Flash vs AllFlash

During 2015 have probably both you and I been well informed about Flash storage, and that will solve all your problems. Especially if you have been listen to EMC, IBM, HP and a bunch of smaller unique vendors.

But not only them have talking a lot of Flash Storage, if you also look at ESG, Gartner, Radar and many more analytic companies.

Just searching on the web on "Why Flash Storage" do we get over +55 million answers.

What is the different between Flash vs AllFlash that a few are calling it.

There is a YouTube video that explains very simple Flash vs Disk and what flash really does for you.

I'm trying to copycat that video but via text and pictures instead.

On top of this do you also have, what is the different between Flash vs Flash, let's say what is the different between IBM FlashSystem vs Violin? Or what is the different between EMC ExtremIO, HP 3Par AllFlash array and IBM Storwize All-Flash?

I'm not going in to the different between does two brands.

So let's start with the basic, what are we try to success with an Any-Flash system. We are trying to cut responds time between each I/O and there for generate as many IOPS as possible.

What that said, we can generate enough IOPS for you with a regular SATA disk solution if we have enough spinning disks up, but what we can't is reduce the latency between the CPU in the server and the disk system without changing disk technology.

Let's start with a basic disk solution from any vendor.

Between the CPU and the disk system do we have everything such a PCI BUS, HBA, Fiber Cable, SAN Switch and another FC cable.

This takes approx 100 ms to go though.

Then do we have the disk system, that's probably takes 5000 ms to responds.

Then it takes another 100 ms to send back the information the same route as it comes.

That mean it takes 5200 ms to get a responds on that I/O request.

If we now change that disk system with a all flash array that is based on SSD disks.

We still have the 100 ms between the CPU and the disksystem and that is something we can't do anything about as long we want to use Fiber Optical solution.

But now can we do a large cut of responds time, we can then probably go from 5000 ms to 800 ms with a All Flash system.

And then another 100 ms to send the data back to the CPU. That mean we have reduced the responds time from 5200 ms to 1000 ms. That is a really good cut down and it's probably enough for most applications and users.

But if you want to cut down even more to even less latency then do we need to insert a real flash system.

What happening then is that we still have our 100ms to send the data to the disk system, but now do we cut from 5000ms or 800ms to only 200ms.

And then back the same route for another 100ms.

That mean we have now cut down from 5200ms or 1000ms to only 400ms.

And if you want to get down even more, then should you probably look a local attach PCI BUS flash system to even cut down the transport between the CPU and the disk.

But if you want to know more, please feel free to contact us and we can help you to choose the right flash system that fit your need and budget.

YouTube Clip: https://www.youtube.com/watch?v=LV_ZKfxCRCE

Thanks

Christian Svensson