We have both SPA and Unisphere installed for VMAX. My concern is how to collect the Data and analyze it. and what are the numbers which should be taken into consideration for performance assessement.

If you're licensed for Symmetrix Performance Analyzer or ProSphere, you can obtain a great deal of information about your existing environment.  You can also work with your local sales team to get the performance desk to assist you.  Is that an option in your case?

You can talk to your EMC accounts team and ask for them to look at symmerge.. it's the easiest tool for looking at information like this...  however you can gather it yourself.. but that will only show information about how your current environment is running.. performance is a complex beast but is broken down into four parts: front end (FA, FA ports, and cpu use), back end (DAs, etc), cache, and disk (physical drives)...  

Each part has various stats,  and it's way to much to get into each one.. but SPA does a really good job of showing you the most important ones by default..

The key to performance really is to answer the question "is my system running ok?". .. You need a baseline for "performing acceptable" so you can look for variations in it..

Do you want to do the assesment to understand how the performance might look on the 20K vs. the 10K?  If so, your EMC team and SymmMerge are your best options for modeling a new platform.

If you just want to get a base line, then compare after you migrate, STP data should be fine.  This can be from the service processor or from a host with SE installed.  I would try to get a weeks worth of data at 5 minute samples.

Thanks mmay

Like you said, Performance for VMAX is broken down into 4 major parts: Front end, Back enc, Cache and disk. Do you have any documentation or procedure which makes me understand what are the sub categoories underneathe these 4 major parts and their affecting factors. Also, What will be the threshold numbers for these factors.

Thanks

Ok, I think EMC should do this for you.  Unless you are familiar with the various performance differences between the two systems, it would impossible for you to guess what things might be like on the new system.

Make sure you have at least a week of "before" data from as many sources as possible, host, application, switch, storage, etc.

Also is this a "New" model 10k?  Is it a single engine or multi-engine?  How many drives?  How many engines and drives on the 20K?  Any other workloads also going on the 20K?

Quincy,

Yes, My manager wants me to do a performance assesment on target VMAX 20k so that we can have an idea how the performance factors looks like on 20k. We have SPA and Unisphere performance tools in our enviornment.

Ok Lets put in this way-

I was actually pulling the information from Unisphere Performance tab & SPA and building the charts. But Customer wanted me to analyze the charts and assess it.

I know how to build up the charts and highlight the areas where the performance is not good but have no deep knowledge, how to assess the charts and individual performance factors in detail and compare with the threshold numbers as per the EMC best practice.

We usually concentrate on four areas for Vmax Performance; Front End, Back End, Cache and disk and then drill down to KPI charts and find the red or hot zone under performance impact, this is what on high level I know.

Finding is quite easy but analyzing!!!!! Analyzing is the area where I need assistance.

The best and the easiest way is for your account team to engage a SPEED certified person to do a SymmMerge simulation of the new VMAX 20k based on TTP/BTP file collection from your existing 10K array. He will aslso utilize the symapi_db.bin file from your 10K array.

It can also assure the hardware configuration of your new array and also include any projected increases of the new load.

It will produce a Utilization Heat Map and and graphs for all Front End and Back End subsystems, Disks, Pools and so on.

SymmMerge is an excellent tool for this purpose.

Generally when EMC sells a new Symmetrix to replace another Symmetrix they will include such modeling for the performance of the new array.

There really is no tool for you to use to model the new array vs. the old array.

As long as you have at least the same number of directors, and disks, you should have the same performance or better going from the 10K to the 20K.