Centera Storage Nodes Capacity

Whenever data is written to a cluster some KB will be utilised for metadata information besides the actual file size
CPM - 2.4MB per node (2.4 MB will also be duplicated to a second storage node)
CPP - .4MB per node (2.4MB divided by 6, as data is stored in fragments in CPP mode)

Dear Mamatha,

thanks for your answers, but unfortunately you didnt answer my original question with that. I want to know the precise value of the maximum capacity (in TB, GB or whatever) per node, if the archived data is expected to be 2,4MB of average size. Since CPP and CPM differ regarding their capacity requirements I would like to know this precise value for each protection mode.

So an excellent answer would be:

Thank you,

Markus

Hi,

You can alternatively get a rough estimate by the capacity of each node.

The following command "show node detail c00xnxx" will list the available capacity in the format below,

       Total Capacity:             XXXX GB

       Used Capacity:                 xxxx GB

       Free Capacity:              xxxx GB

       Faulted Capacity:              xxxx GB

       Total number of objects stored:   xxxxx

Kaja Ismail

Hi people,

thats a very helpful table indeed. But how can i calculate the available capacities for a node using CPM or CPP. I have a feeling that this depends on the avg. file size (2,4MB) of the archived files. Which factor(s) do I have to use on the "available capacity" in your table?

Thanks,

Markus

Hi Markus,

The information you were sent is correct but I do not think that is the question you were asking.

If I understand your question:

"I'd like to know the maximum capacity of a node in storage role if theavg. file size is 2,4MB for CPM and CPP mode"

It makes no difference to a node, the node will store the maximum data that it can based on the size of the node.

The difference if you are using CPM or CPP is how much data is stored on the cluster.

You will get more 2.4 MB files on a CPP protected cluster then on a CPM cluster.

One way to look at this is that:

in CPM it takes 2 nodes and the capacity is 2.4 x 2 or 4.8 MB of capacity used for the file in the cluster.

In CPP it takes 7 nodes and the capacity is 2.4 x 1.25 or 3 MB of capacity used for the file in the cluster.

Hope this is what you are looking for.

Regards,

Dan

Hi Dan,

that's exactly what I was looking for, thank you. But still I think that it's not so easy. If what you say was true, then I if I had a cluster of 8 nodes gen4lp (2TB) with about 7.2TB per node, so 8*7.2TB = 57.6TB in total, I just had to multiply it with 0.5 (CPM) or 0.75 (CPP) to receive the amount of data that could be stored on the cluster. But if you look at the current technical data sheet of the centera (http://germany.emc.com/collateral/hardware/specification-sheet/c932-centera-cas-ss.pdf) you see, that this is not the case.

That's the reason why I am looking for an up to date table likes this one:

which I found in the centera online help pdf powerlink (which unfortunately wasn't updated since 2005). Additionally to that I think, that it is not possible to use all of the nodes in a cluster for storing data since I do need some nodes in the access role too. Or is it already common to use the feature which allows you to use both roles on one node (and does the customer still have to buy a seperate licence for this?)?

I really do appreciate your gelp guys,

Markus

Hi Markus;

You can use as a rule of thumb that under CPM each time you write data it will be written twice or mirrored so the capacity used to write this new data is 2X the original file size. What we use under CPM is that you will use about 1.125 times the capacity for the data we write. There are things that make the numbers harder to compute.

As a quick guide in CPM cluster about 50% of the capacity is "new" data

In CPP cluster you can get between 82% to 85% of capacity is new data.

The spec sheet you looked at has:

Raw Storage Capacity : 16 to 256 TB per EMC Centera cabinet:

A EMC cabinet can hold 32 nodes or 2 cubs of 16 nodes each per cube.

The minimum size of a cluster is 4 nodes so in this speck it seems that they are using 4 4TB nodes to get 16 TB and 32 8 TB nodes to get to 256 TB.

In CPM you would have if all was perfect and no over head 8 TB minimum of new Data in 16 TB cluster but in reality it is 7.7 TB and 125.7 TB in a 256 TB cluster.

In CPP the minimum cluster would be 4 TB nodes X 8 nodes for 32 TB with over head you get about 83 or so % so that is 26.4TB in the speck sheet and a max of about 215.5 in a full 256 TB cluster.

I have no idea what the data table you list is showing and I can not find it to check for a referance.

It is common for Gen3 and Gen4 nodes to be used for both access and storage, I am not sure about the licence.

Regards,

Dan