wirenut Posted July 25, 2013 Share Posted July 25, 2013 Trying to search on those variables but having a difficult time. Any help to point us in the right direction? +1 Link to comment
jumperalex Posted July 26, 2013 Share Posted July 26, 2013 "A good plan implemented today is better than a perfect plan implemented tomorrow." — George Patton Warfare and redundant servers ... not exactly the same mission set. Link to comment
drawz Posted July 26, 2013 Share Posted July 26, 2013 "A good plan implemented today is better than a perfect plan implemented tomorrow." — George Patton Warfare and redundant servers ... not exactly the same mission set. No lives are at stake with unRAID. All the more reason to accept good enough for now with plans to improve in the future. I would argue this principle is valid in many many aspects of life. Link to comment
jumperalex Posted July 26, 2013 Share Posted July 26, 2013 "A good plan implemented today is better than a perfect plan implemented tomorrow." — George Patton Warfare and redundant servers ... not exactly the same mission set. No lives are at stake with unRAID. All the more reason to accept good enough for now with plans to improve in the future. I would argue this principle is valid in many many aspects of life. The ramifications of inaction are still not the same unless there is a current bug putting us all at risk Link to comment
drawz Posted July 26, 2013 Share Posted July 26, 2013 "A good plan implemented today is better than a perfect plan implemented tomorrow." — George Patton Warfare and redundant servers ... not exactly the same mission set. No lives are at stake with unRAID. All the more reason to accept good enough for now with plans to improve in the future. I would argue this principle is valid in many many aspects of life. The ramifications of inaction are still not the same unless there is a current bug putting us all at risk The ramifications of SMB running a little slow for a few people are pretty minimal. Sometimes you just have to accept that it will never be perfect and move on. There is always one more bug or just one more feature. Perhaps this proverb is more appropriate: Perfect is the enemy of good. http://en.wikipedia.org/wiki/Perfect_is_the_enemy_of_good Link to comment
dmacias Posted July 26, 2013 Share Posted July 26, 2013 Trying to search on those variables but having a difficult time. Any help to point us in the right direction? I think he means the tunables in the gui under settings/disk settings. Tunable (md_num_stripes): Tunable (md_write_limit): Tunable (md_sync_window): Link to comment
jaybee Posted July 26, 2013 Share Posted July 26, 2013 The SMB slowness is a major issue in my opinion. NFS has been buggy before and SMB has always been reliable and fast as a fall back option. Now with SMB performance halving, that's unacceptable. I also want a 5 final, but it will never happen if we keep changing the kernel and moving goal posts. Link to comment
drawz Posted July 26, 2013 Share Posted July 26, 2013 The SMB slowness is a major issue in my opinion. NFS has been buggy before and SMB has always been reliable and fast as a fall back option. Now with SMB performance halving, that's unacceptable. I also want a 5 final, but it will never happen if we keep changing the kernel and moving goal posts. I actually don't even see this listed in the "issues list" section of the forum. Has Tom acknowledged it? What work has been done to investigate it? Does it happen to everyone? Is it select hardware? When these issues are worked on publicly, with Tom and the forum members who are effected, it seems that solutions are found much more quickly. Link to comment
bmfrosty Posted July 26, 2013 Share Posted July 26, 2013 The SMB slowness is a major issue in my opinion. NFS has been buggy before and SMB has always been reliable and fast as a fall back option. Now with SMB performance halving, that's unacceptable. I also want a 5 final, but it will never happen if we keep changing the kernel and moving goal posts. I see a lot of SMB slowness as well. I don't know if it's a blocker, but for directories with more than 100 entires, it's pretty darn slow - taking several seconds to more than a minute to provide a listing. By comparison NFS is very fast. Link to comment
nars Posted July 26, 2013 Share Posted July 26, 2013 The SMB slowness is a major issue in my opinion. NFS has been buggy before and SMB has always been reliable and fast as a fall back option. Now with SMB performance halving, that's unacceptable. I also want a 5 final, but it will never happen if we keep changing the kernel and moving goal posts. I see a lot of SMB slowness as well. I don't know if it's a blocker, but for directories with more than 100 entires, it's pretty darn slow - taking several seconds to more than a minute to provide a listing. By comparison NFS is very fast. The only slowdown I can reproduce to list files is for folders with many and large files that haven't been accessed since system booted (i.e. fs not cached) but it is not SMB related at all, just fs thing... if I do an 'ls' on them right after system boot I can see same (and yes I know that there is some script/plugin that lists all files to avoid this, but I just don't feel the need for it until now). Other than that I can't really see any slowdown to list files over SMB even for folders with 1000 files I can open them "in a blink" over smb. Not something at client side (realtime antivirus or something similar)? Link to comment
