3.017: vmax 2011 edition - powerful. trusted. smartest.

In the 20 months since its launch back in April of 2009, VMAX has literally redefined the storage landscape. Back then, EMC focused the messaging around how VMAX was purpose-built for the virtual data center, leveraging multi-core Intel technology to deliver a highly efficient and scalable modular and tiered enterprise storage platform. We introduced the new Virtual Matrix architecture, the first array built upon that architecture, and the first wave of automation that has simplified the whole deployment model of Symmetrix storage.

EMC also did a bit of a Babe Ruth at that launch – pointing to the bleachers where we intended to deliver, in two phases, the innovation of Fully Automated Storage Tiering. FAST v1 for VMAX began shipping just about a year ago.

On December 15th, 2010 the second phase of FAST began shipping, along with more than 50 other significant features and new products in what we now call Enginuity 5875. Included also were some new hardware updates to VMAX – a new native 10Gb Ethernet director for SRDF and iSCSI, plus a new VMAX engine that sports an encrypting back-end to support Data at Rest Encryption.

Today (January 18th, 2010), EMC publicly announces what is inarguably the largest set of new storage products ever to be simultaneously introduced on one day. With over 40 new products and scores of new features, today's launch truly lives up to its Record Breaker theme. (If by chance you've missed all the hype, there's still time to learn about it at the #EMCBreaksRecords web site.)

So, what's all the hype about? Well, for the full effect, you'll have to go see for yourself. But within the context of VMAX, there's lots of new things in this latest release of Enginuity 5875, and I thought I'd lead off my contribution to the launch day communications with a quick run through of the major ones…

 

enginuity 5875: a very long list of features

I have to warn you up front – there's a lot in this release. In fact, it can be argued that 5875 marks the single largest functionality release in Symmetrix history. In recognition of the magnitude, this release's development schedule included extended internal qualification cycles and an extended customer/partner beta program.

To make things a bit more consumable, I'll structure this as a list with a brief description of each. It's still a long list, and the details will be scarce in this first post. But it should give you a flavor for the scope of what has turned out to be the single largest Enginuity release (in terms of features) in the 21 year history of Symmetrix:

