« 1.018: fun with headlines | Main | 1.020: how much does a "free" xiv array really cost? »

August 12, 2008

1.019: (halfway) around the world in (a hundred and) 80 days

I sure hope that nobody has a pea shooter!Well, it seems that the IBM branded version of the XIV Storage System has arrived - in EMEA, at least:

IBM typically does EMEA announcements first - I'm sure these will appear in the US sites later this afternoon. In fact, TonyP is probably getting his blog post approved by IBM Legal as we speak Pinochio.

Seriously, it will be interesting to see how the world responds to the pertinent facts about the XIV product, now that it is no longer being shielded from public scrutiny by the cloak of IBM Confidential.

UPDATED Aug 12, 2008 - 6:00PM EDT

Weird. Nothing from IBM other than the announcement letters posted to IBM.com world-wide. Not even a press release. No customer stories. No insightful spotlight from Tony Pearson. No claims of coming SVC world domination from BarryW.

Just crickets.

In fact the only mention of the announcement anywhere else but here and IBM.com is Chris Mellor's coverage of my scoop over on Blocks & Files.

Maybe this whole XIV thing wasn't really all that big a deal after all!

I think I know why...

what's up with xiv?

Here are a few of the more salient points from IBM's announcement materials to help the industry press and analysts get started in their coverage:

  • With the XIV Storage System, you can use any type of RAID protection you'd like, so long as you like RAID 1.
  • For every 180TB of physical storage you buy in your XIV Storage System, you get to use only 80TB. Not quite sure who buys 100TB more storage than they need in today's economy - even if it IS slow SATA-II storage.
  • You can't even install an XIV Storage System smaller than 180TB. Or bigger (yet, at least). One size fits all.
  • Even if you choose Capacity on Demand, you'll still be spinning all 180 disk drives 24x7 - definitely not "Green Storage," especially by IBM's standards.
  • The XIV Storage System s' cache memory is apparently run without any protection from bit errors or DRAM failures, making it the only storage system I can think of that doesn't at least mirror cached writes. <corrected Sep 23, 2008>
    mirrors write data over a measly 1Gb Ethernet interconnect to a second node
  • You can safely replicate your data within synchronous distances only: limited by the speed of light to a reasonable maximum of circa 200 kilometers between your two sites, you'll not be able to protect your data from anything but campus or metro-centered disasters.
  • IBM intends to deliver several additional features for the XIV Storage System at some point in the future, including <quote> best practice configuration guidance, change management, asset awareness, capacity utilization, performance trending, operational reporting capabilities, single sign-on capabilities, and to add support for XIV Storage System as a disk system managed by IBM System Storage SAN Volume Controller <end quote>. Sounds just a little like the product isn't quite done yet.
  • There is no asynchronous replication being offered for the XIV Storage System. Wasn't even mentioned in the Statement of Direction.
  • They seemed to have forgotten to mention any support for non-disruptive tech refreshes and hardware or software upgrades as well. You'd think you'd want to highlight that if you had it, huh?
  • No benchmark results have been posted yet, but then, IBM hasn't published any performance claims for the XIV Storage System that I can see either.

I'm sure there'll be lots of fun in the next few days as the emperor struts his new clothes. And of course we'll all be looking forward to seeing TonyP's spin over on Inside System Storage.

Meanwhile, if you'd like to learn more about the realities of XIV, my Stranger Danger post of a few weeks back remains accurate without modification.

 


TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d834c659f269e200e553e0b5748833

Listed below are links to weblogs that reference 1.019: (halfway) around the world in (a hundred and) 80 days:

Comments

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

Ed

Moshe doesn't strike me as a person who lacks engineering talent. That being said and reading through the limitations that you spoke about, that would lead me to believe either a system was developed without understanding the market or the market XIV is intended for isn't the one you're speaking about.

Maybe it's not mature yet but I guess I'm trying to understand what if any need the product fills. Or is it an attempt at putting new spin on a different solution that really doesn't solve a whole lot.

the storage anarchist

Excellent observations, Ed.

Regarding Moshe's talents, my prior observations only noted the distinctive similarities of the brand-new XIV Storage System with the pre-DMX Symmetrix that he left behind way back in 2002.

As to purpose and fit, IBM's positioning for the XIV SS is spelled out in the referenced announcement pages, so I'll leave it to readers to decide if the product they've stealth-announced actually matches the needs of their defined target markets.

Were I to pass judgement, I'd surely be tried and hung for slinging FUD 8+)

K. T. Stevenson

Barry,

I can't help but notice that Hulk/Maui isn't even past the vaporware stage yet. Pot calling the kettle black perhaps?

