Ask the Expert: All about VMAX performance best practices

How can we determine that my FAST VP Relocation rate should be set to?

How do we best determine the correct Workload Analysis Period to ensure data is best use of performance tiers?

Initial workload period should be 24 hours.

The more important 1 is Workload analysis period and this should be set to 7 days or 168 hours. This allows FAST VP to work from 7 days of stats, thus creating a balanced workload profile in order to calculate movements

Hi Kevin,

This is a question from a Chinese user:

Companies generally use VMAX to build a hybrid circumstance, will VDI in such circumstance impact other applications?

If I build VDI with VMAX, and only SAS disks being used, how to achieve the best performance?

Thank you in advance!

Kevin Gleeson wrote: Initial workload period should be 24 hours. The more important 1 is Workload analysis period and this should be set to 7 days or 168 hours. This allows FAST VP to work from 7 days of stats, thus creating a balanced workload profile in order to calculate movements

If Workload Analysis Period is set to 7 days, how can we account for odd (but planned) bursty workloads such as end-of-month batch cycles, which compete with a more sustained high-performance workload?  Wouldn't the later be more likely to be promoted, since it has the historical need for performance and the month-end devices have been largely under utilized for the workload analysis period?

When Building VDI the critical to do is spread over a good number of FA CPU's and ensure that Load balancing is OPtimized.

In relation to SAS drives, what config had you in mind and on what VMAX product.

The trick with VDI is to use a good number of spindles.

Also Things like Virus scans etc on the VDI's themselves should be staggered and not all allowed to run at the exact same time.

If Workload Analysis Period is set to 7 days, how can we account for odd (but planned) bursty workloads such as end-of-month batch cycles, which compete with a more sustained high-performance workload?  Wouldn't the later be more likely to be promoted, since it has the historical need for performance and the month-end devices have been largely under utilized for the workload analysis period?

As mentioned best practice for most arrays is Seven Day analysis period.

Fast VP will react during bursty workload at month end assuming the workload is sustained.

I would not recommend changing from seven days.

what happened to all architecture docs (Product guide used to explain operation symmetric to goo detail).?

Is it appropriate to include backup server luns into fast policy.. usually backups will be performed during the non-business hours and moving the backup sever luns to higher tiers would benefit them a lot..??

Is it appropriate to include backup server luns into fast policy.. usually backups will be performed during the non-business hours and moving the backup sever luns to higher tiers would benefit them a lot..??

The not so clear answer is it depends.

There is nothing wrong with having back up luns under FAST VP control Assuming there is constant frequency to the back up. Fast VP will make it's decision based on 7 days of stats against a piece of data. Assuming there is available capacity to promote the data and sufficent evidence of benefits it will promote.

Also Some back up luns have constant traffic ... if that is the case they should be under FAST VP

All release notes and porduct guides should be on EMC support website.

Is there a particular document you are looking for and i will try to provide a direct link for it.

suppose the backup server luns would be very active from 20:00  to 08:00 hours and will be completely inactive during the rest of period, production server luns would be active during the rest of hours..

will the FAST VP would move the backup server luns to higher tier during the 2000-0800 hours and will demote them to lower tiers after that ?

The reason why I ask is bcoz since the backup server luns will be more active coninously they would definitely benifit if they move to higher tiers but it should not be at the cost of production server luns.

The detailed product guide explaining the Virtual fabrics. Data compression etc   is the one I am looking for. The recent version is very high-level and lacks details.

Backups are 100% writes. They will not benefit from FAST policy. As long as you have enough backend bandwidth to destine the data you should be good. I usually used to limit the cache utilization to avoid impacting production luns.

That may throttle the backups though.