Wow, this is a good one! Performance ATEs are fantastic! Will this one be extended to / finished at EMC World?

This event is now open for questions. We look forward to a lively and interactive discussion.

Can you speck in generalities how specific settings / practices to achieving the most performance form a VMAX?  obviously things like spreading your workload, etc, but any more details such as changing the PRC on the SSD pool if you'll never bind a lun to that pool?

Recently my organization has been on a dashboard kick. We're building a VMAX performance dashboard in vCops using the SMI-S provider hooks.  Some of the statistics we ned to show are FA health including CPU and MB/s.  The challenge i'm running into is knowing the max MB/s an FA/FA (They would like to see IO/s as well, but that isn't feasible).  Not limited to FA MB/s, but other status as well.

Does such a document list which give guidelines as to what are acceptable numbers?  I realize each environment is different,  but information on how to get to those number for specific environments would be nice.

Is there a rule of thumb about the amount of hosts that should be attached to a single FA port? And how many ports would you advise to be used for a single host?

I remember when I first installed a Symmetrix 10 or 12 years ago we talked a lot about this - Fan/in Fan/Out ration.  You don't hear as much about this anymore.

The answer has to be it depends.

1. on expected workload for each host. 

2. on expected growth.

3. it has to be noted that while there is 2 ports on an FA there is Single CPU to Manage workload between both.

From a best practice perspective you really need to start by using 4 port across 4 different FA CPU's .

If the workload is expected to be IO intensive, heavy thruput or most importantly respsonse tme critical it well worth considering 6 or 8 ports again spread on 6 or 8 different CPU's .

This will help avoid future issues relating to excessive utilization of CPU and or excessive queuing on the FA itself

Fan in and Fan Out ratio is a good principal.

My recommendation would be to look at Utilization rates and queuing rates on the FA's using unisphere performance monitoring tools. Where possible add to the less utilized FA's and you should end up with a very balanced front end on the VMAX thus getting optimum IOPS and response times.

My question is what tools are you currently using

the latest version of Unisphere can offer some comprehensive data and reports that can show the utilization rates.

it can even give a basic heat Map which give very quick snap shot.

PRC on SSD pool should be 1%

10% is the recommendation on FC pool and this is where all data should be bound initlially

1% on SATA should be sufficent on the basis there is no app bound to it.

please send any other specific parameters and i will try to give you a best practice .

Kevin Gleeson wrote: Fan in and Fan Out ratio is a good principal. My recommendation would be to look at Utilization rates and queuing rates on the FA's using unisphere performance monitoring tools. Where possible add to the less utilized FA's and you should end up with a very balanced front end on the VMAX thus getting optimum IOPS and response times.

What are acceptable ranges for utilization rates and queueing rates on FAs? If I need to color code things (think vCops heat man) green to red.

Hi,

I want to ask : VMAX performance best practices about VDI

Below 65% on FA CPU is optimum. This should manage queuing.

In relation to queuing it is best to avoid medium to high double digit counts on the higher queue ranges.

For VDI it is key to ensure that there is plenty front end processing available so spread onto as many CPU's as you can.

Ensure that Round Robin Settings on the host are set optimally to allow for best possible load balancing. setting round robin to 1 is a good idea.

Do you have specific questions around Disk tiers etc.?

Is the array a FAST VP array with SSDs ?