Extreme Blogging

ORIGINALLY POSTED 15th November 2013

37,106 views on developerworks

I see that EMC have finally pushed yet another technology family on its customers. Extreme I/O arrays. So now you have to decide which one of these things do I want.

OK, let me buy an all flash array, so how will that fit with my VMax and VNX estate – oh it wont –

  • you mean its completely different, and so how do I use this wonderful FAST stuff with it…
  • oh I can’t and, all those SSD you sold me, you mean I can’t connect it all together…
  • oh I could buy a Vplex, and sort of scale it that way… but what does the Vplex buy me… errrm
  • nothing… maybe I need one of those recoverpoint things tooo…
  • how much am I up to now… you want what?

EMC is where IBM was about 15 years ago, a bunch of products, all losely connected but all without commonality… I wonder when they will admit it and come with cap in hand and ask IBM if it can OEM SVC since it could solve all their problems as it has been solving EMC, IBM, HDS and basically everyone’s storage problems for the last 10 years.

Storagebod has been asking what the vendors have been up to, when it comes to blogging, and I too miss the good old days of nip and tuck, pull your punches when you feel you can defend them, and outright attack when someone just plain gets it wrong…

Most of the characters have disappeared, (where for art thou brother ?Barry B?!, some still plod on writing so much without saying anything at all (Chuck) and yes, I do confess Martin that I get prompted to jump out and say hello when we have the next release, or next marketing thing to say… but I don’t want to just post for the sake of it – I’ve said a lot over the years – and don’t like repeating myself. But maybe I should…

Because we in IBM Hursley developerd a differentiator for all storage users – the IBM SAN Volume Controller – we’ve even thought of renaming it the Storwize Virtualization Controller (I liked Contraption for the last C) – but anyway – as we all know it – SVC. It is embedded in all the IBM Storwize family products, which means anyone who has needs of a $10K storage box (IBM Storwize V3700), right up to the 100PB customer (IBM SVC) can benefit – and everyone in between (Storwize V5000, Storwize V7000).

But its times like these I really realise just what a beast we created… and how others, like EMC, just don’t have what we do… if only IBM were an Evangelical Marketing Corportion too…

When IBM acquired TMS (Texas Memory Systems) back in 2012, I was happy that we’d bought a company that “got it right” like we did with SVC. A single rack U footprint with 20TB capable of 500K REAL IOPs. SVC always wanted a RAID brick to use as ‘dumb’ storage, and we’d now got a VERY FAST RAID brick that wasn’t so dumb…

And that is my point… The storage should be a reliable (RAID first stood for Redundant Array of Inexpensive Disks – then became Redundant Array of Independent Disks – now it just means some Storage Object that is Fault Tolerant – maybe SOFT!) So we want some SOFT that is a Super-fast SOFT Device – SSD…) Todays SSD are all Flash based, so we need a system of those SOFT SSD, or even a FlashSystem…

That FlashSystem needs to be a building block, so it has capacity, its super-fast, has fault tolerance bulit in, and will likely need some traditonal HDD storage too. If all FlashSystem were at the $/GB of SATA drives, then we’d be laughing. But they are not. So we need a hybrid solution. We need something that can cope with both.

Yes some applications can benefit from all Flash – but do you have a single application that NEEDS a million IOPs?! I bet you have an appication that needs a million KB of storage though! Not only that but you have backup, replication, site tolerance, and efficiency needs, not only in storage usage, but capex, opex constrainst, and compliance regulations too…

What I am trying to say here, and hopefully what you see it too, is that storage has moved WAY beyond that of raw reliable capacity – so it has also moved way beyond tiers or classes – what it needs is function, flexibility and efficientcy that can be applied commonly across all the entrie storage estate.

You might like to call this “Software Defined Storage” if you are fed up toughting your “Cloud Storage”, but I like to say that its a vision. Not that of a marketeer, but a realist – if you can apply one set of common storage function, flexibility and efficiency, you can build your “cloud” from it, no matter what vendor you choose, and if its all written in software… then maybe its “software defined” too…

So while EMC may trumpet Extreme IO – how does it really fit, how would you use it in a flexible, efficient manner, do you have that million IOPs application, or would you like to (btw its only 250K iops per system they sell you) use those IOPs over your entire estate… thats what our SVC and FlashSystem users are benefiting from TODAY, and have been for over a year

Here are some simple facts for you all related to todays “information”:

IBM FlashSystemXtremIO
20 Terabytes: 1U7 Terabytes: 6U
110us latency (~5x faster)500us latency
Native eMLCeMLC SSDs
400W power750W power

Add 4U of SVC ontop (50us) of either and you get Replication (GM/MM/GMCV), Compression, Mirroring, Split-site HA, Non-distruptive migration, Easy Tier Optimisation, etc etc_

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: