I have a share setup to use cache drive only and I use it as my auto organize folder for emby server. When I copy files to that share from my main PC I get write speeds of 220-240MB/s. If however I copy to a folder within a share setup to use the cache I only get write speeds of 75-80MB/s. Is the overhead from unraid causing this much of a drop in performance? Any thoughts on what I can change, if anything? Running 6.1.3

 

thanks

I have a share setup to use cache drive only and I use it as my auto organize folder for emby server. When I copy files to that share from my main PC I get write speeds of 220-240MB/s. If however I copy to a folder within a share setup to use the cache I only get write speeds of 75-80MB/s. Is the overhead from unraid causing this much of a drop in performance? Any thoughts on what I can change, if anything? Running 6.1.3

 

thanks

Are you sure about those numbers? Do you have 10G ethernet or something?

yes

yes

Yes what? There were 2 questions.

yes, running 10Gig

I can't speak as to your issue, but what I can say is it shouldn't be unRaid overhead causing that much of a slowdown. I can write to a cache enabled share on my server at 112MB/s, basically maxing out GigE.

Are you sure the share is actually using cache?  Try making a small file in the folder and see if it appears on the cache drive

Yes, the share is using the cache. That was my first thought that the cache was not being used, but I checked cache drive when I was copying files to the share and they showed up. I tried different file sizes and copied to different folders and speed stays pretty consistent at 75-80 MB/s, small files are a little slower but not much. Stumped as to the large difference in speeds. Anyone else running 10G that can do some speed tests and compare?

 

thanks

Is the cache drive an older, slower drive?

I have a couple that only make 53 MB/s.

Is the cache drive an older, slower drive?

I have a couple that only make 53 MB/s.

 

It's most certainly an SSD as he stated he gets 220-240 MB/s when copying directly to the cache drive.

Is the slower folder being watched by EMBY?

Or anything else for that matter. If they are competing for resources that could account for the drop in speed.

Yes, the share is being watched by EMBY, it contains all my media, however the cache only share is my auto organize folder so its being watched as well. I will try shutting the EMBY server down and retest the file transfers, more data when I get home from work.

 

thanks

It's most certainly an SSD as he stated he gets 220-240 MB/s when copying directly to the cache drive.

 

I am not copying directly to the cache drive. The speed differences are between 2 shares, one is cache only and the other is cached. That's why  I thought they should be similar performance. It seems I'm out of the comfort zone of many unraid users, I thought that stating I was getting 220-240MB/s would automatically answer the questions yes I'm using at least 10G Ethernet and its an SSD drive.

... It seems I'm out of the comfort zone of many unraid users, I thought that stating I was getting 220-240MB/s would automatically answer the questions yes I'm using at least 10G Ethernet and its an SSD drive.

Since you're not a frequent poster and you hadn't given us information about your hardware, I think we were just looking for more evidence that your numbers were believable. Not everyone who posts can be taken at their word regarding their diagnosis of their systems.

I am not copying directly to the cache drive. The speed differences are between 2 shares, one is cache only and the other is cached. That's why  I thought they should be similar performance. It seems I'm out of the comfort zone of many unraid users, I thought that stating I was getting 220-240MB/s would automatically answer the questions yes I'm using at least 10G Ethernet and its an SSD drive.

 

So this is writing directly to the protected array. 80-90MBs is quite normal performance. This is due to the slower disks and the parity calculations. Hence the reason many people opt for a cache drive.

Since you're not a frequent poster and you hadn't given us information about your hardware, I think we were just looking for more evidence that your numbers were believable. Not everyone who posts can be taken at their word regarding their diagnosis of their systems.

 

And you'd be surprised how many people think MB/s and Mb/s are the same thing.  Obviously you know the difference.

 

My mistake on the "writing directly to the cache drive" statement.  Again, sorry I can't be of more help than to say that writing to a "cache enabled" share on my server I can saturate GigE. 

 

What are the specs of your system? 

So this is writing directly to the protected array. 80-90MBs is quite normal performance. This is due to the slower disks and the parity calculations. Hence the reason many people opt for a cache drive.

 

No, he said one is cache only and the other is "cached".  By that I assume he means "cache enabled", i.e. it is written to the cache drive then copied to the array each night when mover runs.

Since you're not a frequent poster and you hadn't given us information about your hardware, I think we were just looking for more evidence that your numbers were believable. Not everyone who posts can be taken at their word regarding their diagnosis of their systems.

 

And you'd be surprised how many people think MB/s and Mb/s are the same thing.  Obviously you know the difference.

 

My mistake on the "writing directly to the cache drive" statement.  Again, sorry I can't be of more help than to say that writing to a "cache enabled" share on my server I can saturate GigE. 

 

What are the specs of your system?

 

Supermicro SC846 case with Tyan S7025 with dual xeon 5650's and 32gig, 40+TB of mixed WD red and HGST deskstar NAS, using HGST 7200rpm 6TB for parity drive and 500Gig Samsung 850 EVO for cache. Drives are connected to LSI2008 controllers, Mellanox Connectx-2 10Gig network.

I am not copying directly to the cache drive. The speed differences are between 2 shares, one is cache only and the other is cached. That's why  I thought they should be similar performance. It seems I'm out of the comfort zone of many unraid users, I thought that stating I was getting 220-240MB/s would automatically answer the questions yes I'm using at least 10G Ethernet and its an SSD drive.

 

So this is writing directly to the protected array. 80-90MBs is quite normal performance. This is due to the slower disks and the parity calculations. Hence the reason many people opt for a cache drive.

According to OP and his post you quoted, it is not writing directly to the protected array.

Well problem solved at least for now. As suggested I shutdown all my dockers and VM's. Started everything back one at a time and checked transfer speeds and they are now close. After everything was loaded again I checked it with LAN Speed Test and it is showing only about 10% slower writing to the cached share and almost no difference reading between the 2 shares. Hopefully a one time glitch.

 

thanks

Well problem solved at least for now. As suggested I shutdown all my dockers and VM's. Started everything back one at a time and checked transfer speeds and they are now close. After everything was loaded again I checked it with LAN Speed Test and it is showing only about 10% slower writing to the cached share and almost no difference reading between the 2 shares. Hopefully a one time glitch.

 

thanks

It's possible your min free space wasn't large enough.  When writing to the cache enabled share, how did you validate the files written were actually on the cache as opposed to going to the array?  If the min free space setting for the cache was set to a number higher than the amount of data being written to the enabled share, this would cause the data to be written directly to the array.

the minimum free space is set to 0 for the cached share and the cache only share. As for validating, I looked at \\tower\cache and could see the files I had just copied to Z: (my media share). Unless it starts to slow down again lets chalk it up to gremlins and get on with our day.