Highlighted
MTernesto
1 Copper

Smartpools tiering to an Data on SSD tier

Jump to solution

Curious about a question a customer asked me about tiering to SSD some critical data set. My Idea was to use smartpool policies to do this tiering, or to change it manually using the isi set command.

Does Anybody use Smartpools to tier data to SSD by changing the SSD Strategy on the File Pool Policy? what is your experience so far.

0 Kudos
1 Solution

Accepted Solutions
knargi
1 Copper

Re: Smartpools tiering to an Data on SSD tier

Jump to solution

MTernesto,

You have a couple of questions here so I'll do my best to answer both.

There are customers that use SSD's for Data. This option is called Data on SSD's and you must do this using SmartPool's File Pool Policies. Its important to remember that if you do this and put "critical data" on SSD's you no may no longer get the benefits of data written into the cluster having the metadata placed on the SSD's.  If your dataset is really large you may fill the SSD's up with data and not have any space for the metadata to be written forcing it to go to HDD.

To answer your question further, I have have worked with customers that do not have a lot of files or need metadata acceleration but have a workload that will benefit from fast read access from the SSD's. In these cases the customer has defined a directory and created a file pool policy and placed that directory on SSD's.  In these cases, the customer seemed pretty happy with the performance that they were getting by having their data on SSD's.

Also important to know that just because you can put data on SSD's doesn't mean that the performance is going to be any better.  If you are trying to put a transactional database on SSD's you might see a little performance improvement but not much.  You really need to understand what the customer application is and what the workload profile is in order to make a decision as to whether or not using Data on SSD's will benefit.

0 Kudos
2 Replies
mattashton1
2 Iron

Re: Smartpools tiering to an Data on SSD tier

Jump to solution

Hi MTernesto,

Be mindful of the customer's workflow; SSD is not a fix-all for performance.  There are some cases where spinning disks are faster (streaming performance for one example).

That being said I have a few customers that use SSD for both metadata acceleration (that's where it makes a substantial improvement) and for data in a tier.  For some cases, the SSD's perform well.

What is the workflow?

Cheers,

Matt


knargi
1 Copper

Re: Smartpools tiering to an Data on SSD tier

Jump to solution

MTernesto,

You have a couple of questions here so I'll do my best to answer both.

There are customers that use SSD's for Data. This option is called Data on SSD's and you must do this using SmartPool's File Pool Policies. Its important to remember that if you do this and put "critical data" on SSD's you no may no longer get the benefits of data written into the cluster having the metadata placed on the SSD's.  If your dataset is really large you may fill the SSD's up with data and not have any space for the metadata to be written forcing it to go to HDD.

To answer your question further, I have have worked with customers that do not have a lot of files or need metadata acceleration but have a workload that will benefit from fast read access from the SSD's. In these cases the customer has defined a directory and created a file pool policy and placed that directory on SSD's.  In these cases, the customer seemed pretty happy with the performance that they were getting by having their data on SSD's.

Also important to know that just because you can put data on SSD's doesn't mean that the performance is going to be any better.  If you are trying to put a transactional database on SSD's you might see a little performance improvement but not much.  You really need to understand what the customer application is and what the workload profile is in order to make a decision as to whether or not using Data on SSD's will benefit.

0 Kudos