Questions

Slow backups for Oracle 9i

+
0 Votes
Locked

Slow backups for Oracle 9i

Hector Ramos
This one is for my wife. They're trying to do offline backups of an Oracle 9i database. This database is the backend for a Centricity EMR system (5.6). They're experiencing long backup times (2-4 minutes per Gigabyte), on a 110 Gigabyte Database. I have tested 2.8 Gig files on my network and it takes about 3 minutes to copy them. The server has been replaced, and the database reinstalled. We both think that this is a network issue; but, can this be an issue internal to this box? She has tried both xcopy and robocopy with the same results.
  • +
    0 Votes
    OH Smeg

    But it all depends on the hardware inside the Server and what if any bottle necks are built into it.

    Depending on what is being done it's possible to have a bottle neck at the HDD Controller or the LAN Connection as the most obvious places but then again it also depends on the types of CPU/s fitted and the amount of RAM as well so there are any number of reasons why it could be slow from the Server itself or if the place is using some old 10 MBS Switches or Hubs this will slow down data transfer considerably as well.

    Col

    +
    0 Votes
    Hector Ramos

    As I hear it, here are the specs ...

    2 Gigs RAM
    Dual Dual Core (sorry, don't know speed) CPUs
    Gig link

    The really screwy part of this is that it's no a constant failure. Sometimes it flies (30-45 seconds per Gigabyte) sometimes it crawls. I am convinced it's a logical network issue. I did say that it didn't fail al the time; but, it does fail in a consistent basis. I mean that on Saturday night into Sunday AM it's horrible. The nit speeds up duting the afternoon. In the evenings it crawls. The main thing I hope to g et form this thread is some other things to look at OTHER than the network. I'm asking my wife to push hard on her NetOps team to look for a root cause in their network; but, I also want to explore all avenues, OS or Oracle, for example.

    +
    0 Votes
    OH Smeg

    That is that there is something else scheduled to run at the same time and that's taking up a lot of Network Bandwidth and causing the Backup to be slow.

    If the Backup isn't always slow that is what I would first be looking at check the network traffic and then start looking at what is using what so you can begin the fault finding exercise.

    Col

    +
    0 Votes
    Hector Ramos

    Now to get someone else's IT dept to listen to me :-)

    Thanks for the help. Any other input is greatly appreciated.

    +
    0 Votes
    sgt_shultz

    media type on nic? or a bad nic? or needs sp or newest drivers...
    for sure if this is a new server i would check the drivers to be sure latest greatest installed. also i'd bring server up to latest sp's for windows (is this windows??) and oracle...
    i suppose it is far fetched but how big are these files? could you have a black hole router? maybe others can tell us if that would fit your symptoms. I've only read about it...
    maybe you have a bad nic somewhere blatting out traffic. or maybe you have a media autosensing nic that can't quite make up its mind.
    perhaps have a look in the event logs on the backup box. is this windows or linux?

    +
    0 Votes
    Hector Ramos

    Unfortunately, I don't know their topology. The files are 2 Gigabytes in size. Robocopy seems able to handle these fine, as long as this phantom problem doesn't appear. By fine I mean 30 seconds per Gigabyte. The autosensing NIC is a good suggestion that has already been forwarded. I don't know if they did anything about it.

    It's a W2k box running Oracle 9i. I know my wife has the Oracle server with the latest patches approved by GE. They (GE) is very picky about which updates they endorse. This does mean that the system is running on older patch levels than Oracle would suggest.

    In case y'all missed it, the GE tie is due to the fact that this Oracle DB is the back end to a Centricity (Logician) EMR system. Centricity is owned by GE Healthcare.

    +
    0 Votes
    techrepublic

    First thing to do is make sure the database actually stops before the backup. I used Backup Exec 11D, 1GB network cards on the server and switch, with 720GB/Sec Back plane (<1MS Latency). It takes 1.5 - 2 hours to balcup 120GB database to a 1TB HP virtual tape library.

  • +
    0 Votes
    OH Smeg

    But it all depends on the hardware inside the Server and what if any bottle necks are built into it.

    Depending on what is being done it's possible to have a bottle neck at the HDD Controller or the LAN Connection as the most obvious places but then again it also depends on the types of CPU/s fitted and the amount of RAM as well so there are any number of reasons why it could be slow from the Server itself or if the place is using some old 10 MBS Switches or Hubs this will slow down data transfer considerably as well.

    Col

    +
    0 Votes
    Hector Ramos

    As I hear it, here are the specs ...

    2 Gigs RAM
    Dual Dual Core (sorry, don't know speed) CPUs
    Gig link

    The really screwy part of this is that it's no a constant failure. Sometimes it flies (30-45 seconds per Gigabyte) sometimes it crawls. I am convinced it's a logical network issue. I did say that it didn't fail al the time; but, it does fail in a consistent basis. I mean that on Saturday night into Sunday AM it's horrible. The nit speeds up duting the afternoon. In the evenings it crawls. The main thing I hope to g et form this thread is some other things to look at OTHER than the network. I'm asking my wife to push hard on her NetOps team to look for a root cause in their network; but, I also want to explore all avenues, OS or Oracle, for example.

    +
    0 Votes
    OH Smeg

    That is that there is something else scheduled to run at the same time and that's taking up a lot of Network Bandwidth and causing the Backup to be slow.

    If the Backup isn't always slow that is what I would first be looking at check the network traffic and then start looking at what is using what so you can begin the fault finding exercise.

    Col

    +
    0 Votes
    Hector Ramos

    Now to get someone else's IT dept to listen to me :-)

    Thanks for the help. Any other input is greatly appreciated.

    +
    0 Votes
    sgt_shultz

    media type on nic? or a bad nic? or needs sp or newest drivers...
    for sure if this is a new server i would check the drivers to be sure latest greatest installed. also i'd bring server up to latest sp's for windows (is this windows??) and oracle...
    i suppose it is far fetched but how big are these files? could you have a black hole router? maybe others can tell us if that would fit your symptoms. I've only read about it...
    maybe you have a bad nic somewhere blatting out traffic. or maybe you have a media autosensing nic that can't quite make up its mind.
    perhaps have a look in the event logs on the backup box. is this windows or linux?

    +
    0 Votes
    Hector Ramos

    Unfortunately, I don't know their topology. The files are 2 Gigabytes in size. Robocopy seems able to handle these fine, as long as this phantom problem doesn't appear. By fine I mean 30 seconds per Gigabyte. The autosensing NIC is a good suggestion that has already been forwarded. I don't know if they did anything about it.

    It's a W2k box running Oracle 9i. I know my wife has the Oracle server with the latest patches approved by GE. They (GE) is very picky about which updates they endorse. This does mean that the system is running on older patch levels than Oracle would suggest.

    In case y'all missed it, the GE tie is due to the fact that this Oracle DB is the back end to a Centricity (Logician) EMR system. Centricity is owned by GE Healthcare.

    +
    0 Votes
    techrepublic

    First thing to do is make sure the database actually stops before the backup. I used Backup Exec 11D, 1GB network cards on the server and switch, with 720GB/Sec Back plane (<1MS Latency). It takes 1.5 - 2 hours to balcup 120GB database to a 1TB HP virtual tape library.