1 Nickel

Should I use SmartPools in a Home Directory environment?

A "classic" or "typical"  home directory enviroment has two common characteristics.

1) users browse data much more than they access the files in their home directory. This browsing behavior is the basis for recommending SSD's for metadata storage in home directory environments. This will ensure a fast listing response and minimize user complaints to helpdesk concerning the perception of slow performance. 

2) 80% of home directory data goes cold in a very short time, and therefore should be stored on the most cost effective media.

This second characteristic would seemingly point to the logic of adding a high density, NL400 tier to the cluster configuration in order to provide the lowest cost storage for the majority of cold data.

However, in a standard SmartPools deployment the metadata associated with a file, follows that file as it moves from the primary tier (with SSDs) to the NL400 tier. At this point, the metadata will be stored on NL400 disks, which negates the benefits of SSD drives on the primary tier as 80% of the data will no longer have metadata acceleration and fast listing.

If this condition leads to end user complaints, the solution may require implementing GNA (see definition and requirements in other posts) which may include the addition of new nodes. 

Not all home directory workflows are the same, but the metadata placement as data moves between tiers is an important consideration.

Tags (2)
0 Kudos
1 Reply
3 Zinc

Re: Should I use SmartPools in a Home Directory environment?

Very good observation!

It would be sufficient to keep the "permissions" (etc) stuff -- typically fitting in a 512 Byte block --

of the metadata on SSD, while the "structural part" (up to several KB, organizing the layout of the

actual data) could go with data to the HDD!

Now I don't know wether splitting the metadata into two categories,

"permission" and "layout", is on the agenda...(?)

-- Peter

0 Kudos