« 2.000: storage anarchy two-dot-oh | Main | 2.002: meh – ibm really, really doesn't get flash »

May 05, 2009

2.001: ibm's amazing splash dance, part deux

A couple of month's ago, I posted a review of how the various storage vendors were embracing flash drives (or weren't, as the case may be). I then followed that up with a post lamenting the lame (and factually incorrect) white paper describing IBM's approach to enterprise flash drives.

I complained then that IBM was throwing cold water on a very key new technology; the fact that the errors in that white paper STILL haven't been corrected after nearly TWO MONTHS underscores my observation that IBM is totally out of touch with reality, and no longer the "trusted advisor" they once were.

(What happened, BarryW – I know you were working on getting those errors corrected!)

But today's news takes the cake: instead of doing it themselves (today IS IBM-Announcement-Tuesday, after all), IBM let STEC be the one to announce IBM's support for flash SSDs.

In my book, when you trivialize the importance of ANY technology to the point of having your supplier announce GA and availability rather than doing it yourself, it means something. And when IBM's sales force is to this day telling prospects that flash SSDs are "unproven technology" and "not ready for the enterprise," I can only conclude that IBM is embarrassed to admit some huge limitation or inadequacy of their products when used with Flash.

So, I asked myself…
 

…what is ibm trying to hide?

Now, I know better than to ask either BarryW or TonyP for an honest answer to a question like that.

So instead I turned to the paragon of accuracy for all things IBM – RedBooks and RedPapers.

Lo and behold, there is actually a RedPaper titled "DS8000: Introducing Solid State Drives!" Looks like exactly the resource I'd need to understand how IBM is supporting flash in their array…and possibly some insights as to why they aren't making any noise about it.

And indeed it is, even though the current version is plainly labeled "Draft Document for Review April 28, 2009 3:02pm." While there are likely to be some minor grammatical edits, I'm reasonably confident that the actual implementation and configuration details won't change.

Here's what I learned about their DS8K support for STEC SSDs from this RedPaper:

  • SSDs are available only on new DS8K systems – they cannot be added to existing systems (EFDs can be added, non-disruptively, to any CLARiiON CX4-series, Symmetrix DMX4 or V-Max storage array, new or as an upgrade)
  • The DS8K does not support RAID-6 or RAID-10 on their SSDs - only RAID- 5 is allowed at this time (All of EMC's arrays today support RAID 5, RAID 6 or RAID 1/10 on EMC EFDs)
  • DS8K SSD’s must be installed on a DA pair without any other drive types on the DA (CLARiiON, Symmetrix DMX and V-Max allow EFDs to be placed on any drive channel and to be intermixed with any other drive type)
  • A maximum of 16 SSD's can be installed on any DS8K DA pair, for a total maximum of 128 SSDs (Symmetrix DMX4 supports128 EFDs standard and more via RPQ; V-Max supports 256 EFDs and more via RPQ)
  • If you install ANY SSDs in a DS8K, your maximum configuration is limited to 2 expansion bays and your maximum number of drives (SSDs + HDDs combined) is limited to 256 (128 for the smaller DS8K) (EFDs have no impact on the total number of drives or expansion cabinets you can configure on with CLARiiON, DMX or V-Max, and if you want a V-Max with 2400 EFDs, all you have to do is ask).
  • You cannot order a DS8K with only SSDs (both Symmetrix DMX-4 and V-Max can be RPQ'd as flash-only configurations, although using HDDs for vault drives is more cost-effective).
  • Although STEC asserts that IBM is using the latest ZeusIOPS drives, IBM is only offering the 73GB and 146GB STEC drives (EMC is shipping the latest ZeusIOPS drives in 200GB and 400GB capacities for DMX4 and V-Max, affording customers a lower $/GB, higher density and lower power/footprint per usable GB.)
  • The only way to relocate data from HDD to SSD on a DS8K is via host-based copy utilities, which consumes host MIPS and adds load to both server and storage while potentially requiring application downtime to effect the relocation (Symmetrix DMX4, V-Max and CLARiiON all offer array-based non-disruptive Virtual LUN relocation; VLUN on Symmetrix supports both Open Systems and Mainframe volumes, and V-Max also allows for changing the RAID type when relocating a LUN to a different tier)
  • z/OS support of Solid State Drives in the DS8K will be available on z/OS V1.8, or later. (Both Symmetrix DMX4 and V-Max support EFDs for mainframes running any version of z/OS, z/VM or TPF. Additionally, V-Max supports EFDs for use by iSeries and AS/400 servers)

And finally, despite all the bluster and fanfare of IBM's Quicksilver demo last year, there has been no mention to date of anything truly innovative in IBM's use of flash. In particular, not a peep out of IBM (or anyone else, for that matter) that they have anything even REMOTELY similar to EMC's Fully Automated Storage Tiering (FAST) that was announced on April 14 along with the V-Max.

