Sorry Axton, I disagree regarding RAID 5.  It may be suitable for general
storage and backup, but RAID 5 is a performance-killer for databases.  If
reliability (fault tolerance) is important, use RAID 0 plus RAID 1 (aka,
RAID 0+1 or "RAID 10"), which may actually outperform RAID 0.  RAID 5
performance doesn't even compare.
 
References:
 
     http://www.dmreview.com/article_sub.cfm?articleId=2477
     http://www.lifeaftercoffee.com/2006/08/31/raid-5-and-oracle-databases/
 
http://searchoracle.techtarget.com/ateQuestionNResponse/0,289625,sid41_cid46
6681_tax285650,00.html
     http://www.sql-server-performance.com/sql_server_performance_audit3.asp
 
http://www.dell.com/content/topics/global.aspx/power/en/ps3q99_raid?c=us
<http://www.dell.com/content/topics/global.aspx/power/en/ps3q99_raid?c=us&l=
en&s=esg> &l=en&s=esg
     http://www.microsoft.com/technet/prodtechnol/sql/2005/physdbstor.mspx
 
http://www.microsoft.com/technet/prodtechnol/sql/2000/maintain/rdbmspft.mspx
 
http://www.microsoft.com/technet/prodtechnol/sql/70/proddocs/diag/part1/7552
8c01.mspx?mfr=true
     http://www.eisenschmidt.org/jweisen/misc/storage.pdf
 
Pavan, I suggest you research further to see for yourself which RAID
strategy makes the most sense for your database.
 
-- Bing
 
Bradford Bingel ("Bing")
ITM3 California
http://www.itm3.com/
[EMAIL PROTECTED] (email)
925-260-6394 (mobile)


  _____  

From: Action Request System discussion list(ARSList)
[mailto:[EMAIL PROTECTED] On Behalf Of Axton
Sent: Thursday, January 25, 2007 8:30 PM
To: arslist@ARSLIST.ORG
Subject: Re: Reg: Hardware Compactibility


** I would avoid raid 0 for db usage.  One drive failure means you will be
doing a restore.  If you have a good raid controller, raid 5 is the best
bang for the buck.  If no raid 1+0 is slightly more expensive, but has the
best of all worlds (striping for performance, mirroring for redundancy, and
no parity calculation overhead). 

Strongly agree with point 3.  With trunking, there are several arrangements
you can consider, round robin, load balanced, failover, etc.

Networks are fun.

Axton Grams


On 1/25/07, Bradford Bingel <[EMAIL PROTECTED]> wrote: 

** 
Three additional suggestions you might want to investigate further:
 
1.  On your database server, assuming "disk1" is where the database
physically resides, consider using multiple smaller drives with RAID 0
(striping).  Database performance typically improves -- sometimes
significantly -- when the read and write operations are divided among
multiple "spindles" (physical drives).  RAID 1 (mirroring) won't give you
any performance improvement, and RAID 5 (striping with distributed error
correction parity) should be avoided for any database application.
 
2.  Assuming your (Remedy) application server and database server have
multiple NIC's (or a NIC with multiple ports), consider Ethernet trunking
(or "link aggregation").  Properly tuned, trunking offers the twin
advantages of a) improved performance and b) link redundancy (if one of the
Ethernet ports fails, the remaining will continue to handle the network
traffic).
 
3.  If your server NIC's and switches support it, considering using "jumbo"
Ethernet frames.  Recall that the maximum size for an Ethernet frame remains
1,518 bytes, regardless of the link speed (even gig Ethernet), so even a
modest 100Kb data transfer would take dozens of Ethernet frames to complete.
Years ago, Alteon Networks (later acquired by Nortel) demonstrated that
jumbo frames typically doubled throughput, while cutting processing time in
half.  Since then, most of the NIC and network manufacturers began offering
jumbo frames as an option.  I would encourage you to learn more and
determine if it makes sense in your environment.
 
-- Bing
 
Bradford Bingel ("Bing")
ITM3 California
http://www.itm3.com/
[EMAIL PROTECTED] (email)
925-260-6394 (mobile)


On 1/25/07, Pavan Kumar Av (Consultant) <
<mailto:[EMAIL PROTECTED]> [EMAIL PROTECTED]> wrote: 



Hi List,



          As of now my server's are of below configuration. We have gone
live for about two quarters & are working just great. Going forward my
client is expecting ticket load as given below: 



Peak License usage:               100

Incident Tickets per year:   120,000

Change tickets per Year:       5,000



Please assist me if the below Hardware configuration with hold good or not.
If yes, then for how long. Also let me know what other parameters I can look
forward to tune my server to avoid any over loading. Thanks in advance to
all. 




Server

Windows Version

Model

CPU

Memory

Internal Disks


Web Server

2003 Server

DELL

8 CPU, Intel Xeon 3.16 Ghz

4 GB

Disk0 -  68.24 GB, Disk1 - 136.48 GB


Application Server

2003 Server

DELL

8 CPU, Intel Xeon 3.16 Ghz

4 GB

Disk0 -  68.24 GB, Disk1 - 136.48 GB


Database Server

2003 Server

DELL

8 CPU, Intel Xeon 3.16 Ghz

4 GB

Disk0 -  68.24 GB, Disk1 - 136.48 GB



Note: No Clustering or load balancer available as of now. 

 

Thanks & Regards, 

Pavan Kumar AV - [Remedy] 
HCL - AutoDesk 
Work: 408 416 0170 Extn: 5576
Mobile : +91 98409 95070

Home: +91 44 4359 0919 

Mail To: [EMAIL PROTECTED]   


_______________________________________________________________________________
UNSUBSCRIBE or access ARSlist Archives at www.arslist.org ARSlist:"Where the 
Answers Are"

Reply via email to