We are ready to roll out flex-vols and are discussing about the possible performance impact of having lot of flex-vols on a aggregate. We are talking about having 20-30 flex-vols per aggregate.
Please share your experience/recommendations ?
We're using fas3050's as "flex caches". The busiest ones have 60 + flex vols on them in a 600G aggregate. Pay close attention to the NetAPp best practices for Flex Vols, and have a good understanding of your environment, the expected traffic thru each volume, etc, to set the sizes and reserve as appropriate in your environment.
Each flexcache volume has some memory and disk overhead, just to exist. As long as your total active working set doesn't exceed the aggregate size, sizing should be OK. Note to that flex cache volumes impose a load on the origin filer, too, so you'll want to monitor that as well.
-skottie
Premanshu Jain wrote:
We are ready to roll out flex-vols and are discussing about the possible performance impact of having lot of flex-vols on a aggregate. We are talking about having 20-30 flex-vols per aggregate.
Please share your experience/recommendations ?
I mis-red this, and my brain inserted the phrase "flex cache vols". Thats been a focus for us lately. so, when someone asks that question, at least one answer's been posted ;-)
As for local resident volumes, "use them instead of qtrees" is what I've heard from NetAPp, which implese "use alot of them".
20 - 30 should not be an issue.
-skottie
Premanshu Jain wrote:
We are ready to roll out flex-vols and are discussing about the possible performance impact of having lot of flex-vols on a aggregate. We are talking about having 20-30 flex-vols per aggregate.
Please share your experience/recommendations ?
Premanshu> We are ready to roll out flex-vols and are discussing about Premanshu> the possible performance impact of having lot of flex-vols Premanshu> on a aggregate. We are talking about having 20-30 flex-vols Premanshu> per aggregate.
Premanshu> Please share your experience/recommendations ?
As I recall, NetApp limits you to 100 (or is it 200?) FlexVols total. Though that might be per-aggregate. And traditional volumes are less than that, since they take up two slots, one for the private aggregate, and one for the volume itself.
But you have upto 4995 qtrees per-volume.
My question would be why not use qtrees instead of volumes?
In any case, 20-30 flex vols isn't a big deal.
John
The story I know is that total number of aggregates, FlexVols, and TradVols can't be more then 200 per system. Before 7.0.1R1 there was a limit of only 50 FlexVols per system because of the speed of cfo in taking over that number of volumes, which is bug 149737. The eventual promise of FlexVols is to replace qtree's, but NetApp is still in the process of doing this, and there is still a long time before this is possible in larger environments.
Sorry, this is getting kinda of off topic.. moving right along.
-Blake
On 1/27/06, John Stoffel john.stoffel@taec.toshiba.com wrote:
Premanshu> We are ready to roll out flex-vols and are discussing about Premanshu> the possible performance impact of having lot of flex-vols Premanshu> on a aggregate. We are talking about having 20-30 flex-vols Premanshu> per aggregate.
Premanshu> Please share your experience/recommendations ?
As I recall, NetApp limits you to 100 (or is it 200?) FlexVols total. Though that might be per-aggregate. And traditional volumes are less than that, since they take up two slots, one for the private aggregate, and one for the volume itself.
But you have upto 4995 qtrees per-volume.
My question would be why not use qtrees instead of volumes?
In any case, 20-30 flex vols isn't a big deal.
John
-
Blake Golliher -
John Stoffel -
Premanshu Jain -
Skottie Miller