Find Communities by: Category | Product


5 Posts authored by: Cody Garvin

Do you have storage? Do you stretch your vSphere over distance?  Then you’ve probably talked to your EMC rep about VPLEX and running vSphere with VPLEX Continuous Availability.  In fact, you probably use VPLEX to stretch your vSphere cluster between two datacenters.  And lately, you may have heard about VMware’s own solution to this problem, Virtual SAN (VSAN) Stretched Clustering.


Since VSAN Stretched Clustering was released in August of 2015, we have seen a steady chain of questions asking about the key differences between VPLEX Metro and VSAN Stretched Clustering.  I thought it would be helpful to call out the main differences and similarities here, in a central location, for everyone to see.  Nothing to hide here folks!


The most important point to cover, right from the start, is that VSAN Stretched Clustering is a great solution for running vSphere Metro Storage Cluster (vMSC) between two datacenters! It absolutely works, and does exactly what you need it to do.  VSAN’s stretched solution allows you to use VMware High Availability (HA), Dynamic Resource Scheduler (DRS) and, of course, gain instant vMotion over distance, just like you get with VPLEX.  It also gives you the option of running a VM-based Witness to allow for increased survivability for your VMs in the case of a site or network outage.  VPLEX expert confirmed and approved!

Is it strange for a VPLEX blog to go on about VMware’s awesome VSAN Stretched Clustering solution? Not really. Has this author completely lost his mind?  Not relevant.  I can focus on the excellence of the VSAN solution simply because VSAN and VPLEX don’t really compete.  In truth, VSAN and VPLEX provide many of the same benefits but target separate customers. It all comes down to this.  With VSAN a customer must use all direct-attach storage, and with VPLEX all the storage must come from supported storage arrays behind VPLEX.

This really is the major difference to the customer between the two. Is your storage on your vSphere servers, or do you have a Storage Area Network (SAN)?

There are other differences, of course, that are less glaring and less impactful.  VPLEX supports connectivity between sites using either Fibre Channel (FC) or Internet Protocol (IP), while VSAN is only IP based. Even within the support of IP connectivity there are differences in the recommended minimum setups between VPLEX and VSAN: 3Gbps vs 10Gbps, up to 10ms RTT vs 5ms RTT, layer 2 or 3 connectivity vs layer 2 only.  Finally, there is the fact that, although the new Fault Tolerance (SMP-FT) is not supported in any vMSC environments, VPLEX does support using the older RR-FT for automatic FT recovery of VMs between sites; VSAN does not.

As I’ve said, all of these are minor compared to the major point of leveraging your existing storage infrastructure.  Being able to use XtremIO for VDI, or VMAX for mission critical databases, Unity, or even non-EMC arrays means better performance and higher availability for your VM environment.  As soon as you bring SAN storage into the discussion, VPLEX allows you to take advantage of your existing investments and enables the use of targeted best-of-breed technology.

While VSAN and VPLEX can provide the same benefits and cover the same use case of stretching your VMware cluster between two datacenters, with all the benefits an Active/Active solution provides, if you have any existing storage arrays VPLEX is your only choice between the two.  VPLEX Metro is trusted by thousands of customers to provide Continuous Availability for VMware vSphere clusters.  The arrival of VSAN Stretched Clustering doesn’t change that trust.  Pick the option that best fits your environment, and be happy with the knowledge that your vSphere environment and your VMs have been given Metro superpowers!

The Advantages of Data Services Transparency

by Colin Durocher



In Part I of this post, I established that VPLEX is singularly focused on adding mobility and availability to application environments while not subtracting other data services from the overall solution.  I discussed how this purity of mission is behind VPLEX’s success. In this post (part II), I argue by way of example, why mobility/availability aside, customers are generally better served leveraging the fit-for-purpose capabilities of all of the products making up a solution rather than relying on the virtualization layer to provide all the data services.  I do this with a detailed look at space efficiency, operational and disaster recovery, and copy services.


Space Efficiency


One of the main competitive products to VPLEX makes a lot of noise about compression and the space savings it makes possible.  Capacity savings are especially important when raw capacity is relatively more expensive than performance (i.e. with AFAs).  The VPLEX equivalent solution would include XtremIO, bringing both compression and de-duplication to the table.  XtremIO + VPLEX customers get this “always-on” space efficiency functionality compromise-free.


Our competition performs compression at the virtualization layer, using enormous amounts of CPU processing power and cutting the performance of the solution to a fraction of what it otherwise would be.  End result is that the VPLEX solution delivers better space savings (c/o inline de-dupe and compression) with no performance loss vs. native array performance. Competition delivers mediocre space savings and has introduced a major bottleneck into the environment in the process (i.e. storage silos fronted by over-burdened dual-controller storage virtualization).  This clearly shows the performance advantage of offloading data services.  As more workloads transition to flash, this VPLEX performance advantage will only grow.