peter_sm Posted July 27, 2013 Share Posted July 27, 2013 I was doing a quick test of samba performance from my W7 laptop. Used a 1GB large file. all disk was spun up during this test and 1GB network on both server and laptop. Copy file from W7 to disk share = about 20 to 25MB/s Copy file from W7 to cache = about 35MB/s W7: Copy file from cache or disk share to W7 = 35MB/s Sometimes the copy start very slow, from 1MB and go slowly up... Previous this week I saw this transfer speed was much lower. And I think I was up to 50 - 60 MB transfer speed from the server to W7 (laptop) on older RC releases. EDIT On my "Hackintosh" I copy a file using AFP from user share to my mac on about 100MB/s!!!! never seen this on my W7?!! see attcahed images using samba I'm on same numbers that I haver on my W7, doing more test ........... Using samba and copy a file from user share to my MAC it's going up and down 1MB/s to 32MB/s, see second attached image... OSX (afp) to cache = 70MB/s OSX (afp) to Disk share = peak 38MB/s //Peter Link to comment
Harpz Posted July 27, 2013 Share Posted July 27, 2013 What are you using to track the copy in the screenshots? Link to comment
theone Posted July 27, 2013 Share Posted July 27, 2013 Does iperf test the network performance over SMB? http://lime-technology.com/forum/index.php?topic=10771.0 Link to comment
peter_sm Posted July 27, 2013 Share Posted July 27, 2013 What are you using to track the copy in the screenshots? Activity Monitor -> Network //P Link to comment
Thornwood Posted July 27, 2013 Share Posted July 27, 2013 I remember in past RC's i just cant find we were installing samba pakage in the go script to over ride the one in unraid native becouse if some bug you could not deleat files. I cant remember how this is done? But in http://www.samba.org/ they have a version 4.? How would you feel if we try some diffrent versions to nail down a fix for Tom? If this seems like a good idea i can do the reasearch on how to do this. Thornwood Sent from my YP-G70 using Tapatalk 2 Link to comment
SnickySnacks Posted July 27, 2013 Share Posted July 27, 2013 Anecdotal, but my samba copy speed from my server with 16c is a solid 100MB/s via SMB from a user share to Win7 via gig-e. Only things installed are apcupsd and powerdown script. Link to comment
peter_sm Posted July 27, 2013 Share Posted July 27, 2013 I remember in past RC's i just cant find we were installing samba pakage in the go script to over ride the one in unraid native becouse if some bug you could not deleat files. I cant remember how this is done? But in http://www.samba.org/ they have a version 4.? How would you feel if we try some diffrent versions to nail down a fix for Tom? If this seems like a good idea i can do the reasearch on how to do this. Thornwood Sent from my YP-G70 using Tapatalk 2 I was doing that previsly, see my post here -> http://lime-technology.com/forum/index.php?topic=16840.msg174243#msg174243 Please feel free to try //Peter Link to comment
adrian ballard Posted July 27, 2013 Share Posted July 27, 2013 all I can say is: SMB is VERY important to many people and if there is a slow-down issue, it definitely needs to be addressed. Link to comment
drawz Posted July 27, 2013 Share Posted July 27, 2013 Here is the only thread I can find that suggests there is an SMB performance issue: http://lime-technology.com/forum/index.php?topic=28664.0 However, if you look in that thread, the latest posts indicate there probably is NOT an issue. It seems a little more research is needed, but so far Tom has not acknowledged an issue from what I can tell. Link to comment
Thornwood Posted July 28, 2013 Share Posted July 28, 2013 I am doing fine in speeds but the people that are having issues can try to change the SMB to another version. Sorry since I don't see a problem I don't think I can help. But thank you for how to create a package this has worked for me in the past. Link to comment
optiman Posted July 30, 2013 Share Posted July 30, 2013 The best I ever saw on my mombo is low 80's, and it was back on 4.7. Even then, it dropped at some point, not sure why or when. I was so new to unraid, I didn't really give it much attention. With v5, I've always seen 50's and low 60's re SMB speeds going from my Win7 box to the cache drive over gb. I also would like to know the best way to capture these speeds. I use Total Commander and it shows the speed during the copy. I guess my speeds are ok, for my setup. I'd like to hear if others with my same setup are getting better speeds with SMB. Link to comment
Barziya Posted July 30, 2013 Share Posted July 30, 2013 Noticed linux kernel 3.10 has been released, dare I upgrade? Since that was posted, 3.10 has gone to stable 3.10.4, and 3.9 has been labeled EOL. Now it seems reasonable to entertain this idea. Link to comment
prostuff1 Posted July 30, 2013 Share Posted July 30, 2013 EOL kernels don't get much new patches. Now it seems reasonable to at least entertain the idea of bumping up to a 3.10.4 kernel, while we are still in RC. no Link to comment
Barziya Posted July 30, 2013 Share Posted July 30, 2013 no no? Wouldn't you rather have a LTS kernel, than a EOL kernel? (It's not like we haven't waited a few years for a final, I'm sure you can survive if you'll have to wait an extra week or so.) Link to comment
jumperalex Posted July 30, 2013 Share Posted July 30, 2013 Correct no. 5.1 can have the new kernel. There are no blockers to final, even slow smb, since it isn't uselessly slow. Sent from my Nexus 7 using Tapatalk HD Link to comment
Recommended Posts