We're currently implementing a Celerra NS-G2 with a DMX-3 backend. We're at the point where we are ready to mask some data luns, but we are still undecided about hyper size.
We currently use 16.86gb hypers by default. We Currently have 2 FA's with 2 ports each zoned/mapped to the celerra. Our issue comes with the maximum number of hypers.
EMC says each FA can have a max of 1024 address per port. If I use 16.86gb hypers, and give each hyper all 4 ports, I can have a max of 1024 hypers. This is only 17tb. Thats not enough for us.
So my options are:
A) Destroy some 16.86 hypers and make 32gb hypers. 34tb would be my new max and thats enough. This solution is good but I have to then have another size of hyper.
B) Only give each hyper 1 port per FA, thereby giving myself a capacity of 34tb. Again, that enough. This solution gives less load balancing and redundancy but we still would have both.
If anyone has a similary Celerra/DMX setup and can give me advice I would appreciate it.
On DMX3 you can have 4095 devices (including meta members) on FA processor. So if you take one FA processor that has port 8A0 and 8A1, these two ports share that 4095 device limit. So that means that you can map devices to FFF address on the FA. These are DMX limitations, i am not sure if Celerra has a limit of how many devices it can see per FA. We use 500G LCFC drives that are carved out into ~58G hypers to get the most usable space from the drive yet not exceeding 255 hypers limit per drive. What are your concerns regarding having another hyper size in the box ?
Are you sure? I know that I can address to FFF, but EMC sent me a doc saying I can only do 1024 per port.
Also if I list all the available addresses on my FA, it tells me that I have less than 1093 left (I have some other hosts on the FA temporarily, but not 2000 of them) and it tells me those addresses are shared between ports on the FA which supports the above number of 1024. If that output is not lying to me, then how could i possibly address 4096 devs?
you can have 4096 devices per FA processor. If you run:
symcfg list -sid 123 -dir 8c -p 0 -address symcfg list -sid 123 -dir 8c -p 1 -address
if you add available address + including meta members (from 8c0 and 8c1) ..it should give you 4096. I have a pair of dedicated FAs, they do not have VCM enabled so whenever i need to present more devices to my gateway i simply map devices to FA, i don't need to do any masking.
Maybe that doc EMC sent you means Celerra can only address 1024 per port, but DMX can definitely address more than that.
1 DMX-4 is compatible with Enginuity 5772.83.75 codes or later. DMX-4 is also compatible with all 5773 codes 2 DMX-4 requires Enginuity 5772.83.75 and above. 3 DMX-3 is compatible with all Enginuity 5771, 5772 and 5773 codes. Please refer to the Enginuity release note for the limitations of the features.
According to that last field (its hard to read), it says 1024.
So I am even more confused.
Disabling VCM? Never heard of doing that before but not having to mask sounds great. Too bad these arent dedicated FA's
Anyway thanks a lot for the help. Now I need to figure out what to believe!
dynamox
11 Legend
•
20419 Posts
•
87439 Points
177
0
Posted October 22nd, 2009 16:00