Disaster and Operational Recovery


This is not usually the first place that competitors try to poke VPLEX in the eye but I wanted to bring it up to make three points supporting the overall argument:


  1. EMC offers the best operational recovery solution in the industry in the form of RecoverPoint (RP) and this is available with all EMC storage arrays as well as with VPLEX to serve heterogeneous environments.   RP customers who understand the value it brings with its continuous data protection (CDP) would not want to trade that off in the process of virtualizing their storage.  With VPLEX, they can have their cake and eat it too.
  2. When it comes to availability, MetroPoint offers a unique 3-site solution that combines continuous availability between two primary sites, independent CDP at each of them and continuous disaster recovery (DR) to a single third-site copy. Continuous DR means that through a site failure, DR replication instantly fails over without the loss of a single replicated write.  The DR SLA is never violated.  No competitor can currently touch that.  But again, you wouldn’t see this on a direct comparison of VPLEX vs its competitors. The power of MetroPoint is in the combined solution.
  3. Recently we had a large banking customer with a long history of using SRDF for DR going back to DMX days.  Because of the trust they had built up with SRDF over many years of experience, they did not want to switch replication technologies.  With VPLEX, they can continue to use their preferred DR technology without issue.  This customer wanted to use VPLEX to virtualize their DR site too.   Consistent with our DO NO HARM principle, we are making changes to allow the SRDF R2 devices to be encapsulated by VPLEX.  I raise this example because sometimes the choice of data services is more than a financial or spec-sheet based decision. It is sometimes an emotional one based on trust or comfort.  This is also a great example of investment protection. With VPLEX, this customer is able to leverage the investment made in SRDF (licenses, training, experience) while benefitting from the addition of VPLEX’s mobility and availability.


Copy Services


I could write here about any of EMC’s storage platforms as they all have great copy services implementations but let me again take the example of XtremIO which has some unique capabilities in this area.  XtremIO allows you to:

  • Instantly make writeable snapshots without any data/metadata bloat (thanks to native always-on deduplication)
  • Make snaps of snaps of snaps, ad infinitum
  • Combine any set of snaps into consistency groups (even overlapping ones)
  • Refresh a snap from a snap, prod from a snap, or a snap from prod
  • Do all this without any performance penalties associated to the copies


In other words, they are a state of the art copy services implementation that would compare favorably to any other implementation out there.  If you were to virtualize an XtremIO array with one of our competitors, you would lose much of this functionality and be left with the limits inherent in an outdated design not optimized for the flash era.  How is that a good thing?


Speaking of limits… Typically snaps and clones count against the scalability limits of the virtualization platform itself.  With VPLEX, this is only the case if the copy is actually presented through VPLEX to a host. Copies for recovery purposes don’t typically get presented to hosts and copies for test & dev may not require the VPLEX mobility and availability data services and thus can be presented directly from the array.  This gives VPLEX a practical scale advantage vs. its competition, if not one that is immediately apparent on the spec sheet.


There are many additional examples beyond those I’ve outlined above where leveraging the native data services of the virtualized storage array is a winning proposition but I don’t want to belabor the point. The bottom line is that EMC products are laser-focused on doing what they do best, and hence they do those things very well. Customers are well served when they allow all that value to shine through the solution.  As we covered in part I of this post, VPLEX is focused on availability.  Competitors to VPLEX, on the other hand, aim to be jack of all trades - an attractive proposition until you consider that they may also be master of none.




About Colin Durocher...


Colin Durocher is a Consultant Product Manager with EMC and has spent over 10 years working on the VPLEX product in a variety of roles within Engineering and Product Management. He currently owns coordination with storage array partners, as well as pushing VPLEX performance and scale to meet increasing customer requirements. He’s also defining future VPLEX product strategy, direction, and innovation.

Outside of work, Colin enjoys racquet sports, brewing beer, camping, and spending time with his wife and two children in Montreal, Canada.

Clarity of Vision, Purity of Mission

by Colin Durocher


Strategy is about making choices, trade-offs; it’s about deliberately choosing to be different

            - Michael Porter


The storage virtualization market is divided into two camps.  In one camp sit those who aim to commoditize the underlying storage arrays and provide all the needed storage intelligence in one package within the virtualization layer.  I call these the Jacks of all Trades virtualization products.  In the other camp sits EMC’s VPLEX.  VPLEX’s raison d’être is very clear and the data services approach it employs is unique in the industry.  In this post, I lay out the VPLEX mission and data services strategy and I explain why they are key to customers’ success with VPLEX.