Chuck Hollis

IBM's actions are just plain weird. I can't come up with a logical explanation for what they're doing.

I'm waiting for the "splaining" part ...

the storage anarchist

Hulk and Maui have been shipping since Q2'08.

But now that you mention it, other than discussions about them at EMC World, their GA and FCS was pretty low-key. Word back then was to expect more fanfare at NAB, IIRC.

Maybe there's a showdown coming...until then it seems BOTH sides are playing hide and seek :)

stewey

we have a gen 1 frame. it's great! extremely fast. very, very fast (counter to your insinuation about sata II drives).

however, no reporting, lack of SRM support (even from Tivoli), and inability to perform NDU's is a problem. sync-only repl is not that big a deal for us, but anxious to get async.

btw, infinniflex (er maui/hulk) apparently is not ready for prime time either. this, according to folks from e=mc2...and it's shipping?!?! i'd love to get an eval unit in. :) need to stay anon, sorry

the storage anarchist

I'm not sure why all the hulk/maui/infiniflex references are popping up here. IBM clearly has positioned XIV as "Tier 1.5" for enterprise storage for the past several months. In fact, the Product Positioning section of their announcement materials clearly targets the standard "mid-range and high-end" storage markets, not the bulk storage targets of hulk & maui.

Product Positioning

The IBM XIV Storage System (2810-A14) provides the hardware platform required for the IBM XIV Storage System Software. The combination of hardware and software enables a revolutionary grid-based architecture designed to provide an exceptionally easy to use, high performance, scalable, reliable enterprise disk system for UNIX, Linux, Windows, and other supported distributed open server platforms. It can provide a platform to address the need for reducing complexity while

    keeping pace with midrange to high-end disk capacity demands.
This system is a great addition to the IBM disk storage family, core products in the IBM Information Infrastructure. It is a good fit for clients who want to be able to grow capacity without managing multiple tiers of storage to
    increase performance and reduce cost.
These users also may want to improve their backup capabilities, as well as reduce the task load on storage administrators. The XIV system is especially well suited as a consolidated utility storage for fast growing, dynamic mixed, and emerging workloads.

(emphasis mine-tsa)

The cost thing gets me - 180 1TB drives to get 80 TB usable doesn't make sense, when you can get the same usable with 120 drives or less using RAID 5 (including spares). SATA drives are cheap, but not so much you can afford to buy nearly twice as many as you really need!

That said, I'd be most interested to understand what you mean by "extremely fast. very, very fast", since IBM hasn't published any performance metrics that I can find. I do know that most customers are pleasantly surprised by the performance and reliability of SATA drives in both the Symmetrix DMX and CLARiiON platforms - could you quantify the performance you've been experiencing - even as compared to another system? I'm curious as to the workloads you've been using as well...

And bummer about not being able to upgrade from the pre-IBM XIVv1 to the new hardware XIVv2. Is IBM at least supporting the new v10.0.0 software on the old boxes? I couldn't tell from the announcement materials.

Chris Mellor

Ahem. On a pedantic point of order Mr. Chairman, I believe I scooped everybody with fulsome XIV configuration details in my August 2nd XIV story. (http://www.blocksandfiles.com/article/6284) Now if only I can get the goods on InfiniFlex and Maui.....
Chris

the storage anarchist

Actually, sir, if you check the date on this post of mine, you'll note that I actually scooped you by a month - July 3, 2008.

In fact, the IBM materials on XIV have been floating around for a couple of months now. I personally was surprised that you published so much detail, however, since almost every document I saw with that level of detail prior to your post was marked "IBM Confidential." I tried to limit myself back in July to only those things I'd learned from non-confidential sources.

And you can't argue that I got the links to IBM's official documents up before anyone else. In fact, come on - admit it - you learned IBM had published these from my blog first, didn't you?

But I do have to admit that I owe yesterday's scoop to another EMC employee (TH) who alerted me to IBM's postings in the first place. I don't know how they did it, but they seemed to have found the documents within minutes of them going up on IBM.com. Maybe it was an American Indian scout trick or something.

Martin G

About a month a go I sat in a customer presentation given by IBM about XIV and I got the feeling that IBM do not yet feel that this product is really ready for the large corporates out there. For instance 80TB useable is just not enough for me. But it shows a certain amount of promise if they can get a Gen-2 box out next year.