• Fully Automated Storage Tiering for Virtual Pools (FAST VP): Policy-driven sub-LUN Fully Automated Storage Tiering simplifies storage allocation and allows customers to utilize Flash & SATA capacity to attain the same performance as an all-FC configuration for a lower $/GB – and/or – to realize performance better than all-FC for costs similar to an all-disk based solution. Using algorithms derived from real traces from more than 3500 customer production workloads, Symmetrix FAST VP will set the true standard for automated tiering far above where the competitive wanna-bees have targeted with their attempts at pre-empting FAST VP.
• Federated Live Migration (FLM): Delivering on another target defined at the VMAX launch, Federated Live Migration enables the first truly non-disruptive tech refresh migrations for moving open systems hosts and applications off of existing DMX1, DMX2, DMX3 and DMX4 arrays onto VMAX. Unlike traditional virtualization solutions which require an interruption to insert the virtualization layer, FLM is a peer-to-peer solution that literally orchestrates the LIVE transfer of LUNs from old arrays to a new one requiring nothing more than a qualified multipath driver installed on the host (PowerPath 4.x and 5.x to start, DMP and MPIO to come). This approach gets customers using VMAX faster and with less operational complexity, and best of all, everything required (except the multi-path drivers) is provided at no incremental charge with the addition of FLM to the Symmetrix Migration Package.
• Data At Rest Encryption (DARE): Implemented within a new encrypting VMAX engine and based upon RSA technologies, DARE provides integrated drive-level encryption for those requiring protection from media loss. Symmetrix DARE automatically encrypts ALL data written to EVERY drive, be they Flash, Fibre or SATA. Each and every drive has its own unique key that binds it to a specific array and drive slot, and keys are never made visible in any format to users, operators, administrators nor service personnel. Implemented in hardware there is no measurable performance impact.
• Native 10Gb Ethernet: Supporting both iSCSI and native SRDF, this new hardware interface controller supports up to 8x the bandwidth of the existing VMAX 1Gb controller, allowing customers to reduce their costs when their replication and/or iSCSI workloads exceed the capabilities of 1Gb
• Simplified Storage Allocation: In our continued focus on reducing the time spent to provision storage, Enginuity 5875 cuts the provisioning times in half again over the initial 5874 release.
• Default Port Flags: A somewhat late addition to 5875, Enginuity now defaults the port flags to support the most common set of hosts as part of the auto-provisioning process. These can still be overridden, but henceforth most customers will likely never have to deal with port flags again.
• Dynamic Back End: As we proceed in our efforts to deprecate the notorious BIN file to the point of irrelevance, Enginuity 5875 makes another great stride: hardware configuration changes no longer require that array configurations be locked-down in preparation for adding new drives or other hardware.
• Virtual Storage Integrator (VSI): The premier block-storage integration add-in for vCenter since its first delivery in 2009, the VSI package has been extended as a set of Microsoft MSC plug-ins for Hyper-V deployments as well. Coupled with the storage allocation automation foundation in VMAX, VSI truly streamlines operations for virtual environments in both VMware and Hyper-V domains.
• VMware VAAI Integration: Enginuity 5875 adds native support for the Bulk Zero (Write Same), Bulk Copy and hardware-assisted locking, off-loading I/O intensive operations from the server into the storage where these operations can be executed far more efficiently.
• T10 Space Reclamation: Another set of new APIs are the T10 UNMAP and WRITE_SAME(UNMAP) SCSI commands. These allow the host file system, database and/or volume manager to notify a supporting "thin" device of blocks that are no longer in use so that the storage can release those blocks to the free pool. As I wrote in my 3.011 hot air reclamation post, the Symmetrix implementation adds a few innovative twists in how we utilize these APIs to drive additional efficiencies within the VMAX.
• NWBH SRDF & Rebuilds: And in fact, two of the related innovations are that both SRDF and RAID rebuilds are now aware of blocks that have never been written to by hosts. These blocks will be skipped over when making replicas or rebuilding a failed drive, based upon the knowledge that these blocks must be all-zeros (since that's what VMAX initializes unwritten – and UNMAPped blocks to). In fact, if a drive has any significant portion that has not yet been written to by hosts, VMAX can now rebuild it FASTER than the rated drive write speed…
• Zero Space Reclaim during Migration and Replication: And related to THAT is the new ability for SRDF and Open Replicator to scan data during a migrations for all-zero blocks. Only non-zero data is written to the target, eliminating the unnecessary I/O overhead and reclaiming the unused space during the move (if the target is a thinly-allocated VP device).
• VLUN for Virtual Pools: First introduced for DMX3&4, VLUN now supports relocating a device from one Virtual Provisioning pool to another. This is particularly useful along with FAST VP, because it allows you to define the default pool for a given device (where writes will be "landed" unless FAST VP overrides the default). It also allows storage admins to quickly "get out of sing-sing" if they suddenly need to relocate a VP device to a different tier or pool for performance or workload balancing purposes. Ideally, one should never need to use VLUN for any device, but administrators are thrilled to know it is there if they ever need it.
• Copy QoS: With the movements of FAST VP and VLUN and replication and the like has come the need to set relative priorities of the various copy operations. Under normal loads, VMAX will do a yeoman's job in moving data without any discernable impact on performance, but inevitably there comes the time that something is more important than another, and Copy QoS provides a mechanism to guide the system to focus on what is most important.
• "Lights On" NDU: As Symmetrix arrays have gotten larger and are supporting more and more servers, non-disruptive software upgrades have been both a requirement and a challenge. This latest release significantly overhauls the NDU implementation of VMAX such that the entire software upgrade process is performed without ever logging out of the Fibre Channel fabric nor even turning out the lights. This serves to avoid any "login storms" that might occur after the new code comes on-line, and also avoids errant "link down" recoveries that occasionally occur in the host multi-pathing software during NDU. The result – even LESS disruptive non-disruptive upgrades.
• Vault to Flash: For customers wanting to use 100% Enterprise Flash Drives (EFDs) in their VMAX arrays, power-failure vaulting is now supported to EFDs. Prior only FC or SATA devices were allowed, but some customers wanted their VMAXes to be totally free of all spinning rust. Who are we to argue?
• Local & Remote Replication Features and Enhancements
  • TimeFinder/Snap Duplicate Snaps
  • TimeFinder/Snap off SRDF/A R2
  • Concurrent SRDF/A – two Async R2's from the same R1 array
  • Optimized Copy Engine (request-based Copy)
• Mainframe Specific Feature and Enhancements
  • Simultaneous TimeFinder/Clones for Mainframe
  • Compatible z/OS Global Mirror Multi Reader support to improve performance
  • Global Mirror and Compatible z/OS Global Mirror Incremental Resync to increase resiliency
  • Support of zHPF multitrack for better performance with large block IO
  • Faster provisioning for CKD
  • Increased I/O Concurrency with SRDF
• Electronic License Manager: Providing customers with a means to track licensing compliance, Enginuity 5875 now includes an integrated license manager to track both licensed features and registered capacities. Customers can get daily reports, and license keys are easily applied using a variety of standard management tools (SMC, CLI, etc.).
• Up to 2.5x Large-block Cache Miss bandwidth: last, but certainly not least, Enginuity 5875 more than doubles the maximum large block read miss GB/s that VMAX engines can deliver. This added bandwidth will benefit a broad range of applications, from DSS/OLAP/BIDW to backups – not to mention reduce any impact of FAST VP data relocation.

but wait, there's still more!

Enginuity is the firmware that runs the Symmetrix hardware, and within that code base are implemented the array-based functionality. Surrounding Enginuity is a variety of host-based products, applications, toolkits and enablers, and these are simultaneously released along with Enginuity 5875. Each has its own list of features and enhancements, but for now I'll just list them for completeness' sake:

• Solutions Enabler V7.2
• Mainframe Enablers V7.2
• SMC V7.2
• SPA V2.1
• EzSM V3.1
• iSeries STM V7.2
• SMI-S V4.2
• VSS/VDS V4.4
• VSI V4.0 for VMware
• SRDF SRA V2.2
• SRDF/CE V4.1
• InfoMover V5.1
• TF/IM V4.6
• VSI for Hyper-V Qualification with 5875
• GDDR V3.2 Qualification with 5875

Whew! The Symmetrix team sure have been busy, huh!

vmax, 2011 edition

Given the significance of this release, it is almost like getting a brand-new VMAX without a hardware upgrade. Non-disruptive upgrade to a system with more performance, the world's most intelligent automated sub-LUN tiering, seamless migration from older Symmetrix arrays, and tight integration with virtual server infrastructures large and small.

There's a lot more going on at this launch than just VMAX, but perhaps this gives just a little context as to why we're proud to participate in today’s Record Breaking launch.

Symmetrix VMAX: Powerful. Trusted. And now the Smartest.
 

technorati tags: Breaking Records, Record Breaker, EMC, Symmetrix, VMAX, FAST, FAST VP, Federated Live Migration, Data at Rest Encryption, 10Gb Ethernet, 5 Million VMs, Powerful, Trusted, Smartest, SRDF, TimeFinder, VAAI, UNMAP, T10, WRITE_SAME, sub-LUN

Comments

[Philip Leighton]

As my organisation is just about to go through a DMX4-VMAX migration I was quite excited to read about FLM.
I had a look at the FLM documentation on PowerLink and it appears there is a bit of a gap between the statement "FLM is a peer-to-peer solution that literally orchestrates the LIVE transfer of LUNs from old arrays to a new one requiring nothing more than a qualified multipath driver installed on the host " and the reality of the Simple Support Matrix (300-000-166 REV A02).

Cluster volumes are not supported (SCSI3 reservation issues).

Boot volumes are not supported.

Windows 2003R2 only supported with a single specific Emulex or QLogic HBA.

Windows 2008 or Windows 2008R2 are not supported.

As this takes out most of my environment unfortunately it’s not quite the panacea I initially thought it might be. Disappointing for this customer considering its being called out as such a major new feature.

[the storage anarchist]

Many of your concerns are a simple matter that the qualifications have not completed for the combinations you list. I suggest you submit an RPQ before abandoning hope.

[Aran Hoffmann]

Phillip,
If the RPQ does not pan out, which it should, then you can still use the traditional Open Replicator with Hot Pull and Donor Update option. The only downtime required for this type of migration is the time it takes to shutdown a host, switch the masking, then boot it up again.