0.019: dmx-4 and oh so much more
As you probably already have seen, EMC had a pretty major platforms update announcement on Monday, with updates to virtually every storage platform product we offer.
Given the size of the company and the scope of our product portfolio, gone are the days when we have the luxury of launching only on a single product at a time -- there just aren't enough weeks in the year (and it does take weeks, with all the pre-briefings, scheduling, invitations, etc.). So we pretty much have to do these so-called "roll-up" announcements, and this one was pretty significant.
In fact, there were so many, many things announced across Symmetrix, CLARiiON, Celerra and Centera that it's near impossible to summarize them all in one place. For the next several weeks, if you want to learn about specific updates, I invite you to visit www.EMC.com directly. You can select the key product line areas from the rotating landing page banner, or you can follow the link to the Executive Summary presented by Barbara Robidoux, the Storage Platforms Marketing VP (but be forewarned, she has a LOT to summarize!).
I'll throw in a special recognition of EMC's marketing folks who created this launch - they've done a bang-up job of announcing a boatload of stuff without the overwhelming use of superlatives and hyperbole. In fact, it's obvious that there was a focused effort to communicate this announcement in terms of realizable customer benefits, instead of mind-numbing speeds-and-feeds. It certainly makes everybody's job easier when we don't have to answer questions like "so what?"
But if you really want to know the inside scoop about the launch, you've come to the right place. Well, so far as the Symmetrix side of the launch is concerned that is. My pal Storagezilla has adeptly covered the CLARiiON, Celerra, Centera, and RainFinity announcements over at his blog.
So, want to know more about DMX-4? Read on, Gunga Din!
Let me get this out of the way first. With all the attention I've been giving the competition about truth-in-advertising, I know darn well that some of you are reading this solely for the opportunity to catch me in a slip up. And while I won't deny that this article is just as much "blogketing" as any of the corporate-sponsored blogs, I have every intent to stick to the facts, to call out what's been announced vs. vision, and to be specific about availability of the features. Even though my blog isn't EMC-sponsored, the last thing I want to do is to jeopardize the revenue recognition of our products.
So, I admit up front, that virtually everything we announced related to Symmetrix DMX is targeted for availability later this quarter, in August 2007. There are a few exceptions, which I will point out as I come across them below. And while you can't get a DMX-4 installed today (unless you're a beta site), we have all the incentive in the world to ship them before the end of this quarter. Most importantly, if you want the new software functionality for your existing DMX-3, you are probably entitled to it at no charge under your current warranty and/or maintenance agreements.
So let's review the announcement...
dmx-4, as in 4Gb/s fc front and back
The major Symmetrix-related hardware component of this launch is the new DMX-4, and more specifically, its new 4Gb point-to-point FC back-end infrastructure. Complementing the 4Gb FC and FICON front-end support added to the DMX-3 at the end of 2006, the new 4Gb back-end allows the DMX-4 to support the latest in 4Gb FC disk drives.
You may have noticed that there weren't any specific performance claims attributed to the new 4Gb FC back-end. This wasn't an oversight, it is in fact intentional. The reality is that when it comes to massive-cache storage architectures, there really isn't that much of a difference between 2Gb/s transfer speeds and 4Gb/s. Transmit times are really only a tiny portion of I/O overhead, and just don't make that much difference when a massively-cached system is pre-fetching reads, buffering/delaying writes and reordering I/O requests to minimize seek times. Not that 4Gb/s won't help some applications, but most people just won't see any noticeable difference. No - the performance improvements we noted in the announcement were 100% related to the new release of Enginuity, which I'll cover later.
The real benefit of the new 4Gb FC back-end is in fact the new point-to-point architecture. In addition to supporting both 4Gb/s and 2Gb/s signaling, the new hardware interfaces the drive enclosures themselves implement point-to-point "switching" of I/O requests to the specific target device within each enclosure. In effect, this shortens the length of the arbitrated loop infrastructure and reduces the number of devices that each I/O traverses. And while this affords some theoretical performance efficiency, more significantly it improves the fault isolation and recovery capabilities of the back end. With this new hardware, individual drives can be isolated and even re-initialized without impacting any other drive on the same FC loop pair - thus improving the resiliency and availability of the system overall.
