PAM II

[babukish11]

Hi PAM II users,

Have you come across following BUG in PAM II, Please share your experiences with PAM II.

http://now.netapp.com/NOW/cgi-bin/bol?Type=Detail&Display=383820

I do have Ethernet based workload.

Thanks

[reinoud7 · Accepted solution]

Hi,

I use today PAMII in 4 FAS6040's and in 4 FAS3170's.

The 3170's are "only" be used for supporting all our  databases (MS SQL, Sybase and Oracle) and they are connected by iSCSI, FC and NFS (depending on the database). The results are amazing. We save between 1000 - 2000 read IO's/second. The CPU of the database servers is dropped (almost 10% for sybase) and a lot of batch jobs go many times faster.

For the 6040's, those controllers are general use. But also here, we have got call's of our users who tell us that some applications go much faster now. Also our Exchange is running on that system and you can see that some calendar operations (open a shared calendar, availability, ...) are going much faster.

We have never seen the problem descriped in that BUG report. We never set the lopri blocks on ON and after a while, your PAM cache is always used for more than 90%.

Was this helpful?

Reinoud

[babukish11]

Thanks for the information, I have followed same configuration, Interested to hear that calendar access is faster in exchange.

Apart from “ stats show ext_cache_obj and stats show -p flexscale-access”, do you use any other tool to  watch performance

thanks

[reinoud7]

We use most of the time: stats show -p flexscale-access.

Most of the time, we just enable the PAMII card for all the volumes (metadata and normal data blocks). But for 2 controllers, we use flexshare to decide if the data of a volume must be cached or not (default only metadata and with flexshare we turn on also the data for some volumes). We only do this when there are volumes with a lot of read io, but there's no raison to cache those data.

Reinoud