Keith Brown wrote:

Alexei...

I'm going to leave your shelves-per-loop and loops-per-filer in the cluster scenario questions for our clustering product manager to address. He will have the latest information on exactly what configurations we have tested and are supporting on which filer. However, I will take a shot at the obvious question...

Thanks Keith, as the product manager for Clusters I guess I better speak up, although I don't manage storage so this isn't really my area. In a Cluster there is no change to the maximum number of disks/shelves that a single filer might have. Once a takeover occurs, you are left with a single filer managing all the disks, therefore the max limit for a single filer is the max limit for a Cluster pair.

I'm gong to straighten out a few other things and then go back to your original message and try and answer your questions

...

Ok, so the hardware guide says that a f740 and an f760 can both only have 112 fcal drives, 2 fcal loops max.

How does the 760 then do 1.4TB of storage and the 740 960GB? Seems there

are

...

56 or so drives missing.... another loop?

This was/is just a quality assurance and testing lag. The F760 should be going to the full three FCAL loop configuration in the near term (if it hasn't happened already?).

I'm afraid is was more than a testing issue. One of our vendors hard a serious flaw in their part. We waited for a respin of the chip involved and still found a flaw. We have a new strategy to substitute a different part, but it won't be a near term solution. I wouldn't expect to be able to have three loops (four with clusters) on an F760 until after New Years.

Our latest GA release, 5.1.2, just officially gave the full 3 FCAL loop / 168 drive capability to our former high end unit, the F630, so hopefully the new F760 isn't too far behind! :-)

This is not true. The 5.1 release which went EA in June added three loops to the F630. We have been running a full 1.4TB or 168 disks on an F630 for some time now.

The F760 is behind not because of software problems but due to hardware issues (ya, I know, spoken like a true Software Product Marketing dweeb and yes my pals that handle the Hardware Product Marketing will certainly give me a hard time after they read this and no I'm not worried cause there are more Software Product Marketing people than Hardware Product Marketing People in the group :-) ).

...

So, my bottom line question is: in a CF setup with 2 f740's, what is the

maximum

...

*useable* storage space I will have (combined)?

Don't take it as gospel, but I *think* you'll probably be looking at a couple of FCAL loops-worth. In other words, ~960 GB combined (~480 GB per filer). The actual *useable* space is dependent on some other configuration specifics.

Sort of, but if you want to calculate it you'd want to know how many RAIDgroups (actually how many parity disks), how many spares, and then use the 8.6 GB per disk formatted capacity to figure usable data and recognize that about 10% reserve is set aside for each file system.

For instance lets say you wanted 14 disk RAIDgroups, one volume per F740 Cluster node, two spares per node [Spares are global to a node but not shared between node members]. You start with 122 disks, minus 4 parity 2 spares per node or minus 12 disks - so you'd have 100 data disks, times 8.6 GB equals 860 GB minus the reserve leaves about 775 GB. If you went with 28 disk RAIDgroups and only one spare per node you'd have 106 data disks or about 820 GB of useable space.

...

Also, why does a f630 get to have 3 fcal loops?

Because it's been tested, certified and stamped as "approved" in a 3 FCAL loop configuration by our quality people.

In a Cluster configuration the F630 is allowed to have up to 4 loops, but the maximum of 168 disks can not be exceeded. This allows for things such as balanced disks where for instance 12 shelves or 84 disks are on each node. Of course you don't have to have balanced disks, you might want to put 16 shelves on one node and only 8 on the other in which case you'd only need three loops. That's fine, in fact you can even do 23 shelves on one F630 and one shelf on the other. We have a few customers that like this sort of hot standby config. The 168 disk restriction is required since once a takeover occurs, the F630 will have to support all the disks and as Keith has said out QA group has only qualified a max of 168 disks for an F630.

...

Boy, I am just full of questions tonight...

Tell me about it! :-)

Keith