One thing we did focus on in the announcement was not to use the terminology "4Gb end-to-end" -- this in deference to the fact that the middle of the DMX-4 is clearly much, much faster than 4Gb/s. We support lots of 4Gb/s front-end connections, and up to 2400 4Gb/s disk drives on the back-end, and the Direct Matrix in the middle provides the bandwidth necessary to keep it all running. So that's why there was none of that "end-to-end" talk from us.
Correction (July 23, 2007): It turns out that the DMX-4 announcement did in fact specifically include the phrase "4Gb/s end-to-end" in several places. Oh well...the fact remains, the DMX-3 and DMX-4 architectures provide far more than 4Gigabits per second of data bandwidth through the Direct Matrix that is at their core - from 32-128 GigaBYTES/s in fact, dependent upon configuration.
One thing perhaps not clear in the launch were the details around the models and configurations of the DMX-4 that will begin shipping later in Q2'07. Basically, the core lineup of the DMX-3 remains intact - there will continue to be both DMX-3 and DMX-4 1500, 2500, 3500 and 4500 configurations in the price book. The DMX-4 950 configuration will also be available this quarter, and new for the DMX-4 entry-level configuration is native FICON support, making it a great platform for smaller mainframe environments, now that the DS6000 has apparently all but disappeared from the planet.
Customers will choose their generation based on their own qualification schedules combined with pricing, availability and configuration requirements.
Oh, and yes, it's true - the DMX-4 is not the first high-end storage array to ship a 4Gb/s FC back-end. The USP-V, announced way back in May, has that honor (but only if it meets the promised first shipments in July 2007). DMX-4 will be in August '07, so I guess that leaves the DS8000 a distant 3rd.
(Actually, it's SATA II, as in Serial ATA version II).
That's right - the Symmetrix DMX-4 is the first enterprise-class storage array to offer support for SATA-II drives within the array. You may recall that Symmetrix DMX was also the first high-end array to support the 7200rpm 500GB LC-FC (low-cost fibre channel, aka FATA) drives as well. Combined, these drives make up what I call the "fat and slow" tier of storage, although these terms are meaningful only in relation to the smaller/faster 10K and 15K rpm drives more traditionally used in high-end storage (the current suite of SATA-II drives we are supporting are all 7200rpm).
EMC's strategy to support for high-capacity, lower performance storage within Symmetrix DMX is one of the key distinguishing factors between DMX and its competition. Specifically, the ability to provide multiple tiers of storage within a single array helps to lower the TCO of tiered storage for our customers, and minimizes the complexity of managing multiple different devices to accomplish tiering. With tiering in-the-box, you can avoid both the added capital and operational expenses (power, cooling, interconnect, floor space, etc.) of separate arrays for your lower-cost ATA storage requirements.
SATA-II support is particularly important to this strategy, because the market demand for SATA storage is driving both capacity and cost improvements faster in SATA than in any other hard drive form factor and interface. So to get out in front of the $/GB curve, SATA is the place to be. By tapping directly into the high-volume SATA market, DMX-4 customers will benefit from continued capacity increases and the corresponding $/GB erosion, further reducing the CAPEX and OPEX for this class of storage.
While the launch announced that the new 750GB SATA II drives would be supported by all of EMC's storage platforms, we did note that the DMX-4 support for these 750GB drives was targeted for Q4. When the DMX-4 first ships in August 2007 it will support 500GB SATA-II drives as its largest drive.
the how behind sata support
You may be interested to know how SATA II support is being added to the DMX-4's 4Gb/s point-to-point Fibre Channel back-end. The DMX-4 approach is different from prior generations of SATA support on EMC's CLARiiON platforms, where SATA-specific DAE's had been employed. For the DMX-4 we are utilizing a SATA-II / FC "bridge" on the back of each SATA II drive. This bridge serves a several purposes:
- Interface translation (from 4Gb FC/SCSI to SATA II)
- Dual port FC interface for load balancing and loop fault resilience(SATA II drives are single ported)
- Advanced command support (things like command queuing and reordering)
- SATA-to-FC SMART translations (for error monitoring and reporting)
The net of this is that qualified SATA II drives can be installed in a DMX-4 right alongside FC drives. In fact, for the most part, the Enginuity software doesn't have to do much special to support a SATA II drive - most (but not all) of the differences between FC and SATA are handled entirely by the hardware.
