+ Reply to Thread
Page 2 of 2 First 12
Results 26 to 38 of 38
  1. New Member royal's Avatar
    Join Date
    Jul 2006
    Location
    Chicago, IL
    Posts
    3,373
    #26
    A possibility is 2 disks for OS/Logs and 4 disks for Database in Raid 10. If you do this, make sure you create a dedicated partition for logs; not for performance so if their logs fill up the partition it doesn't cripple the OS. Or 2/2/2.

    Personally, I'd probably do 2 Raid 1 and 4 Raid 10 as database uses up quite a bit more IOPS than the logs do. Raid 1 = Bad Read/Good Write. Raid 10 = Great Read/Great Write. Logs need better write and Database needs good both but more so reads than write.
    Reply With Quote Quote  

  2. SS -->
  3. Senior Member
    Join Date
    Mar 2007
    Posts
    12,308
    #27
    That's exactly the info I was looking for; thank you so much
    Reply With Quote Quote  

  4. Senior Member
    Join Date
    Jan 2008
    Posts
    1,941

    Certifications
    MCITP: EA, EMA; MCSE 2000/2003: M; MCSE 2000: S; MCSA 2000/2003: S; MCTS: ISA 2006; VCP3/4
    #28
    You're gonna need far more RAM than that for E2K7. You need 4GB to run HT, MB, and CAS roles plus 2 to 5MB's per user depending on their usage. That's just for Exchange alone. Tack on additional resources for anything else like AD or AV running on the server.

    For that few users, you could get by with the tlogs on a mirror with the OS drive, and the DB's on their own disks.
    Reply With Quote Quote  

  5. Senior Member
    Join Date
    Mar 2007
    Posts
    12,308
    #29
    Yea, I think I'm going to get another pair of 2gb sticks and see how that works. Thanks again.
    Reply With Quote Quote  

  6. Cisco Guru mgeorge's Avatar
    Join Date
    Jun 2006
    Location
    127.0.0.1
    Posts
    800

    Certifications
    A few...
    #30
    I never was a big fan of microsoft upgrades per se, "upgrading" stuff seems to cause more problems then they fix for a lot of microsoft products but of course I'm speaking from my past experiences.

    Given the scenario, I'd probably virtualize the server using vmware converter then rebuild the new server from scratch, afterward migrate exchange, AD and other stuff to the new build and deploy it with as little as a few minutes of down time and still have the old server build archived if it was ever needed after the new deployment.

    but that's just me

    Also note that SBS2008 is x64 only, if the server is currently running x32 architecture, this may complicate stuff a little bit more.
    Last edited by mgeorge; 02-02-2009 at 08:04 AM.
    Reply With Quote Quote  

  7. Senior Member
    Join Date
    Mar 2007
    Posts
    12,308
    #31
    Thanks for the response Matt

    Exchange 2003 to 2007 will always involve a clean installation because you will always be going from a 32-bit platform to a 64-bit platform (though there is an unsupported 32-bit version of 2007 for testing).

    It's tricky in this scenario because I only have the one machine to work with, which is why virtualization is required, as you suggested. I was just going to load up Hyper-V, but now that I know I have an ESX-compatible machine, I might go with ESXi.
    Reply With Quote Quote  

  8. Cisco Guru mgeorge's Avatar
    Join Date
    Jun 2006
    Location
    127.0.0.1
    Posts
    800

    Certifications
    A few...
    #32
    Performance wise, I can definitely tell you that ESXi is a better choice. The installation of ESXi hypervisor is around 2-300Mb

    I'm not all that familiar with Hyper-V, mainly because when companies think virtualization, they think VMWare. I can probably count on one hand the mount of companies I worked with that used Hyper-V.

    I know with VMWare you can trunk a physical server and create different virtual servers in different VLAN's from a single trunked port from a Cisco switch. This way you can create another server or even multiple servers to put in a DMZ VLAN used for web mail access or different services of your choice.

