Disk Partitioning/Binning

  • Hi All,

    Here's my post..My cmpy has bought new hardware for the datacenter of (5)-140gig hard drives.We are planing to have the new sql server and LCS servers 2005 on those drives.

    The set up is,we'll have an active/active cluster of SQL server 2000 and active/passive version of LCS servers.can anyone tell me,how I can group them and how the logical partitioning is to be made.Meaning,I just want to know,how the disk has to be partitioned for efficient performance of these servers on these drives.

    I'm new to this concept of disk partition and I've been involved in this work..can anyone help me with this..

    Thanks in advance..

    AK.

  • I've never been involved in anything quote as large, so I'm curious to see what others have to say as well..

    But, I would suggest that you have the transaction log on a fast RAID 0 volume and have the data on a RAID 5 or RAID 10 volume... You can split the database into different files to help spread the load amongst a variety of spindles.

    There's a LOT more to it than that - someone else will have more to offer I hope

  • Tht was a good and an encouraging start...

    Even I do wish,tht someone wud come with somethin more to offer.this wud be great start for many of the newbie's here as this wud be the starting point in any environment to start a process from scratch..

    also,i'll keep posting the best practices and the issues we encountered as we walk thru this process..

    but,anyone pls give me some acceleration..

    thanx..

    AK.

  • Hi Arun,

    I have collected details from members here and working on it as our company is heading for 2005 setup. I have 2 server [HP DL-380] with 6 drives and based on that here are my recommendations:

    On Drive 0, create a C: partitions of about 8 Gb for the swap file and software (Windows and SQL Server) and then use the remaining space to create a D: partition for the tempdb data file and log.

    On Drive 1, create one partition for the data file (*.mdf) with an NTFS cluster size of 64K

    On Drive 2, create one partition for the transaction log (*.ldf) with an NTFS cluster size of 64K

    As the performance of a disk is affected by fragmentation, one way to reduce fragmentation is to isolate static files, such as the Software, from dynamic files, such as tempdb. This way, you can defrag C: once and be done until you next modify the software files via a service pack. 

    For partitions housing SQL Server data files, the NTFS cluster size should be 64K but for file system, 4K is the typical choice.

    You can get the actual reads and writes for all database files by running

    SELECT * FROM :: fn_virtualfilestats(default,default)

     Keeping Logfiles and Database files on different RAID you will get better performance when you put them on different RAID's (physical disks). You'll also gain from having the log files on a RAID1/10/01 rather than on a RAID 5

    Hope this helps,

     

    Kindest Regards,

    Sameer Raval [Sql Server DBA]
    Geico Insurance
    RavalSameer@hotmail.com

  • That was sweeeeeeeet Sameer..

    That helps a lot.I've started working on keeping this as the base.will get bk,if incase i'm stuck somewhere...

    do've any idea about how the LCS 2005 set up has to be carried out.coz,it goes along with the SQL stuff on those drives..

    Thanks a lot to everyone and to the group..

    AK.

  • Hey,

    it's me again..Now,itz the next level..I believe these stuffs should be carried out by the system's folks,if am not wrong..but,my input is required too...but,am on my own now..

    We need to identify as to how many IP addresses are needed for SQL clustering (virtual IP, private, public, heartbeat etc) . The list should have the IP's for SQL and for LCS 2005 SQL.

    Any thoughts?

    have you got anything Sameer?

    Thanks,

    AK.

  • Here's what you need

    • physical machine A network card
    • physical machine B network card
    • virtual windows server
    • sql server instance
    • msdtc
    • physical machine A network crossover
    • physical machine B network crossover

    RegardsRudy KomacsarSenior Database Administrator"Ave Caesar! - Morituri te salutamus."

  • Rudy,

    Can u explain me a bit abt this..i don't understand this clearly..moreover we don't have msdtc running in our setup.

    does this pertain to the LCS servers too?

    Thanks,

    AK.

  • This is the minimum IP addressess needde for SQL2K, SQL2K5 should be the same since clustering is performed on the OS level. Here are a few links that you need to check out:

    Granted most of the information is for SQL 2K, but the majority of it should apply to SQL2K5 as well. Additionally, going through each link and some of their successors will enrich your knowledge base in this area.

     

    RegardsRudy KomacsarSenior Database Administrator"Ave Caesar! - Morituri te salutamus."

  • No one mentioned SAN yet; this is an active/active cluster, so your binaries/OS should go on a local mirror, the data files on the SAN.

    It's unlikely you'll hit any I/O limits on your SAN, but I have heard of it happening.

Viewing 10 posts - 1 through 9 (of 9 total)

You must be logged in to reply to this topic. Login to reply