313 reputation
115
bio website
location
age
visits member for 3 years
seen Jan 26 at 16:03

May
16
comment Water-Cooled Rack Door: does it work well with Blades or GPU nodes?
it's a designated room of about 80 m^2 with no windows and a single AC (pretty weak, perhaps ~20 kW or so to keep a constant room temperature). there's a water supply that has enough capacity to provide cooling for ~100 kW. there are no strong space constraints.
Nov
26
comment ZFS performance: do I need to keep free space in a pool or a file system?
syneticon-dj, thank you for this explanation! Increasing RAM seems to help indeed.
Jul
24
comment ZFS: Why does resilvering go over all disks in a pool rather than only those in the raidz1 vdev?
Solaris 10 8/11. The resilvering is over by now, but I can tell you it was a disk inside raidz1-1. Thank you for the pointer to iostat, I will look into it next time I lose a disk. I'll also have to stop nfsd first in order to isolate the resilvering load from the usual load.
Jul
24
comment CUDA: Is it possible to dynamically restrict the number of cores / threads / clock freq. while a process is running on a GPU?
Thank you for your answer! You're right with everything you said, but I'm still looking for some technical details of how to perform the throttling. nvclock was a good start, but it seems that it doesn't support my video cards. Perhaps there are other ways?
Jul
23
comment CUDA: Is it possible to dynamically restrict the number of cores / threads / clock freq. while a process is running on a GPU?
Of course you're right, working cooling is absolutely important. Thank you for the pointer to nvclock!
Jul
23
comment CUDA: Is it possible to dynamically restrict the number of cores / threads / clock freq. while a process is running on a GPU?
Yes, this was the first thing I thought of, but unfortunately I can not modify the code to reduce the load. That's why I'm curious if it's possible to do this on the Linux kernel level (tell scheduler to give the process less shares?) or the CUDA driver (same idea).
Jul
23
comment CUDA: Is it possible to dynamically restrict the number of cores / threads / clock freq. while a process is running on a GPU?
I'd appreciate a work-around though. Proper cooling is a different topic, but for a moment I'm looking for a way to control the GPU resources as I described.
Jul
2
comment Pass kernel boot parameters from dhcpd.conf for a host group?
OK, I solved it with PXE now. Thank you, dawud!
May
27
comment ZFS performance: do I need to keep free space in a pool or a file system?
OK, thanks a lot for your help!
May
27
comment ZFS performance: do I need to keep free space in a pool or a file system?
Hehe.. Yes, this is a quite common situation we face. So, are claims like this: "On filesystems with many file creations and deletions, utilization should be kept under 80% to protect performance." unprecise, because it's really about the free space within a pool rather than file system?
May
27
comment ZFS performance: do I need to keep free space in a pool or a file system?
Yeah, but it's a pain to have these "please clean up your sh.., the fileserver is awfully slow" discussions in the mailing lists every couple of weeks...
May
27
comment ZFS performance: do I need to keep free space in a pool or a file system?
Thanks! There's no fair way in our setting to use quotas, so everyone uses the same mount point and can fill up the space, leading to a drop in performance. My idea was to guarantee some free space with a reservation so that the overall system never gets too slow. But IIUC, I can have this guarantee by limiting volume to 8.5T and never think about it again. Is that correct?
Jan
23
comment Can I safely use a 1U server with an actively cooled Nvidia GTX 680 video card?
Well there's an active ("K20C") and a passive version of the K20 video card. The passive one is designed for server units like in those specs, so I guess this is also what you've used?
Jan
23
comment Can I safely use a 1U server with an actively cooled Nvidia GTX 680 video card?
We did some comparisons with a K20c and the difference in performance (with our application) was only about 10-20%, while the K20c is about 4-5x the price of a GTX680.
Jan
23
comment Can I safely use a 1U server with an actively cooled Nvidia GTX 680 video card?
It's for CUDA based scientific computing, so I expect the video cards to get pretty hot most of the time. I will look for a 2U server that fits our needs. Thanks for your answer!
Jan
23
comment Can I safely use a 1U server with an actively cooled Nvidia GTX 680 video card?
What kind of chassis are you using - 2U? Or more?
Jan
23
comment Can I safely use a 1U server with an actively cooled Nvidia GTX 680 video card?
1U or 2U doesn't matter that much to me. I'm more curious if I can use cheap GTX video cards that are actively cooled while the fan is pretty close to the chassis top. Does it work well for you, or have you used passively cooled GPUs?
Jan
11
comment Howto set up SGE for CUDA devices?
I still have one more question on this configuration: now I have multiple queues, one of which offers "gpu=2" and the others offer "gpu=0". My problem is, that since "0" is the default value and "0<=2", all users who don't request any GPUs are blocking the GPU-capable queue most of the time with their jobs. How can I tell SGE that this queue is to be used for jobs requesting "gpu>0" only?
Nov
22
comment How to find maximal value for “zfs set quota” without changing quota?
Thanks, that was what I was looking for!
Nov
22
comment How to find maximal value for “zfs set quota” without changing quota?
Thanks! In my case, there is no parent FS, the file system I'm using is the top most one. "zfs get available" thus only returns the remaining free space (1TB actually; the storage should be ~40T). How do I find out the available storage space then?