Greetings;

Concerning VMware View - With linked-clone desktops, FAST Cache improves overall performance by reducing I/O accrsses to hard-disk drives and directing them to Flash drives. Specific use cases include boot storms (up to 99 persent reduction in hard-disk I/O access), and virus-scan operations (up to 77 persent reduction in time required for a full scan of the desktop. Since less I/O accesses are done from the HDDs, the same number of users can be supported with 44% percent less HDDs.

Oracle, SQL and Celerra unified file services also benifited with FAST Cache.

I have uploaded our whitepaper titled EMC Fast Cache - A Detailed Review that discusses FAST Cache and FAST VP.  On page 18 there is a comparison chart between FAST Cache and FAST VP that will help you decide what is best for your environment.

Nick

The documet is located at:

https://community.emc.com/docs/DOC-12903.

Nick

Personally I am not aware of any tools.  However, I am not in a position that would utilize them.  I will leave your request open to the Community for responses.  In the meantime I will check my internal sources.

Nick

We have our 'Analyzer Helper' tool, which shows the results in charts.  Ask your TC to arrange a demonstration with one of our local 'Speed Gurus'.

A simple way to look at this is that FASTcache looks at 64KB chunks with a live algorithm and FAST auto-tiering looks at 1GB chunks on a timed in an on-request fashion.  If the choice is one or the other, I would suggest FASTcache.  The granularity of the choice for what data resides in cache is my primary reason for this choice.

I would definitely continue to carve the NL-SAS drives from a pool so that you can at any time in the future place SSD drives in the pool and get the auto-tiering advantages.  With both enabled you would get the benefit of seemless tiering of course, but also of being able to ensure certain LUNs are served from your fastest drives at all time.

Hi Nick,

could you please provide link to this document "EMC Fast Cache - A Detailed Review" ?

Thank you for this link. I have read it and found that it is possible to analyze current workload before deployment. Do you have any special tools to make such task easier ?

I know the EMC TC's can pull some information out of the VNX (NarData) and can give you a pretty good idea of the benefit of using the EFD's for FastCache. I would reach out to them.

You had asked above:

"Which option: FAST Cache or FAST VP would be the most effective from performance point of view ?"

My opinion is FastCache is where I would put my EFD's if I had to decide between the two choices. It gives more flexibility. I would then put the rest of your spindles into a FAST Pool (without EFDS) and let the array handle it from there.

Now, there are certain uses cases where FastCache wouldn't be a good fit but from what I see above, you have a pretty good chance of it being pretty valuable.

The great thing about fastcache + FastPools is if you do get a block or 2 that makes its way down to Tier 3 and then it gets "hot" again, instead of waiting a few days for it to migrate back up the stack, FastCache copies those blocks right into EFD's so you go from 100's of Milisecond response times (YUCK!!) to less than 10's of milliseconds of response time.  I like to refer to it as the "Turbo Boost" for data blocks !!

Make sense?

Just a quick update - i published a blog called "Fast Cache - the answer to the Automated Storage Tiering Problem" that might help shed more light on why you may want to pick FastCache for now.  it's here: http://vtexan.com/2011/11/22/fastcache-the-answer-to-the-automated-storage-tiering-problem/

Message was edited by: Tommyt

przemol, EMC has tools like TAV for USD to analyze the busy luns which  may be suitable for FAST VP or FAST Cache. Small size IO will benefit more from Fast Cache while FAST VP is better for larger size IO. Sequential read and write need to be eliminated. You can ask your TC to submit a case to SDC in the SalesForce.com

What array (make/model) do you currently have in your environment?