For while flash as a "Tier 0" is interesting and useful ("Tier 0" is an EMC term that IBM has copied, by the way), the true full power of flash will only be realized when the system can automatically and dynamically utilize flash capacity to accelerate the LUNs and sub-LUN data chunks that are used most frequently and/or that achieve the greatest benefits from significantly reduced response times.

Come to think of it, the DS8K doesn't even offer thin provisioning – with Symmetrix and CLARiiON, you can use Virtual Provisioning on EFD capacity to maximize the utilization of that expensive capacity!

the true meaning – ibm is abandoning storage

Or at least, they clearly aren't taking storage seriously.

More importantly, the aging, decrepit DS8K is no longer a viable storage platform for the enterprise. Later this year, IBM will finally get around to upgrading the aging DS8K processors to the p6, but it won't be enough to catch up with the Symmetrix V-Max.

And the XIV debacle is almost over, as customers learn first hand that everything I've said about the data corruption and performance is in fact true (thanks to all the free eval units seeded over the past 12 month). Moshe has cashed all his checks, and the shine has worn off for that "special XIV sales overlay," as they all get hit with real quotas, the end of their draws, and the ultimate insult of getting folded back into the mainstream sales force.

And while IBM has been handing cold cash to Moshe, it has been starving the DS8K. Meanwhile, Joe Tucci has upped funding on storage platforms, and the Symmetrix V-Max has totally re-defined enterprise-class storage. True scale-out architecture, built on industry standard components, and leveraging the time-proven Enginuity operating system (instead of cobbling together a loosely connected hodge-podge of Linux-based storage services), Symmetrix V-Max has defined a storage future that inevitably won't include any IBM-manufactured storage.

IMHO, at least.

This post is from the storage anarchist.

 


TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d834c659f269e201156f7a910e970c

Listed below are links to weblogs that reference 2.001: ibm's amazing splash dance, part deux:

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Bas Raayman

Hey Barry,

just a short question that came up first when I read your post. You state that "although using HDDs for vault drives is more cost-effective".

What about performance? Independant of what type of SSD I use (EFD or consumer type), I will see the biggest benefit in parallel reads when using flash storage. Can you give any info on the performance of a cache destage with the standard HDD's and when using EFD for the vault drives?

Thanks,
Bas

the storage anarchist

Bas -

Vault drives on a Symmetrix are used to retain an image of global memory in the event of a total power failure or emergency shutdown. Roughly 5GB of capacity is reserved on a subset of the drives in the system for this purpose, and data is written sequentially and in parallel to all the vault drives simultaneously such that all of global memory can be vaulted in less than 3 minutes (integral standy power supplies provide power for at least this long in the event of power loss).

The STEC EFDs indeed excell at parallel (and random) read operations, but they are also up to 10-15 times faster than HDDs for writes - especially random write workloads.

But using EFDs for vault would arguably be more expensive than necessary. Since the destage is 100% sequential writes, even a 7200 rpm SATA drive is sufficient to handle the load.

But as I said, if you're willing to sacrifice enough EFD capacity to handle the memory vaulting, your RPQ will likely be approved.

jason arneil

interesting line: "the true power of flash will only be realized when the system can automatically and dynamically utilize flash capacity to accelerate the LUNs and sub-LUN data chunks that are used most frequently "

I agree wholeheartedly.

Does a Clariion actually have the ability to do this?

I'm currently comparing features Clariion against a Sun 7400, to decide which to buy.

The Sun does exactly as you say and moves the most used data to SSD. Can a Clariion do that?

Chris M Evans

Barry

Isn't there a restriction on where SSDs can be placed in the DMX-4? I thought they needed to be in the first enclosures in the first disk bays. If that's still true, how to you (easily) deploy SSD into an existing configuration?

the storage anarchist

Jason -

The Fully Automated Storage Tiering (FAST) that was announced with V-Max will also be provided on CLARiiON and Celerra platforms.