Finally, one of the most compelling aspects of SATA II support is the implications for environmentals - with the larger drives, not only is the $/GB reduced, but so also are the Watts/GB, ft^3/GB, and BTU/GB. Meaning that using large SATA II drives for your storage (where appropriate) will actually save you money as compared to smaller/faster FC drives.
Hitachi executives are on record as saying they don't believe slow/fat SATA or FATA storage belongs in their high-end storage products:
Intermixing, slower, less reliable SATA or FATA disks in tier 1 storage systems will impact that system’s performance and availability with more frequent drive failures and longer rebuild times which consumes internal bandwidth, cache, and controller cycles. For these reason’s we do not offer FATA disks in our USP/NSC, preferring to attach them through external disk systems.
Oddly, Hu buries this statement within an article ostensibly complaining about 20 year old storage architectures (he can't be talking about Symmetrix, since Symm is only 16-going-on-17 years young). Anyway, despite the negative insinuations, Symmetrix has managed to include support for slow/fast drives in across all of the DMX's since the first one (introduced in February 2003). Many of the performance heuristics I described in my earlier blog enable Symmetrix to cost-effectively support multiple storage tiers within the DMX - all without the issues that Hu calls out. And without the cost, complexity and overhead of adding external storage into the mix (overheads that David Merrill had discussed in a now-missing entry on his blog, as was covered by Beth Pariseau over on StorageSoup).
So if Hitachi says that they can't support SATA inside the USP/USP-V because they can't manage the performance implications of slow fat drives, who am I to argue? That just leaves DMX-4 as the only viable in-the-box tiered storage alternative. I think EMC is OK with that.
another round of performance - for everybody!
Alongside the DMX-4 hardware announcement EMC announced an updated version of Enginuity 5772 - a "plus" version of the code release that was GA'd back in March 2007 for the DMX-3 platform. Importantly, this update is not only for the new DMX-4 platform - in fact, the same exact code release will run on both DMX-3 and DMX-4. True investment protection for DMX-3 customers.
The primary enhancements to this 5772 "plus" update are to squeeze more performance out of the hardware, and the engineers have made so many improvements that it seems there's something in there for everyone. Some of the key enhancements include:
- Sequential and large-block read response-time improvements and overall system bandwidth increases by up to 30%, thanks to more efficient memory utilization and optimized internal I/O transfers
- Up to 25% faster RAID5/RAID6 performance by dynamically redistributing the XOR calculations between the DA CPU and disk drive processor, dependent upon workload and CPU utilization
- Up to 10x faster TimeFinder/Clone create & terminate operations, bringing Clone operations ever closer to the operational speed of TimeFinder/Mirror
- Increase of space-saving Copy-on-Write (COW) snapshots per source device from 16 to 128, enabling the creation of hourly COW recovery points for 5 days or more
- SRDF/Sync protocol optimizations that reduce response times for
8K/16K Journal 0512-16K byte block transfers by up to 33%, or that enable the use of Synchronous replication over greater distances with these block sizes (as much as 2x further maintaining the same response time)
The net of all these is that existing DMX-3 customers will get more out of the hardware they already own, be it more headroom to add more applications (or tiers), better application response times, faster backups, more recovery points or new alternatives for implementing BC/DR strategies. All available at no additional charge to DMX-3 customers under warranty or maintenance - a pretty compelling customer benefit.
And in addition to these announced enhancements, there are several others that are still being finalized but that could not be fully quantified in time for the announcement. You can expect to hear more about these when the software goes GA later this quarter. I do know there is some pretty exciting news in the works for FICON performance, so stay tuned!
ease of use
Another major area of focus for Symmetrix DMX is ease of use improvements, and a key component of this is the Symmetrix Management Console. Introduced a little over a year or so ago, SMC is rapidly becoming the GUI front-end of choice for managing and operating a Symmetrix (any Symmetrix in fact, back to the Symmetrix 8000 series). Essentially a lightweight GUI alternative to the SymCLI commands, SMC also includes numerous wizards to simplify and automate device management, storage allocation and even replication relationships - EMC World, we even demonstrated the ability to create and allocate 1TB of DMX storage in less than 3 minutes using SMC. The latest version, SMC 6.0, integrates closely with the recently announced ControlCenter 6.0, and owners of Symmetrix Manager under current maintenance agreements can get SMC at no additional charge.
