Struck by Lightning?

by teri.shiozaki on 02-11-2012 02:13 AM - last edited on 02-14-2012 05:13 PM

By Tom Joyce

 

In my last update I commented on the status of solid state disk technology.   This is one of the most

interesting areas of activity in the storage industry today, and we are in the very early stages of seeing how this will change the landscape.  HP currently supports solid state disk technology in or arrays, including HP 3PAR, and we are investing heavily in data services technology that can deliver unique capabilities with SSDs.  One of the most important areas we are focused on is tiering.  We have powerful tiering technology in 3PAR that allows customers to not only move workloads between classes of storage within the array, but to create “hybrid” tiers that spread the workload across a mix of magnetic disk and solid state storage to hit specific price/performance objectives.  This takes advantage of 3PAR’s unique wide striping architecture, and is a key technology differentiator for HP.

 

When it comes to performance, architecture matters.  The 3PAR architecture is decades younger than some other leading competitive storage platforms.  This architectural advantage is the reason the new 3PAR P10000 set the world record on the SPC-1 performance test a few months ago when we released the product. We did this with regular hard disk drives on one array, and beat results posted by a number of competitive all-solid-state arrays.   

 

Since then, IBM posted an SPC-1 number that managed to just exceed our 3PAR benchmark.  They did it with a total of 16 IBM arrays behind a SAN volume controller.  3PAR did it with one array, in a real customer configuration, that costs a heck of a lot less and is a lot less complicated. 

 

Also, EMC recently announced their VF Cache solid state flash card, aka project “Lightning”.  This card will go in a server and will be used to increase the performance of applications running on EMC arrays by moving I/O out of the array onto the server.  The goal is to eliminate the latency associated with EMC arrays.  This is probably a good idea, because EMC arrays are getting long in the tooth and need a performance “kicker”.  EMC does not participate in SPC-1 or typically allow publication of any benchmarks of that nature, but suffice to say, we are more than comfortable with competing against EMC’s older arrays in customer bakeoffs. 

 

Here is just one simple example of why architecture matters, and why EMC may need extra cache:  data center storage today REQUIRES thin provisioning technology.  Without it you have to over-configure your storage configuration and run with low disk utilization to handle today’s workloads. This is why so many EMC customers have low storage utilization.  3PAR was designed with thin technology built in from the beginning.   EMC had to add their thin provisioning later, and it is very limited.  In their implementation all thin devices are cache devices.  So the more thin you try to use, the more cache you need, because each thin device has a cache overhead.  In some configurations you might need TERABYTES of cache, just to use thin provisioning.  EMC actually limits the number of read requests that can go to a thin device because of this.  They recommend that customers should only use their thin provisioning for applications for which some performance impact can be tolerated.  They have more “beware” messages in their materials that I can mention here (you can look it up). 

 

3PAR, on the other hand, runs thin for the whole box, for all applications, period.  Turn it on when you install the box, and blast off.   

 

With regard to Lightning, what customers should ask themselves is, why should they have to band-aid poor performance on older storage arrays?  Customers should also evaluate VFcache with a critical eye, and note that it is SLC only (expensive) and has limited capabilities for VMware.  For example, it doesn’t work with VMotion.  Given that a majority of servers are virtualized today, and that VMware is one of the most significant areas where solid state might help improve performance (and that they own VMware), you would think that EMC might have sorted that one out in advance.

 

I should note that HP also has flash server cards.  We have the good fortune to have an entire server business!  Beyond flash cards, there are a great number of optimizations on the server side that you can perform when you own the server business.  The same is true in networking, and we are in that business too.  EMC is not.  VF Cache is really EMC’s first foray into the server hardware business.  A flash card does not a server make… 

 

At HP we are delivering true converged infrastructure, across servers, storage and networking.  We have now delivered new technology in each of those businesses, and integrated that technology.  We have delivered complete end-to-end converged systems, including CloudSystem and VirtualSystem.  And yes, over the coming months and years you will continue to see exciting new solid state announcements from HP, in both our server and storage products. 

 

But there is one thing you won’t see:  expensive band-aids.  We don’t need them.

 

For more info:

HP 3PAR

Converged Infrastructure

 

We encourage you to share your comments on this post. Comments are moderated and will be reviewed and posted as promptly as possible during regular business hours.

To ensure your comment is published, please follow our community guidelines.

Post a Comment
Be sure to enter a unique name. You can't reuse a name that's already in use.
Be sure to enter a unique email address. You can't reuse an email address that's already in use.
Type the characters you see in the picture above.Type the words you hear.

Find HP in Social Media

Facebook Twitter YouTube SlideShare Flickr
About the Author
Labels