Alas, I'm not permitted to provide details on those platforms' implementation or schedules, but your EMC sales team should be able to connect you with people who can discuss them with you.

the storage anarchist

Chris -

You're right - the limit exists today on the DMX4...I got ahead of myself. That restriction will be lifted later this quarter with an upcoming 5773 Service Release.

On V-Max, EFDs are allowed to be installed in any DAE or drive bay.

Barry Whyte

Barry,

"Announced" is very different to "released"

So give us some dates please on when all those funky DMX-5 features are actually released?

Brainy

@Jason:

I would go with the Sun Storage. EMC is famous for charging for every little feature.

Just two days ago Sun released a major software update, bringing 15 (!) new features free of charge.

Many interesting features are already on the roadmap, and they will be free too. (See slide 14 http://wikis.sun.com/download/attachments/63226450/summit09.pdf)

Funny, that the so called "leader" is not allowed to talk about FAST publicly in detail, while others are already delivering solutions and even provide the sourcecode for it for free!

While Sun is truly doing "OpenStorage", EMC is clearly doing "ClosedStorage(tm)".

(Disclosure: I'm an EMC, HP, NetApp, Sun Customer)

the storage anarchist

BarryW -

Everything I discussed in my V-Max launch blogs save for FAST has been shipping since BEFORE the V-Max public announcement.

FAST v1 (automated relocation of "fat" LUNs) ships 2H'09, FAST v2 (automated dynamic relocation at the sub-LUN granulatiry) ships in 2010.

And there's no shame in pre-announcing the capability...at least we've explained how Flash will deliver even more value going forward. And customers can start planning for their new future.

You'll also notice that all that "what you REALLY need to do with Flash is..." discussions have ceased now that we've disclosed the FAST objective.

Now we just wait for everyone else to announce they're doing the same thing. Even if someone else gets there first, at least we all know where we're going...

Unless you're going to say that FAST is a bad idea, that is...

the storage anarchist

Brainy -

Dude! Normally I might have considered deleting your comment as spam, but it's too funny for me to deprive my readers from it.

Somehow the irony of Sun's "Open and Free" business model and their outright collapse as a viable business just makes me chuckle!

G'Luck with the "charge-you-TWICE-for-every-feature" future you'll soon be living with Oracle!

Brainy

Don't spread FUD.

I know, and you know that Oracle will have a very aggressive pricing for the next years to come.

Sun's Amber Road Project is a massive success, and a proof that Sun's Open and Free business model works. They just run out of gas. With Oracle, the business model now just got a direct pipeline.

And if that's not the case and Oracle "charges-me-TWICE-for-every-feature" it's still half the price than what EMC charges me, to achieve the same functionality.

But hey, thanks for not marking me as spam. Otherwise this would have convinced me once more, that Sun's OpenStorage is the way to go!

williamwbishop

I have to get a bit snarky here...You're statement belittling IBM "was throwing cold water on a very key new technology".....

I'm grinning a bit.

The comments to this entry are closed.

anarchy cannot be moderated

about
the storage anarchist


View Barry Burke's profile on LinkedIn Digg Facebook FriendFeed LinkedIn Ning Other... Other... Other... Pandora Technorati Twitter Typepad YouTube

disclaimer

I am unabashedly an employee of EMC, but the opinions expressed here are entirely my own. I am a blogger who works at EMC, not an EMC blogger. This is my blog, and not EMC's. Content published here is not read or approved in advance by EMC and does not necessarily reflect the views and opinions of EMC.

search & follow

search blogs by many emc employees:

search this blog only:

 posts feed
      Subscribe by Email
 
 comments feed
 

 visit the anarchist @home
 
follow me on twitter follow me on twitter

TwitterCounter for @storageanarchy

recommended reads

privacy policy

This blog uses Google Ads to serve relevant ads with posts & comments. Google may use DoubleClick cookies to collect information (not including your name, address, email address, or telephone number) about your visits to this and other websites in order to provide ads about goods and services of interest to you. If you would like more information about this practice and your options for not having this information used by Google, please visit the Google Privacy Center.

All comments and trackbacks are moderated. Courteous comments always welcomed.

Email addresses are requested for validation of comment submitters only, and will not be shared or sold.

Use OpenDNS