Hi Glenn,

the points you mentioned are exactly the points which make me worry. But there are some interesting points. Typically when you talk about IOPS und latency. I will make up my mind about this on the weekend :-)

Best Regards and thanks a lot for your ideas

Jochen

-----Original Message----- From: Glenn Walker [mailto:ggwalker@mindspring.com] Sent: Friday, September 15, 2006 2:40 PM To: Wischer, Sebastian; Willeke, Jochen; NetApp Toasters List Subject: RE: Performance Indicators

I believe that this is more along the lines of QoS (though not true QoS and not guaranteed).

What Jochen is looking for is a bit difficult: you can't really guarantee performance on the head due to workload differences.

Each of the machines Jochen listed is a very capable machine - but selling IOPS to a user is not easy. Each workload is different, each application would require a different amount of IOPS in order to facilitate functionality (hard to define these sometimes), you can't easily segregate workloads and adding workload from one user\system would definitely affect existing workloads.

For example: CIFS typically requires about 5 IOPS to open a file as a minimum - if you add some newer functionality from 2k3\XP and MS Office, you could be talking about hundreds of IOPS to open a file (pulling file metadata).

Perhaps the best thing to do is to talk to sales to get actual numbers, but simply list the capabilities of the filer and manage how the workload is spread among them.

One last point: when you speak of performance, you mentioned IOPS - this is workload, not performance. Performance is more a measure of latency at a particular level of work: ie, how much latency for a given amount of IOPS. If I stated that the R200 was capable of about 12000 IOPS (CIFS) per second, that would be true - but I could also position it as being capable of higher if the latencies of the workload were to increase (still the same amount of work, but response time would decrease). You could speak to the user-base about performance in terms of latency and IOPS.

Glenn

-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Wischer, Sebastian Sent: Friday, September 15, 2006 5:18 AM To: Willeke, Jochen; NetApp Toasters List Subject: RE: Performance Indicators

HI Jochen,

have you think about FlexCache?? There you give volumes priorities. I know it is not exactly what you are looking for but perhaps with the help of FlexCache you can sell categories to your departments. Here is a small summary from a NetApp slideset regarding FlexCache:

FlexShare(tm) is enabled when there are system bottlenecks CPU NVRAM Disk

Control of priority by volume Five priority levels

Sebastian

-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Willeke, Jochen Sent: Friday, September 15, 2006 10:32 AM To: NetApp Toasters List Subject: Performance Indicators

Hi Toasters,

right now i am thinking about any kind of performance indicator for my fileservice. We do supply storage to other departments of our company.

Of course one thing i can sell is space. But there is another indicator which is performance. Has anybody good ideas how to define the performance i can grant to a special service?

I was thinking about I/O's per second but do not know what a good number would be. Has anybody ever had to work out those values and can give me something to think about? I am interesting in any experience you had with this topic...

I/O's, data-throughput, responsetime....And what are good values which i can grant with netapp systems? We use R200, FAS960 and FAS3050.

Thanks in advance

Jochen