I may as well have titled this post "I KNEW IT!", as it was my first reaction.

The rather excellent You Had Me At EHLO blog has a good article on Exchange 2010 losing Single Instance Storage (SIS). 

First up, and almost as an aside, I've been saying for years that the sweet spot for Exchange is around 100 users.  Maybe 250 at a push.  Its design always felt like that was the sweet spot - as if someone had wanted to build a new MS Mail server, then suddenly realised that maybe they should scale it up. 

  "Exchange 4.0 (and, to a certain extent, Exchange 5.0 and Exchange 5.5) was really designed as a departmental solution."

Couldn't agree more.  It's been crippled by design since day one, and whilst it has improved, it's still crippled by some of those early design decisions.

You should go and read that article for some very interesting thoughts from the Exchange Team on storage.

It's an article article full of nice little asides like the quote I chose above.  And like SIS not being something you should factor into capacity planning.  Which isn't news to anyone who had studied Exchange properly, but will probably be news to some of Microsoft's salespeople, PR people, and many of their MVP's and MCSEs out there in the community.

(It's also not a very well written article in a couple of places.  Apparently the Exchange ESE database schema has remained relatively the same for 19 years, but then with Exchange Server 2000....  Wait?  Someone was running Exchange Server in 1981?  Gosh.)

But what I really want to talk about is that SIS is gone, and that attachments aren't single instanced either.

FORWARDS, TO THE PAST!

Microsoft have instead given you compression.  IBM gave us design element and database compression in R8, so the compression bits they talk about aren't very interesting - we have direct comparisons.  In fact, I recently saved around 25% on four archive servers just by implementing R8's compression.  25% may not sound like much, but 25% of ~750Gb is a nice chunk of disk to have back.

Anyway, what's really interesting is the strategic directions at play here: 

Microsoft:  Disk is cheap, buy as much as possible and use Retention Policies to keep things under control.

IBM: Disk is cheap, but let's try and use it more efficiently.  Downtime for expanding later can be expensive, after all.

It's nice that Microsoft has retention policies, and IBM's policy so far has been to leave that to 3rd party products.  (You can set up automated archiving if you like, which is a crude retention policy, I suppose.) 

But I don't buy that storage is cheap.  Even with a SAN to make it cheap to manage, you have to cut the storage, present it, expand volumes - all of which have risk, and the usual way of mitigating that risk is to shut services down whilst you expand your storage.  The individual platters may be cheap, but unless you're 100 users or smaller then there will be costs associated with extending storage.  Not using a SAN?  Well, expanding that RAID array may be even more expensive in downtime and have even higher risks...

And I don't buy that this is all in the name of performance, either.  From Domino 6 to 7, and then from Domino 7 to 8, IBM have optimised their NSF structure and software engine without significant on-disk space increase.  I know that the two aren't really directly comparable, but...  Was ESE really that badly architected that it needed such drastic changes to get better performance?  And wasn't the move to 64-bit with Exchange 2007 supposed to give better performance?  Was that not enough?

So strategically, this is becoming an interesting role reversal.  Microsoft is moving Exchange Server to a place where its storage layer can be criticised for exactly what they used to criticise Lotus Domino Server for.  Meanwhile, IBM is giving you more options - you can carry on as you were, or you can have some single-instancing and compression to help you manage storage capacities. 

What makes it really bad for Microsoft is that there is now no defence.  Their best shot is just saying "storage is cheap, and this gives us better performance" - which means IBM can say "Yep, and that benefits us too, and we also have the options there if you want them - at no cost."

I've always felt suspicious and somewhat used when Microsoft do their "What we sold you last time is crap, buy this new one which is Not Crap!" dance.

And in a way, the blog entry is not the most interesting thing.  The most interesting thing is the complete lack of cognitive dissonance in some of the comments.  A was good, B was bad.  Now A is bad, B is good.  And nobody's questioning the sudden reversal.  They're just applauding, like the Emperor's new clothes are the best they've seen yet...

Anyone care to take a bet that in a year's time we'll be hearing all our old criticisms of Exchange SIS being rephrased and bounced back at DAOS?

BOOKMARK THIS CONTENT

	Bookmark at:

CREATE A NEW COMMENT

	Name:
	Email:
	Website:
	Content:	HTML is not allowed. Hyperlinks will automatically be converted.