Forum Discussion

BEI's avatar
BEI
Level 5
9 years ago

File recall target storage latency

Hi all,

I have a problem with storage latency when restoring files from EV.

The file server is a Windows 2008 R2 and it resides on VMware 6 and it’s located on a Dell Equallogic 6110E with 10GB controllers.
The EV file archiving is installed on a Physical server together with Backup Exec 2015 sp1, and it archives to Dell MD3600i with 10GB controllers.

The EQ and MD are located on separate physical 10GB network including Switches, the servers and storage is located in same rack.

Everything seems ok until someone tries to recall a 3 GB file from archives.
Our latency on the EQ box reaches 135ms and everything halts to a stop until the file is re-called

I have set the recall to 1 min time-out however it’s still a problem.

Any suggestion would be greatly appreciated.

 

Cheers

10 Replies

  • what type of file is it? have you tried pass-thru recall?

  • Thanks AndrewB

    It's a Research files, 3 GB is a single raw data file.

    Have not tried pass-through yet.

    I guess it's something I can look at.

    Cheers

  • I wonder if this could be related to the 14MB/s recall transfer rate?

     

  • I'm refering to this post by me today

    https://www-secure.symantec.com/connect/forums/archive-restore-speed-confused

    Could the latency be due to the fact that the frontend restores takes forever to get the file and hence the latency as it's waiting for the file.

     

  • Hi AndrewB,

    sorry for the late reply.

    I have read about pass through and it might not resolve all of our issues as there are circumstances where the file needs to go back to the original location ie. 1.6TB + for analysis.

    Since the file is located in cache I might have to create 3TB cache - Not ideal.

    Can people still copy the file to another location? doco only mentioned file will go to cache.

    if you can copy will the file first copy to cache and then copy to the users new location? - seems like waste of resources if it does.

     

    Cheers

  • Hi Andrew,

    I have finaly tried the pass through and the speed is 15MB. not much better.

    Cheers

  • i guess the next step would be to trace the entire restore process and narrow down where the delay is coming from. you might want to open a ticket with support since it's going to be pretty involved.

    i just thought of something i remember from one of my projects. the customer didn't have jumbo frames configured on the network interfaces.

  • Thanks for your help and suggestions so far.

    It appears that the Tech dep cannot help me any further with my front end restore speed issues.

    My self and the tech dep have no idea as to what the software is capable of when it comes to the speed. I have asked for an idea of what I should be expecting and in fact I have asked for some kind of matrix that allows you to see what the software is capable of. Apparently there is no such thing and it’s beyond me to understand why.

    I have built a new VM guest win2012R2 server for testing
    Setup a new archive share on a separate store
    all traffic is a 10Gbits
    Copied some files 20GB worth

    Played with front end restores and I get between 18 to 20MB per/sec (its ok if you want recover few hundred Meg.)

    We have even played with IIS and changed some parameters as apparently the restore process uses the HTTP protocol for the front end restores.

    Once files are restored to the original location then I get the copy speed of up to 250MB per/sec.

    So it's not the network or the store or the server, it has to do with the way the files get restored.

    Cheers