    I've seen some pretty crazy designs with vmware.
    Last edited by mgeorge; 02-03-2009 at 04:25 AM.
    Reply With Quote Quote  

  9. Senior Member
    Join Date
    Mar 2007
    Posts
    12,308
    #33
    Matt, you need to take a break from your CCIE studies. You just spilled Cisco all over my virtualization

    I know what you're saying though; VMware allows for some pretty complex scenarios. And as far as ESXi goes, the footprint is only 32mb
    Reply With Quote Quote  

  10. Cisco Guru mgeorge's Avatar
    Join Date
    Jun 2006
    Location
    127.0.0.1
    Posts
    800

    Certifications
    A few...
    #34
    Yeah >.< Sry bout that, I'll go grab the Cisco mop and try to clean up.

    Yeah the 32MB hypervisor footprint is nice but I was referring to the complete installation size when its booted, which includes the v3i client installation files, tmp directory etc.. (of course what is 200-300MB anymore?) Most thumb drives are larger then that now.

    Man what happened to the days of 1.44MB 3.5's
    Reply With Quote Quote  

  11. Senior Member
    Join Date
    Jan 2008
    Posts
    1,941

    Certifications
    MCITP: EA, EMA; MCSE 2000/2003: M; MCSE 2000: S; MCSA 2000/2003: S; MCTS: ISA 2006; VCP3/4
    #35
    Quote Originally Posted by dynamik View Post
    Matt, you need to take a break from your CCIE studies. You just spilled Cisco all over my virtualization

    I know what you're saying though; VMware allows for some pretty complex scenarios. And as far as ESXi goes, the footprint is only 32mb
    It's not the 32M footprint that results in the better performance. A lot of it is the single instance memory storage that HyperV can't do.
    Reply With Quote Quote  

  12. Occasional Member dave0212's Avatar
    Join Date
    Nov 2007
    Location
    UK
    Posts
    283

    Certifications
    A few..... here are the highlights - CISSP, SSCP, CISA, CSTA, CSTP, CWSA, C|EH. and a bucket load of vendor specific certifications
    #36
    Quote Originally Posted by royal View Post
    Raid 1 = Bad Read/Good Write.
    Is this not the other way round?

    Twice the Read transaction rate of single disks, same Write transaction rate as single disks
    Reply With Quote Quote  

  13. One Man Wolfpac NetAdmin2436's Avatar
    Join Date
    Mar 2008
    Location
    Minnesota
    Posts
    1,077

    Certifications
    AAS in Computer Networking, MCSE 2003, Network+, Security+, A+
    #37
    Quote Originally Posted by dave0212 View Post
    Is this not the other way round?

    Twice the Read transaction rate of single disks, same Write transaction rate as single disks
    I think it depends what it's being compared to (ie, other raid levels or a single disk)
    RAID Level 1
    Reply With Quote Quote  

  14. Senior Member
    Join Date
    Jan 2008
    Posts
    1,941

    Certifications
    MCITP: EA, EMA; MCSE 2000/2003: M; MCSE 2000: S; MCSA 2000/2003: S; MCTS: ISA 2006; VCP3/4
    #38
    Quote Originally Posted by dave0212 View Post
    Is this not the other way round?

    Twice the Read transaction rate of single disks, same Write transaction rate as single disks
    Depends on the RAID controller. Some RAID controllers don't offer faster reading on a RAID1 mirror. In virtually all cases, any RAID offering striping has better read rates than RAID1, so it generally is safe to say RAID1 compared to RAID5 or 10 has a poorer read performance. When you add the fact that a RAID5 volume has a minimum of three disks, and RAID10 minimum of 4, he's spot on.

    However, RAID1 does offer good write performance compared to say RAID5 since it doesn't have to write the data plus a parity stripe of the same data. RAID1 is typically therefore faster than RAID5 for heavy writing I/O.

    In short, Royal is usually correct in his statement.
    Last edited by HeroPsycho; 02-03-2009 at 03:20 PM.
    Reply With Quote Quote  

+ Reply to Thread
Page 2 of 2 First 12

Social Networking & Bookmarks