ONTAP Hardware
ONTAP Hardware
I have tried different setting for the eseutil verification throttle but they seam to change filer load and throughput randomly. Can you recommend any good rules of thumb when setting this value?
Thanks
Bren
Solved! See The Solution
Yes unfortunately there really isn't a magic guide as to what will yield what results. This tends to vary alot for different environment so there isn't really a best practice or guide table.
Brad
What is the current verification throughput and what are you trying to achieve from that perspective. Verification throttling will allow throttle the amount of I/O load the verification process places on an Exchange Server. So as far as to which way to adjust these settings depends primarily on the current load being placed on the server during verification and what you are targeting.
Brad
The SME reports it at between 60 and 110 Mbs on iSCSI. On my FCP and new FAS3170 boxes it has climbed to 160Mbs. I need to balance the load on the filer verse throughput on verification jobs. I have seen I can get more throughput and lower filer load using the throttle option but the hard part is getting the correct number.
Bren
The basic rules for verification throttling are based more on disk latency. The trick is to figure out how long to configure the the verification to "sleep" to lower the latency to acceptable limits.
The first step is to monitor the disk latency during verification. So you'll want to look at the Average Disk/Sec. Read and Average Disk/Sec Write counters. To be in line with Microsoft practices these need to be below 20ms. You will want to look at these counters for each LUN containing the DB's being verified. During the verification time if the latencies are above 20ms then the verification throttle can be adjusted in order to keep the latency within the acceptable limits.
Brad
That is more or less what I have been doing but I was looking for a 'magic' guide for want of a better term.
Bren
Yes unfortunately there really isn't a magic guide as to what will yield what results. This tends to vary alot for different environment so there isn't really a best practice or guide table.
Brad