Several other components of the announcement fall under this Ease of Use umbrella in my mind, although they may not have been presented as such in the external launch materials. For example:
- Secure audit lag integration with RSA enVision - automated collection of the Symmetrix DMX internal audit logs into the external enVision repository, where events from virtually all of the IT infrastructure can be readily correlated, searched and reported upon for compliance and monitoring.
- Additional automation in SMC and SymCLI to simplify the new Virtual LUN capability first introduced in 5772 back in March. These new enhancements make it easier than ever to relocate RAID groups non-disruptively across different tiers of storage within a DMX-3 or DMX-4
- Dynamic Cache Partitioning and Priority Controls (delivered in the March release of 5772) allow customers to insulate the storage for different applications (or tiers) from each other within the array
- Virtual Device Addressing (also March release of 5772), which simplifies LUN mapping and addressing by eliminating the need to calculate LUN offsets across multiple hosts sharing an FA port
Environmental efficiency (aka The Green Thing) has become an underpinning of virtually every storage and server technology announcement of late. I won't take a lot of space here to rehash what was announced, but I will make a few clarifying points:
- The DMX-3 requires less Watts/GB in an apples-to-apples comparison of capacity and ports against both the USP and the DS8000, using the same exact disk drives
- The DMX-4 requires virtually identical power as the DMX-3, even with the new 4Gb/s FC back-end
- The performance enhancements in Enginuity 5772 enable both DMX-3 and DMX-4 to do more I/Os with no additional hardware
- The DMX-3/DMX-4 power advantages are even greater using the slow/fat LC-FC/FATA and/or SATA-II drives, the largest drive on the USP/USP-V is 300GB and on the DS8000 it is 500GB
- Both the DMX-3 and DMX-4 support more than twice as many drives as a single USP, USP-V or DS8000, so the Symmetrix power advantage is even larger when you need more drives than fit in a single competitors' array
- All of the above power comparisons are based upon actual measurements and not on spec sheets or data sheets
Finally, many of you noticed that Monday's announcement included a specific mention that thin provisioning support for Symmetrix DMX-3 and DMX-4 is targeted for availability in Q1'2008. This announcement apparently surprised some of you, perhaps because of my recent focus on the risks and realities of thin provisioning technology. But those of you who attended EMC World already knew - there EMC actually stated that thin provisioning would be supported across all of our storage platforms. Celerra support has been shipping since January 2006, Symmetrix DMX-3/4 support is now targeted for Q1'08, while the CLARiiON support data has not yet been announced (thin provisioning doesn't really apply to Centera).
Given the interest I've generated around this topic across the industry, I have checked, and I've been told it's OK to share a few additional details about the Symmetrix DMX implementation plans for thin provisioning:
- Symmetrix thin provisioning will be delivered as a software feature within a new version of the Enginuity microcode. This version of Enginuity will be available for all DMX-3 and DMX-4 platforms under warranty or maintenance; customers will not be required to purchase a new storage array to get this new functionality
- Target availability is Q1'2008, as a separately priced product (pricing TBD)
- The implementation will support array-based local (TimeFinder), remote (SRDF) and third party (Open Replicator) replication from the beginning, using standard syntax and commands
- The implementation will support all of the different storage media available within the DMX - both Fibre Channel and SATA-II disk drives
- Management will be provided via SMC, SymCLI and the related APIs. All monitoring and reporting capabilities of ControlCenter and related products will also work with Symmetrix Thin devices.
This is intentionally NOT a detailed product definition, and I can't yet discuss the specifics of the implementation or what specifically is being done to help customers deal with the inherent catch-22's of thin provisioning that I've discussed earlier. But I can say that we (EMC) are taking the time to deliver to our customers a comprehensive and robust solution from the start. Leveraging our experience with Celerra, and the feedback we're getting about competitive products, we think this is more prudent than trying to rush to market with a incomplete implementation.
And that's pretty much all I can say on the subject (for now)...
...and in summary
Zzzz... If you've made it to here, congratulations. I hope this has been informative, and if you have any questions, please don't hesitate to ask. EMC employees have access to comprehensive announcement information via internal PowerLink and the EMC Proven Professionals program; existing customers and partners have a customized PowerLink page as well. And if you don't fall into one of these communities, and EMC.com doesn't answer your questions, please contact your EMC representative for more information. I'm sure they'll be happy to help...
And I'm sure that at least some of you won't hesitate to let me know if you think
my blogketing crossed over that fine line into misinformation