Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Excessive writes (and an apology to Time4VPS/Delimiter)
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Excessive writes (and an apology to Time4VPS/Delimiter)

I'm guessing that doing nearly 20MB/s of write i/o on a VPS for what looks like several hours would be considered abuse?

In which case I'd like to apologise to Time4VPS and Delimiter for that not quite sure what's going on there but I'm pretty sure that's NOT supposed to happen.

Given I only sent what should have amounted to about 2GB of data quite why I was writing so much to disk I'm not sure but it seemed to be triggered by the uploading/transferring proccess.

Will await feedback from the application devs on that one as it doesn't seem right almost as if it was writing the data to disk/throwing it away and then writing it again.

Comments

  • Lol. No need to apology

  • Depends on the block size I guess.

    Thanked by 1ricardo
  • ricardoricardo Member
    edited February 2016

    The write performance on a well written
    program is much kinder to the disk than say, PHP without buffering. So it may not be as bad as you think.

  • dragon2611 said: Will await feedback from the application devs on that one as it doesn't seem right almost as if it was writing the data to disk/throwing it away and then writing it again.

    What were you running?

  • @dragon2611 said:
    I'm guessing that doing nearly 20MB/s of write i/o on a VPS for what looks like several hours would be considered abuse?

    In which case I'd like to apologise to Time4VPS and Delimiter for that not quite sure what's going on there but I'm pretty sure that's NOT supposed to happen.

    Given I only sent what should have amounted to about 2GB of data quite why I was writing so much to disk I'm not sure but it seemed to be triggered by the uploading/transferring proccess.

    Will await feedback from the application devs on that one as it doesn't seem right almost as if it was writing the data to disk/throwing it away and then writing it again.

    That's why I ruled VPS'es out.
    Anything is a no-no...
    Lol!

    :D

    Thanked by 1inthecloudblog
  • raindog308raindog308 Administrator, Veteran

    dragon2611 said: Will await feedback from the application devs on that one as it doesn't seem right almost as if it was writing the data to disk/throwing it away and then writing it again.

    image

    Thanked by 2jar elflord
  • Why choose a vps for such use when you can have a server for $10 from wsi

    Thanked by 1BeardyUnixGuy
  • @inthecloudblog said:
    Why choose a vps for such use when you can have a server for $10 from wsi

    Multiple VPS, actually ;-)

    It's a distributed filesystem but I'm pretty sure it wasn't working properly.

  • Is this using Infinit?

    If so, make sure you have equally/near equally performant filesystems otherwise things start going a little crazy, block copies start timing out.

    I have this running on ObjSpace and a couple of VMs

    Thanked by 1vimalware
  • @dragon2611 said:
    It's a distributed filesystem but I'm pretty sure it wasn't working properly.

    Pretty sure the @time4vps nas arrays are not the best place to run in testing mode.

    Have you checked out ovh's ssd vps? Low latency between their 3 locations too.

    Thanked by 1time4vps
  • time4vpstime4vps Member, Host Rep
    edited February 2016

    vimalware said: Pretty sure the @time4vps nas arrays are not the best place to run in testing mode.

    Our NAS arrays designed to store and retrieve backups. And it's recommended to do only this. Testing and producing server load intentionally - not the best idea.

    Synthetic tests are not the answer if you wondering "does this storage server will be enough for me". Just host that damn content and see your-self how service behave.

    Sometimes, it's funny when people host cold backups in SSD array... Resource and money waste.

    Thanked by 1wolfman
  • no need to apologize, you are being too kind! paid for resources? you have all the right to use them!

  • I pay road tax; doesn't give me the right to stop my car in the middle of the road.

    Use the best tool for the job like @time4vps said. The NAS is great for cold data.

  • time4vps said: Sometimes, it's funny when people host cold backups in SSD array...

    Literally funniest thing ever.

  • shovenoseshovenose Member, Host Rep

    @Maniac said:
    Literally funniest thing ever.

    Hahahahaha yeah so hilarious

    Thanked by 1deadbeef
  • @MarkTurner said:
    Is this using Infinit?

    If so, make sure you have equally/near equally performant filesystems otherwise things start going a little crazy, block copies start timing out.

    I have this running on ObjSpace and a couple of VMs

    Hmm I wonder if that's the problem because my nodes are so far apart.

  • dragon2611 said: Hmm I wonder if that's the problem because my nodes are so far apart.

    Latency is a small part of the issue, but the real issue is IO latency.

    I have servers running US/Asia/EU and my biggest issue was IO latency. It seems to cause Infinit to timeout the block placement and then try again, and so on until you are just disk thrashing.

    Thanked by 1vimalware
  • Infinit uses UDT for moving packets. UDT has some issues in dealing with packet loss and latencies, so maybe the retransmits are killing your disk.

Sign In or Register to comment.