For example the RAID-1 isn't really RAID-1 as far as I can see; it works at the sub-disk level and distributes the chunks across all the spindles. This means that in a failure scenario, a much large number of disks take part in the rebuild. I don't know about you guys but rebuild times and potential data-loss when we go to 2 Terabyte drives worries me somewhat. So forget about RAID-5 and RAID-1, we are probably going to end up with even more than two copies of each block written. Still, if you have enough grunt perhaps alot of this can be deduped in flight (no not the mirror copies!)

At the moment, there are some bizarre design decisions in the XIV array, why build the interconnects around 1 GbE? Why not Infiniband? Why not 10 GbE? If you build from commodity hardware running Linux, why no SOD on getting a clustered-filesystem and turning it into a NAS device as well?

But it's probably too soon for either side to start crowing yet. I'm hoping to see some interesting products over the next few years as the technologies evolve be it EFD, dedupe, FCoE and how parity-based RAID evolves.

Just some thoughts from user-land!

the storage anarchist

You're right, XIV's "RAID-X" isn't really RAID-1. Properly it is "mirrored chunks", where the XIV chunks are 1MB in size - this is an approach employed by EMC Centera back in 2002 (albeit with smaller chunks - 256KB IIRC).

But the net effect on usable capacity is the same as RAID-1 - it takes 2x as much space as the data (plus sufficient "reserved" space for rebuilds in the various failure modes).

And also you're right - total elapsed rebuild time can be reduced with this "mirrored chunks" approach. B availability is necessarily better than RAID 5 (and nowhere near as good as RAID 6), because the probability of data loss due to a second drive failure across the dependent set of drives during a rebuild is actually higher with RAID-X than it is for RAID-6 and for some N+1 RAID-5 configurations (this because the dependent set of drives is significantly smaller with RAID-5 than it is with RAID-X).

And as I said, Centera started out with a similar "mirrored chunks" approach, but early customers pushed back (hard) on the resultant $/GB - even though Centera too is based on SATA drives. Today Centera employs a "RAIDed chunks" approach that offers nearly the same availability with significantly lower cost per usable GB.

It'll be interesting to see if IBM copies this aspect of Centera as well in a future XIV update.

william Bishop

I don't know, I'm a huge EMC fan(own a dmx and several hundred TB of centera as well as some celera), but from some of my brothers out there, this thing is FAST. To me, like you, it feels like a centera on steroids(but centera is too expensive for the speed and usable space)...but it's unanimous among those who have been able to use it, it works well. As to disk waste, there are about a hundred stories out now about waste amongst the vendors, netapp, emc, etc....I don't really know how to look at this, By the time I lose enough money on power consumption vs. purchase price, I've probably already sent the box off for replacement for another array after a few years(power isn't that expensive for me) and the box(I understand) is quite a bit cheaper than a comparable array from our favourite company....Too many subjectives on this. We're going to have to wait until it's been out a while before anything can firmly be said.

Bhalachandra Banavalikar (Bhal)

Barry,

I came across your blog while searching for answers to XIV system architecture. I work in Infiniband area and I wondered from day 1 of the XIV acquisition news about the choice of 1GB Ethernet for interconnect. I strongly believe that Infiniband would be a perfect replacement for 1GB Ethernet links in backboane and will provide great value addition in terms of performance and reliability of the end-to-end XIV solution. Do you have any thoughts? Do you have more details on XIV architecture?

Thanks,
Bhal

the storage anarchist

Bahl -

I agree that the choice of a raltively slow 1Gb Ethernet is an odd choice, except that it is probably the cheapest interconnect in terms of $/GB/s.

Inifiniband might have been a better choice, but only if the costs come down rather dramatically.

IMHO, that is...

Wendy Darling

I have heard there are about 20 customers that have experienced data corruption and data loss on XIV!! Anyone know more???

Grizzly

We test drove four storage solutions in 2008. Isilon, XIV Gen 1, Sun, and XIOTech.

Isilon has a way cool NSA solution for growing storage easily and making it look like one big file server. Support for our mid-size company has been disappointing.

XIV was very easy to administer. Support was 5 stars, super excellent. However, we could not get the system to scale very well when you started adding servers of different workloads. We tested up to 6 servers. Performance basically came out at about RAID5 for reads and RAID1 for writes. IBM added more RAM and more management nodes in Gen 2 to help boost performance but at additional $/TB. We never tested Gen 2. In the end the product just wasn't feature-rich enough yet to justify the $/TB.

We eventually purchased XIOTech. They had one of the best SPC AND a 5 year warranty on the hardware (self-healing technology). Our in-house tests verified the published performance numbers. Support has been extremely good for a small company.

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