Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Getting the best out of new server hardware: a disk setup for Oracle database

Getting the best out of new server hardware: a disk setup for Oracle database

From: Heikki Siltala <abcwebmasterxyz_at_abcheikkisiltalaxyz.abccomxyz>
Date: Tue, 13 Jun 2006 00:55:44 +0300
Message-ID: <upljg.2107$5x.30@reader1.news.jippii.net>


Hello all Oracle&server experts, what you think about the following setup for Oracle database?

HP DL585, 16 GB RAM, 2 dual-core AMD processors, 64-bit Red Hat Linux 4, 64-bit Oracle database software. The disk setup for Oracle data consists of six PCI-X RAID cards with BBWC set to 0 percent read, 100 percent write. Each card has six 146 GB disks on RAID 1+0 setup and one hot spare. 128 kB stripe size is used. Each RAID 1+0 set (about 410 GB) is divided into two logical volumes: 40 GB and 370 GB and these volumes are seen as raw devices at the server.

I'm planning to put all six 370 GB raw devices together into one diskgroup using ASM and ASMlib and let ASM stripe the database data (tables, indexes) over them (external redundancy). I'm planning to group the 40 GB volumes into two ASM failgroups (3 volumes each) and then let ASM do striping and mirroring over this diskgroup (normal redundancy). This smaller diskgroup would then store SYSTEM tablespace, control files and online redo logs. With this setup I get good protection for disk failures for all the data and additional protection against RAID card failures for SYSTEM tablespace, control files and online redo logs. There is a separate FC connected disk system that will be used for archived redo logs and backups. And separate mirrored disk pairs for OS and Oracle binaries.

Does this setup sound feasible? Could I make a better use of the hardware by changing something? The system will be used mostly for data warehousing weighted on reads and has no 24/7 requirement - some downtime can be tolerated. I have tested the disks using dd and orion, dd shows 235 MB/s read speed per six disk RAID 1+0 set, orion shows 800-1400 MB/s read speed over all the disks depending on the parameters. RAID card caches was set to 50/50 for these tests.

One additional question: if I configure "flash recovery area", where do the archived redo logs go? Do they go only to "flash recovery area" or will the area have just copies of the archived redos?

--
Heikki
Received on Mon Jun 12 2006 - 16:55:44 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US