General discussion

  • Creator
    Topic
  • #2287328

    Examine common RAID levels

    Locked

    by debate ·

    Has your organization implemented RAID arrays? If so, which RAID level do you use? Have you found them effective? Share your comments about using RAID arrays to protect data, as discussed in the March 2 Disaster Recovery e-newsletter.

    If you haven’t subscribed to our free Disaster Recovery e-newsletter, sign up today!
    http://nl.com.com/acct_mgmt.jsp?brand=techrepublic

All Comments

  • Author
    Replies
    • #3381656

      RAID for small business

      by david_dundas ·

      In reply to Examine common RAID levels

      We are a small buiness that has mantained a single server for almost 20 years (not the same one). I found this article to be useful but it missed a few critical points:

      The first is that a particular RAID controller is usually specific to the drives it controls, ie the striping of the drives is the “fingerprint” of that controller, so if it fails (we have had this happen twice over the years)a replacement may not be able to read the hard drives, then the tape backup is your only lifeline. The RAID controller is the weak link in the RAID scenario. This is a problem that the IT industry ought to tackle and solve.

      The second point is that hot swappable arrays are available that allow a drive to be replaced without downing the system, but you need drive surveillance software such as IBMs ServeRAID manager to alert management if a drive fails.

      • #3381598

        Problem Solved

        by draco vulgaris ·

        In reply to RAID for small business

        Digital Equipment Corporation solved this problem years ago with dual redundant controllers. The SCSI controllers were the HSZ40, HSZ50, HSZ70, and HSZ80.

        The HSZ50 was a wide SCSI controller that supported narrow SCSI disk. The HSZ70 was Ultra SCSI. Both are available used. I think that HSZ80 is still available from HP. You would need to buy one of the RAID Arrays supported by these controllers. You get (optional) dual redundant controllers, N+1 or full redundant power supplies, battery backup for the cache memory, redundant paths to the host (if the hardware/software supports such), etc.

        This stuff is not cheap! If you need this sort of reliability you pay willingly, if not gladly. I’ve been running a DEC RAID Array 450 for five years with only one hardware failure (a power supply) and that did not cause any down time.

    • #3381653

      RAID 10 with 3 disks?

      by ethan ·

      In reply to Examine common RAID levels

      Ummm… RAID 10 uses 4 disks, and the difference between RAID 10 and RAID 0+1 is the order that you stripe and mirror. RAID 0+1 is striped, then mirrored, so if one drive from each mirror dies, you lose . This means 4/6 configurations of 2 drives failing kills you. RAID 10 is mirrored, then striped, so it is better for redundancy since you only die if 2 mirrored drives fail. This means only 2/6 configurations of 2 drives failing kills you.

    • #3381592

      RAID 1 performance

      by draco vulgaris ·

      In reply to Examine common RAID levels

      RAID 1 (mirroring) does carry a write penalty.

      There is, however, no penalty on reads. Multichannel controllers can read different records from multiple members of a RAID 1 set simultaneously.

      The better hardware RAID controllers can select, from the members not otherwise occupied, the member most advantageously positioned to retrieve the required record; e.g. if the requested record is on cylinder 200, the member with its heads at cylinder 195 will be chosen in preference to the member with heads at cylinder 0.

    • #3381591

      In answer to your question

      by draco vulgaris ·

      In reply to Examine common RAID levels

      Mike,

      I use a pair of mirrored 80GB on my PC at home (Adaptec IDE controller).

      At work we use both RAID 1 (SCSI) in PC servers and RAID 5 in our mid-range systems.

      We make regular tape backups; RAID defends against hardware failure but not against software error or human error; both of which are more common than hardware failure!

      • #2729147

        Agreed

        by miketalonnyc ·

        In reply to In answer to your question

        RAID is designed for hardware errors. Tape and other solutions are designed for software errors.

        The focus of the article was specific, which is unusual for this column, but becuse of the tight focus I was limiting what solutions I spoke of.

        Thanks.

        Mike Talon
        miketalonnyc@yahoo.com

    • #3381483

      Raid – the second data saver

      by johnnyhb ·

      In reply to Examine common RAID levels

      Yes we do use raid on all our 5 servers. The unix, Novell and Win2k (master and replica) servers all uses level 5 with hotswap scsi disks. One minor and less critical application server uses level 1 and IDE disks.

      The raid is only a insurance for not crashing a server if a disk crashes.

      The main Disaster recovery is the daily backup to tape. The backup is often needed due to “user failure”: overwriting or deleting. From time to time it is also needed due to damaged exel and word documents.

      The raid has only been “usefull” once since we introduced it in our servers in january 1996, but the benefit was great. When located “remote” from available spare parts, it keept our system running with a dead harddrive until a replacement was available.

      The real crashes has been mainboard failures and scsi controller failures (that connects to the tape streamer) and then the raid system had nothing to add to system uptime.

      I am aware of that others has not been that lucky with harddisks that we have, and had wery good use of raid configuration to keep their server running when a harddisk crash strikes.

    • #3383288

      Raid Configurations

      by gordon.briffa ·

      In reply to Examine common RAID levels

      At work we have over 13 Win2K Servers using Raid 5 and 2 HPUX machines using Raid 10. Altough using Raid configuration (especially when using SCSI HDD) it is still very useful in case of HDD failure. It is also very important to have a good raid controller since this will be useful when rebuilding the new disk (ie when a HDD fails and is replaced by another HDD this new disk needs to form part of the RAID). I strongly suggest this Raid configurations especially if your server rooms lack any good UPSs and there is a tendency of flactuation in electricity cause this can easilly damage your disks.

    • #3383233

      RAID10 is great – RAID5 – er no.

      by richard kirk ·

      In reply to Examine common RAID levels

      RAID5 is great for reading, but has real drawbacks when it comes to writing data as it has to calulate and then write parity data.
      Points:
      1) Disks are the slowest part of a system
      2) Disks are also very cheap – about ?300. SCSI disk
      3) Decent servers like Dells come with RAID built-in
      4) RAID 10 is superbly fast and only costs ?907 ( 3 additional disks + ?7 to split SCSI backplane on a Dell PowerEdge)
      5) Your data is priceless!

      Cheers

      Richard

Viewing 6 reply threads