Tony, thanks for posting this question on the forum!

The best answer is really, both! FAST Cache along with FAST-VP....and here is why;

FAST Cache is going to help the entire storage array overall, it's going to give a boost in performance for all writes coming into the array as well as being able to help with those unplanned read attacks. A little piece of information you may or may-not know, if a block of I/O is accessed 3+ times, it will be promoted into FAST Cache....this is very different from the algorithms within FAST-VP.

FAST-VP works a little different, you can of course choose where data should be written first (highest tier/lowest tier/etc) for the best performance/utilization. However, the array will also look at how often certain blocks of I/O are being accessed and how their accessed and when their accessed. Then based on a schedule (which you set) or via manual effort, FAST-VP will migrate those blocks of I/O to place them on the tier of storage that best meets their needs.

For example, that stale archive data may get into the system and sit on EFD or FC disks at first, but if FAST-VP determines that it is stale it will move it to the inexpensive SATA disks.  Or on the flip side, you have batch jobs that run weekly or monthly but only 10% of the data is actually active....FAST-VP will move that 10% into the faster EFD or FC disks and allow the 90% of inactive data to sit on the inexpensive SATA drives.

So as you can see, the combination of FAST Cache and FAST-VP is a winning one.  FAST Cache for that instant boost of speed, for the entire array...and FAST-VP for better utilization of specific data along with better performance when you need it.

If you have any other questions I strongly suggest you reach out to the community again, or ask your local EMC account rep for more information.

Cheers,

Rick Scherer - vSpecialist

Rick-thanks for the quick response! How/when does a IO block get demoted from FAST cache? Reallocation of blocks with FAST VP is cool, but I don't like to scheduling piece. It seems like for a virtualized environment with many dynamic sub environments, the reallocation will be already be outdated at the next interval. I see how FAST cache can compensate for this.

Tony,

FAST and FAST Cache are complementary technologies and can be used together. If you have a small dataset that fits within the FAST Cache size for the array you can use just the FAST Cache and a single storage tier. However, if you have a larger working set that outstrips the size of the FAST Cache for your array then a FAST enabled pool with two or more tiers is a good approach.

You can learn some more about FAST and FAST Cache here: https://community.emc.com/docs/DOC-8123

~Aaron

Aaron-I agree. Calculating the number of hot blocks/slices of your workload for FAST cache and/or FAST VP seems impossible with the lack of tools available to process this.

Aaron-thanks for your response. Regarding your points:

1. I've inquired about these tools, however I am unsure if these tools exist. If they do, they may be not available for CX/VNX.

2. I totally agree!

3. Great information. FAST VP is all about TCO.

Tony,

FAST Cache behaves in the same way that standard array cache does. Once the array determines cache to be stale and/or you hit a cache watermark (writes) the data will be pushed to disk.

In terms of FAST-VP for virtualized environments, the technical response is it depends. Believe it or not, most virtual environments are fairly inactive, you only have 5-10% of your total capacity actually driving the I/O…this is where FAST-VP shines, keep 90% of your data on inexpensive media and the active 10% on drives that can deliver the performance.

Like Aaron stated, EMC Field technicians have the tools available to show you the benefits of both technologies. I would highly recommend reaching out to your local sales staff.

-Rick

Tony,

Couple of points:

1. There are tools available to the EMC field folks to assist you in sizing a FAST/FAST Cache setup. During the sales process these folks can be engaged to help you figure out what size the FAST tier and FAST Cache should be.
2. FAST and FAST Cache do not duplicate data between them. I.e., if you have data that is located on EFD drives in a FAST tier, that data is not eligible to be promoted to FAST Cache. Since the data is already on the fastest tier of storage there is no benefit to moving it to another set of EFD drives :-)
3. The FAST VP process keeps a running histogram of I/Os to the data in the FAST tiers and makes recommendations (or automatically relocates based on your settings) on where to put the data. We keep the highest tiers highly utilized and move cold data down to lower tiers as we start to fill up the fast tiers. We do this so that response times on the active data are as good as possible. This also ensures that only the most inactive data gets moved to the lowest tiers which is exactly what you want to happen.

Just as a general point, consider that FAST Cache is a 'need it now' performance boost type of solution whereas FAST VP is more of a total cost of ownership (TCO) tool to make sure that you don't buy any more storage than you need over time.

Ryan-thanks for the explanation. I think I've learned more about the technology with this posting than the white papers.

Thanks Rick! I think I got just about all the information I need from this post rather than the white papers. One more question: Do you know when the fundamentals white paper for VNX will be available? I see references to that paper in other white papers. Thanks again.

Tony

Unfortunately I do not know when the document will be released. However, there are a number of application focused whitepapers and reference architectures available on Powerlink that may be able to help your situation.

-Rick