Forum Discussion

felonious_caper's avatar
13 years ago

Staggered Scheduling

I have read that it is a good practice to "stagger" the scheduling of jobs so that netbackup can allocate and manage memory and resources more effeciently.

Currently I have all of my policies set to start at the same time, they then backup in order of priority and only one policy can run one job at a time.

 

I now have a TON of jobs that are queued. Do these queued jobs eat up resources and memory like an active job does? Should I actually set the open windows to different times so there are fewer queued jobs at any given time?

6 Replies

  • It sounds like you have really tied your system down if you have one job running per policy so to be honest i wouldnt worry about it

    They wont eat much in teh way of resources NBU will just check it all occasionally but wont be dong much else with it

    One of my managed customers runs about 9000 jobs every night, a lot of which kick off at similar times and run to a host of Media servers each of which has lots of advanced disks all in one big group

    We dont see any real issues so unless you have thousands queued and only a handful running then I dont think it will cause issues

    My concern based on what you have sadi would be to actually get a lot more jobs running to get everything under way - as long as you have the media resources to cope - that is where disk really comes in handy

    Hope this helps

  •  Unfortunately, we only have two drives (one in each robot), and one disk to backup to. At one point we had 15 jobs all writing to disk at the same time, but our performance stats came to a screeching halt. This is because even though 15 clients are being read from at a given time, the disk can only write one at a time. (Hence setting all of the policies (and storage units) to one job at a time.

  • "the disk can only write one at a time" - is this just due to it performance?

    I like to hit disk hard (at least 25 to 50 jobs at a time) - that is the idea behind it - good disk is really cheap now and well worth the investment

    It usually negates a tape library expansion or upgrade so pays for itself.

  •  Well after my NBU training class I'm under the impression that unless a disk is part of a disk array, you can only write one at a time (in parrallel) for as many physical disk that you have. Otherwise you the I/O performance drops drastically.

  • If you just have a single physical disk drive then I guess you are about right but I would tend to use external disk in a RAID

    So a RAID 5 set I would happily hit with 25 jobs

    SIZE and NUMBER disk buffers making a great difference

    When you can get so much disk these days in such small units then bigger is better - but it all down to budget

  • Until such time as you can replace your current disk STU, just bypass the disk and backup directly to tape.
    You should be able to safely write 4 streams per tape drive.