Hi,
We have a file server cluster in windows 2012 that is used as backend access to our app servers. This is a large environment and lots of access 24/24h a day by our app servers.
Each node has 64GB of RAM, MS NIC Teaming with 10GBbps per nic, etc...
During business hours the server performance is acceptable, but we have a serious performance hit at night when the backups start, during the first 1 or 2 hours the access to the files in the file cluster become very slow...
I already check several times this problem with no effective solution, hopefully you guys can throw other different ideas for us to test.
So here is the scenario:
- The problematic share is one folder with aprx. 4.000.000 files that range between 4k and 1Mb.
- During working hours the average access time to the files is between 3ms and 1sec MAX. At night when the BKs starts, during the first hour or two of the Backup process, the access time can reach up to 60 seconds and this causes the app servers and web servers to timeout which causes problems for our clients.
- Note the backup (full backup) takes aprox 12 hours. The daily backup takes 4 Hours but we still have the issue (the drives are backup one of the time to not cause large impact)...
-I already check the storage and everything is apperantly ok, the network appears to not be a problem as well (the app servers and the file cluster is in the same vLan segment), no CPU increase, and so on...
- I also ran perfmon, windows performance toolkit (WPT) and others... From my perspective the problem is related with the file access in the disks/luns... But I don`t see any credible evidences of that, howerver in WPT reports the first 2 hours of large operations may be related with snapshots being taken to all the files and this may be the problem...
So, for windows 2012 what recommendations or ideas do you suggest?
Thank you.