Infinidat’s sweet AFA challenge

by Robin Harris on Monday, 5 June, 2017

StorageMojo has observed, many times, that great marketing of a mediocre product beats mediocre marketing of a great product all the time. Thus it is always of interest when someone comes up with an innovative marketing wrinkle.

That’s what Infinidat has done with their Faster than all flash challenge. Their claim is that their system will outperform any all flash array (generally available prior to 1-1-2017) or they donate $10,000 to the charity of your choice.

A safe bet?
Infinidat says their system beats AFAs because:

  • Massive parallelism.
  • Machine learning to optimize caching.
  • DRAM is used to cache hot data.
  • Missed cache hits are usually serviced by NAND.

In theory, they might have some longer-tailed I/Os than an AFA. But if most of the workload is being served out of DRAM, then sure, they should get MUCH higher performance. I doubt they’ll be donating to many charities.

The StorageMojo take
Good marketing does a credible job of presenting and resolving customer pain points. Great marketing links the product to something greater, something meaningful, and taps into the powerful human desire to make a difference. This does both.

StorageMojo supports the Verde Valley Sanctuary, and I’d love to see them get $10k from Infinidat. But I don’t have an enterprise workload to try them out on.

If you have, or will, take Infinidat up on their offer, please drop StorageMojo a line on how it goes. What’s the worst that can happen?

Courteous comments welcome, of course.
Update: From Infinidat’s Drew Schlussel:

Just to clarify, INFINIDAT will make a donation regardless of who wins. The fine print is that the “winner” determines where the donation goes. It is the best example of “it’s all good”.

End of update.

{ 5 comments… read them below or add one }

Drew Schlussel June 5, 2017 at 10:25 am

Thanks for helping to get the word out about our Faster Than All-Flash Challenge!

Just to clarify, INFINIDAT will make a donation regardless of who wins. The fine print is that the “winner” determines where the donation goes. It is the best example of “it’s all good”.

Paul Goldberg June 5, 2017 at 12:38 pm

We are an early partner and reseller of Infinidat and will let you know if any of our prospects take up the offer. We’ve had a few inquiries, so will certainly let you know what happens. Thank you – Paul Goldberg – CEO – Alturna-Tech.

Patrick Mann June 6, 2017 at 1:44 am

This is good marketing but covers up the issue with an architecture like Infinidat. The Infinidat architecture is very much like the XIV architecture – spread data on a lot of drives so that you get decent latency when you go to drives, with the addition of the flash and memory tiers. This works if you have a lot of cold data that you need to store along with the data that is used by your performance applications. But again aren’t there cheaper ways to store cold data?

Let’s look at the two main issues. For say 10K IOPs you need 100 disk drives while you can get the same performance from a couple of SSDs.
1) With a higher failure rate of HDDs vs SSDs, you have more failure to deal with
2) Power. You have 98 more devices that consume more power and have to be cooled and take up more space.

How does this make any sense.

elan elbracht June 10, 2017 at 3:44 am

the idea is:
how to get higher performance AND less energy consumption AND economic feasibiliy under one ombrella.
-optimization techniques within the system lead to power consumption just give me a call 😉

Paul N June 19, 2017 at 11:20 am

The offer says that “evaluated response times are system internal response times.” Your application could run 10x slower, but as long as the array’s self-reported numbers are faster, Infinidat wins. Of course remote support would never dream of adjusting the numbers as needed.

Leave a Comment

Previous post:

Next post: