Symmetrix V-Max Systems: SRDF Enhancements and Performance

September 10th, 2009 No comments

v-max image 2So this was one of those posts that I always wanted to write related to Symmetrix V-Max and SRDF enhancements that were incorporated with the 5874 microcode.

Yesterday morning had a chat with a friend and ended up talking about SRDF and then later in the day had another interesting conference call on SRDF with a potential customer. So I really thought, today was the day I should go ahead and finish this post.

Back in April 2009 when the V-Max systems were initially launched, Storagezilla had a post on V-Max and SRDF features, he covers quite a bit of ground related to the Groups and the SRDF/EDP (Extended Distance Protection).

Here are the highlights of SRDF for V-Max Systems

SRDF Groups:

  1. 250 SRDF Groups with Symmetrix V-Max (5874) Systems. In the prior generation Symmetrix DMX-4 (5773), it had support for 128 groups. Logically even with 2PB of storage, very seldom do customers hit that mark of 250 groups.
  2. 64 SRDF groups per FC / GigE channel. In the previous generation Symmetrix DMX-4 (5773), there was support for 32 groups per channel.

SRDF Consistency support with 2 mirrors:

  1. Each leg is placed in a separate consistency group so it can be changed separately without affecting the other.

Active SRDF Sessions and addition/removal of devices:

  1. Now customers can add or remove devices from a group without invaliding the entire group, upon the device becoming fully synced it should be added to the consistency group (with previous generation Symmetrix DMX-4, one device add or remove would cause the entire group to invalidate requiring the customers to run full establish again).

SRDF Invalid Tracks:

  1. The “long tail” – last few tracks search has been vastly improved. The search procedure and methods for the “long tail’ has been completely redesigned. It is a known fact with SRDF, that the last invalid tracks take a lot of time to sync as its going through the cache search.
  2. The SRDF establish operations speed is at least improved by 10X; see the numbers below in the performance data.

Timefinder/Clone & SRDF restores:

  1. Customers can now restore Clones to R2 and R2’s to R1’s simultaneously, initially with the DMX-4’s this was a 3-step process.

SRDF /EDP (Extended Distance Protection):

  1. 3-way SRDF for long distance with secondary site as a pass through site using Cascaded SRDF.
  2. For Primary to Secondary sites customers can use SRDF/S, for Secondary to Tertiary sites customer can use SRDF/A
  3. Diskless R21 pass-through device, where the data does not get stored on the drives or consume disk. R21 is really in cache so the host is not able to access it. Needs more cache based on the amount of data transferred.
  4. R1 — S –> R21 — A –> R2 (Production site > Pass-thru Site > Out-of-region Site)
  5. Primary (R1) sites can have DMX-3 or DMX-4 or V-Max systems, Tertiary (R2) sites can have DMX-3 or DMX-4 or V-Max systems, while the Secondary (R21) sites needs to have a V-Max system.

R22 – Dual Secondary Devices:

  1. R22 devices can act as target devices for 2 x R1 devices
  2. One Source device can perform Read write on R22 devices
  3. RTO improved with primary site going down

Other Enhancements:

  1. Dynamic Cache Partitioning enhancements
  2. QoS for SRDF/S
  3. Concurrent writes
  4. Linear Scaling of I/O
  5. Response times equivalent across groups
  6. Virtual Provisioning supported with SRDF
  7. SRDF supports linking Virtual Provisioned device to another Virtual Provisioned device.
  8. Much more faster dynamic SRDF operations
  9. Much more faster failover and failback operations
  10. Much more faster SRDF sync’s

Some very limited V-Max Performance Stats related to SRDF:

  1. 36% improved FC performance
  2. FC I/O per channel up to 5000 IOPS
  3. GigE I/O per channel up to 4000 IOPS
  4. 260 MB/sec RA channel I/O rate, with DMX-4 it was 190 MB/seconds
  5. 90 MB/sec GigE channel I/O rate, with DMX-4 it was almost the same
  6. 36% improvement on SRDF Copy over FC
  7. New SRDF pairs can be created in 7 secs compared to 55 secs with previous generations
  8. Incremental establishes after splits happen in 3 seconds compared to 6 secs with previous generations
  9. Full SRDF establishes happen in 4 seconds compared to 55 seconds with previous generations
  10. Failback SRDF happen in 19 seconds compared to 47 seconds with previous generations

To read more about V-Max systems follow

To read more about SRDF systems follow

The Blue lights on EMC Symmetrix V-Max Systems

September 10th, 2009 6 comments

If you were to walk in a Datacenter and see an EMC Symmetrix V-Max for the first time, you will end up giving it a look.

It’s those Blue Flashy lights in the front of the unit that just catches your eyes.

It gives the Symmetrix V-Max the Sleek and Sexy look..

Here are some pictures to prove that..


v-max image 2

v-max image

Visible in the picture below are Cisco UCS blades, NetApp systems, HP systems, Cisco Switches, Xsigo Systems, but you can surely spot the Symmetrix V-Max

vmworld 2009

EMC Symmetrix V-Max

A video from EMC World 2009

So these lights are USB powered through USB cables, a very simple idea though a genius one. Enterprise class arrays and the use of USB ports at the backend of these systems keep these Blue flashy lights on.

Here are the pictures of those USB connectors from the front. Believe it or not, they are redundant as well.

(Look at the USB cable connectors that go on the front door, 2 of them right above the blue light)


As Storage Anarchist says in his Blog post “The first thing you’ll probably notice about the new Symmetrix V-Max is the packaging – and specifically the glossy-black panel with the blazing blue LED light bar that underscores the name Symmetrix on every door. The design team had a lot of fun blending the modern gloss-black look of today’s popular personal communications devices with the image of stability and security that customers expect from Symmetrix.”

Yea that is right,,,,,,,,,, this post was about the Blue lights on the Symmetrix V-Max Systems :-)

Oh…the big question, will it call home through the EMC ESRS Gateway if one of these Blue lights fail?

CDP: Blurring the Line Between High Availability and Backup

September 8th, 2009 2 comments

Josef Pfeiffer

For as long as people have been protecting data there have been a myriad of products to help. High availability and backup are two general categories of products that can assist but they offer very different benefits.  On one end of the spectrum, high availability includes technologies like clustering, replication and shared file systems and they really allow for near zero recovery time when a problem occurs.  If a clustered server fails, then it automatically fails over to another server and helps to ensure the application stays up and running.  What high availability lacks is the ability to roll back to older points in time.  For this reason, high availability is almost always complemented with backup products that make additional copies of the data at specific moments in time. Together they help create recovery point objectives and recovery time objectives that can be tailored to the importance of the data.

Recently however, continuous data protection (CDP) has started to blur the lines between these two separate product categories. CDP is often correctly viewed as a different way to protect data.  It tracks all changes to a disk continuously block-by-block, as opposed to scheduled points in time. CDP’s key differentiator is how it changes recovery.  Backup products always store data in a different location, whether on tape or disk (deduplicated), it has to go through some process to get copied back to its original location.  You simply can’t run your server or application off of that backup storage.  

CDP changes this however by virtualizing the backup storage and presenting read/write volumes that can be used.  The mere fact that you no longer have to copy data back to another location means that your recovery time is dramatically reduced to near zero. Sound familiar?  Yep, just like high availability.  Can’t find replacement storage in production when a problem occurs?  No problem, just run the application off a CDP server until a more permanent recovery option is available and you can fail the data back once things are fixed.  The benefit is near zero downtime. If the corruption is copied to CDP, you can simply rewind to a previous moment in time and present a virtualized disk volume of how that original volume looked at any point in time.

While there usually is a trade off between recovery points and recovery time, CDP gets pretty close to reaching near zero on both.  While not every application needs high availability or CDP, it is becoming an easy option to add to your existing data protection environment.

Vmworld 2009: Day 4, 3rd September 2009

September 4th, 2009 No comments

Unfortunately my Vmworld trip came to an end today. Started to head back home at 6 AM this morning, entire day of travelling.

Vmworld 2009 officially comes to an end today at 4 PM. Most of the people will be leaving later in the day or early tomorrow morning to head back home.

Some people are planning to spend some personal time in San Francisco Bay area with their families and friends since it’s a day off on Monday (long weekend). San Francisco is just amazing, the Golden Gate bridge, the food, the wine, the downtown and oh the Moscone center. For this trip, my walking has only been between the hotel and the Moscone center, absolutely no sightseeing.

The weather has been great so far, almost 65 to 75 degrees Fahrenheit. everyday Overall it was been a joyful trip.

Looking forward to going to Vmworld 2010.

Stay tuned for the final post on Vmworld 2009 tomorrow.