CIFS performance problem on HP StoreOnce systems

HPs StoreOnce systems do a great job as a backup repositiory. With integrated dedup and compression you can use these systems as a Virtual Tape Library or as simple NAS target using NFS or CIFS to store backup data. If you use them as CIFS NAS, please be aware of the following problem:

There seems to be a big performance problem on all HP StoreOnce systems even running on latest firmware/software that causes file transfers to these systems via CIFS to be very slow.

The problem was found by Veeam support that has several incidents regarding slow performance with StoreOnde systems configured as repository. Tom Sightler from Veeam decided to dig a bit deeper and found out that using Windows client for NFS and accessing the StoreOnce systems from the same Windows machine was significantly faster (up to 4x) than using CIFS protocol. His source machine was running Windows Server 2012 R2. Having this information he switched the OS from Server 2012 R2 to Server 2003 and he instantly got 10x the performance over CIFS than with 2012R2 which is nearly the same as you get from a native NFS implementation. So this problem seems to be related to an inefficient way Windows Server 2012 R2 handles CIFS traffic in combination with HP StoreOnce systems. A general CIFS problem on Windows Server 2012 can be excepted as this problem only appears with StoreOnce systems.

HP is currently working on that problem and until a solution is provided the recommendation from Veeam is to use Windows Server 2003 or up to 2008 (not R2) as repository to avoid this problem. Or switch to NFS either by using Windows client for NFS or using native LInux NFS.

If you see weired performance by using StoreOnce systems with CIFS from a Windows Server 2012 system try switch to Server 2008 to determine if you are also affected by this problem. BTW, this is not a only-Veeam-related problem, they just found the probable cause and a workaround but the problem will probably hit you in any environment where the forementioned componentes coexist.

Leave your comments

Post comment as a guest

Your comments are subjected to administrator's moderation.

People in this conversation

  • Guest - Julian

    found out the VEEAM v8 update 2 has a fix that improves the performance to StoreOnce and Data Domain. here is a copy of the release notes.

    Slow backup storage optimization. Target data mover now caches metadata from backup files, instead of requesting it from backup files residing multiple times over the course of the job run. This significantly improves performance of jobs targeting storage with poor random I/O performance (such as EMC Data Domain), while reducing the overall storage load. Important notes:
    1.The cache is only enabled when using a 64-bit OS on backup repository (or gateway server), and increases RAM consumption of each job by 2GB on average, depending on backup chain size. Using non-default backup block size changes cache RAM requirements proportionally.
    2.Using the default "Automatic selection" gateway server setting for Shared folder, EMC DataDomain or HP StoreOnce based backup repositories disables the cache.
    3.Disabling built-in deduplication disables the cache.

Powered by Komento
joomla templatesfree joomla templatestemplate joomla
2018   globbers joomla template