 1999 12:03 | |
| Robert Seward | 02 Dec 1999 15:12 | |
| brdo...@oro.net | 02 Dec 1999 15:46 | |
| Paul DuBois | 02 Dec 1999 16:40 | |
| Sasha Pachev | 02 Dec 1999 16:44 | |
| brdo...@oro.net | 03 Dec 1999 06:02 | |
| Sasha Pachev | 03 Dec 1999 08:32 | |
| Steve Ruby | 03 Dec 1999 15:19 |
| Subject: | Re: mysql speed - raw i/o under linux, Hoard memory allocator, SMP![]() |
|---|---|
| From: | Robert Seward (rsew...@bluestone-consulting.com) |
| Date: | 12/02/1999 03:12:54 PM |
| List: | com.mysql.lists.mysql |
The Hoard memory allocator looks like you can drop it into any codebase and use it. It is just a replacement to malloc. Try compiling MySQL with it and benchmark it. If MySql performs better than without Hoard you know the answer to your question. And please share your results with list! ;-)
Raw I/O I have mixed feelings about that. The gains will be minimal at best, and you will lose a bit of stability when you crash. But it may help with 2 GB limits on file systems.
Thanks, Rob
srf...@mistik.net wrote:
On Thu, 2 Dec 1999, Ed Carp wrote:
As raw i/o support is available for Linux as kernel patch, I would like to know if mysql can be run on raw i/o device under Linux and, if so, how would this impact mysql performance. Did anybody try this?
I brought this up a while back. Seems it wouldn't run any faster. And I didn't know you needed a kernel patch to do I/O direct to a disk partition.
i think he meant raw i/o device, not using the system buffers (sendfile() and the like). i believe this feature is partially available in Linux 2.2.x - extended by the rawio patch, but it's planned as one of the main features in Linux 2.4.x.
you still need to use the facilities in mysql (for example) to get the desired performance. of course, mysql doesnt have these in place. but maybe in a future version (?) *hint* :)
======== Integrity Systems Professionals =================
Robert Seward - Oracle Consultant
voice: 800 542-9512 x27529 pager: 313 684-9843 email: rsew...@noplace.com




