Isilon OneFs Cluster size

See if the following makes sense:

8 TB per disk is disk maker's capacity (1000 based), which gives you 18 PB for the cluster.

Software disk storage (1024 based) converts it to about 6.8 TB, which gives you about 15.7 PB for the cluster, which is round up to 16 PB.

Perhaps the cluster is taking your protection level into account?  At N+4n, you are asking for sufficient redundancy to survive four offline nodes simultaneously, which at 59x8TB means each node is ~472TB of storage.  That's just about 1.8PB of space reserved for redundancy. Add in the 0.2PB of utilization and you get up to 2.0PB of unavailable storage, resulting in 16.PB remaining.  Based on that, it would seem this graph is showing you the available space with overhead factored in.

What version of OneFS?

Also was it the Isilon Sizing Tool that you used?

Thank you All ,,

bang11

if we convert it to 1024 base it should give 7.8 TB right ? 8000/1024=7.8 TB

mikekozar

I think you are right but I am confused also inside each node its showing 425 TB HDD available so how this calculated ? it should give me 460 TB ROW after converting to 1024 base.

RobChang-Isilon

OneFs Version 8.0.1.0

Isilon Calculator on EMC web site there is like table showing the 39 node with n4+ how much usable it will give you

The pie chart and the outputs of isi status and df -h /ifs

should all be consistant, and are measured in 2-based units 1 "KB" = 1 KiB = 1024 Bytes and so forth.

df -H /ifs  shows size and usage in decimal units, 1 KB = 1000 Bytes.

All outputs display the "raw" capacity before protection overhead is applied.

On your Isilon, have a look at the outputs side-by-side:

df -h /ihs

df -H /ifs

hth

-- Peter

@peter_Sero

thank you , correct

its showing 16 PB RAW in the command 10 base ,  but i am trying to calculate the capacity RAW

* 8TB disks to 10 base

8000/1024 = 7.8125

* 7.8125 * 59 disk to calculate 1 node RAW capacity

7.8125 * 59 = 460.9375

* node capacity * total number of nodes 39

460.9375 * 39 = 17.9 PB

Apologize for the quick and dirty job earlier.  Instead of using 1024, I used 1040 in my calculation due to typo.

8 TB disk => 8/1.024^4 = 7.3 TB software

With 59 disks per node and 39 nodes => 16.7 PB, which is round off to 16 PB (matching the 16.0 PB in your graph).

Still difficult to see where base-2 and where base-10 units got applied...

Lets start with base-10 and, for sake of simplicity, 1 disk = 8,000,000,000,000 B = 8e12 B

=> 39 * 59 disks = 2301 disks = 1.8408e16 B = 18.408 PB (still base-10)

divide by 1.024^5:

=> 16.3496 PiB (base-2)

I think that's close enough to the pie chart: 16 PiB + 208.2 TiB + 17.3 TiB = 16.22 PiB

also taking into account that OneFS uses slightly less than the nominal disk capacity.

Another approach to really nail it down is to inspect the output of

sysctl efs.bam.disk_pool_db

(here, 1 block = 8 * 1024 Bytes = 8 KiB)  [EDIT: fixed typo]

-- Peter