VPLEX Vision:  Providing peace of mind through continuous uptime and the flexibility to move applications (and data) anytime, anywhere


VPLEX is an inter-array data services platform.  VPLEX’s mission is to keep applications running, no matter what happens.  It does this by providing availability and mobility-related data services.  These generally break down into 2 core use cases:

  • Eliminate planned downtime:
    • Tech Refresh of virtualized arrays
    • Storage Tiering and Load Balancing
    • Maintenance Avoidance
    • Datacenter relocation
  • Eliminate unplanned downtime:
    • Stretched Clustering for protection against site level disasters
    • Local Mirroring to protect against storage failures


What about all the other data services that I need to keep my business running?  A core VPLEX goal is to “DO NO HARM” to the storage environment. VPLEX aims to add the above capabilities while not subtracting any existing data-path capabilities.  As a result, VPLEX has adopted a strategy of encapsulating storage and transparently leveraging the existing data services on the underlying arrays.  This provides for investment protection in cases where those data services are bought and paid for, and it provides new VPLEX customers with the capability to select their storage platforms based on their unique capabilities and fitness to the task at hand.  VPLEX provides all the needed data services in the solution, rather than in the product.


It is, of course, no coincidence that EMC would employ the pass-through strategy for its virtualization platform as its storage arrays (ie VMAX3) have best-in-class data services. Why would we seek to replace those?  Making use of native array data services allows VPLEX to focus efforts on the core product value proposition – Continuous Availability. Which is why VPLEX is able to offer industry-leading availability for mission critical applications.  This is why customers purchase VPLEX and as a result VPLEX adoption and market share both continue to grow.  [UPDATE: My colleague Parasar documents some customer examples in this recent blog.]


Deciding what not to do is as important as deciding what to do.

            - Steve Jobs


Our competition have all chosen a strategy of replacing the functionality of virtualized arrays with their own data services implementations.  While this can add value in older environments and can certainly simplify licensing in heavily heterogeneous environments, it often renders otherwise superior data services on the arrays completely unusable.  Not surprisingly, they love to setup feature-by-feature comparisons to VPLEX on its own.  Since VPLEX does not (by itself) offer space efficiency or copy services functionality, these comparisons predictably end up very one-sided.  But in the age of converged infrastructure and outcome-minded customers, it is more useful to compare solutions.  In these comparisons, the EMC solution leads the industry by layering best-of-breed availability (VPLEX) over best-of-breed data services from arrays that are fit-for-purpose.  And that is how EMC and, more importantly, customers win.


In Part II of this post, I delve into concrete examples why customers are better served with offloaded array-based data services.  Hint:










About Colin Durocher...


Colin Durocher is a Consultant Product Manager with EMC and has spent over 10 years working on the VPLEX product in a variety of roles within Engineering and Product Management. He currently owns coordination with storage array partners, as well as pushing VPLEX performance and scale to meet increasing customer requirements. He’s also defining future VPLEX product strategy, direction, and innovation.

Outside of work, Colin enjoys racquet sports, brewing beer, camping, and spending time with his wife and two children in Montreal, Canada.

I believe in stupid questions.  They are born from a lack of knowledge and can’t we all relate to that?  For instance, a coworker recently asked me: “With VMware’s vSphere 6 coming out, how does that change things for VPLEX?” To which I replied, “Not much.” Because when you ask a stupid question, the person who answers is allowed to be deliberately obtuse.  That’s a rule I just made up.


The real answer is that, really, not much changes.  The three big items that launched with vSphere 6 were VVOL support, Long Distance vMotion, and expanded Fault Tolerance support.  I’ll get Fault Tolerance out of the way right now.


The new Fault Tolerance changes in vSphere 6 are amazing!   Finally you can run FT on your mission critical VMs, on VM’s of substance!  But excellence comes at a price, and today that price is distance.  FT is not allowed to run in Stretched Metro environments.  In VPLEX Local environments, FT runs great.  On VPLEX Metro environments FT is ruled out.  It’s as simple as that.


Since FT was ruled out, what interested my coworker were VVOL support and this new vMotion.  I’ll start with VVOLs.  For many people, while they know about VVOLs, they may not really get it if you know what I mean.  I will now attempt to greatly overly simplify what VVOLs are, and why they are important, in a general hand-waving sense.  The two aspects of VVOLs that are important are 1) per VM storage and 2) control path interface with storage arrays.  For each VM you now have 3-5 LUs that hold things like the VM configuration, running state, and of course the VM data.  This collection of LUs can then move or grow or shrink or be protected on a per-VM basis.  The control path aspect means that there is a well-defined way for vSphere to now tell the storage array what needs to be done for any particular VM.  All of this leads to a policy-based style of management.  Setting up a new VM? Great! Here is our menu. Would you like flash storage with that?  May I also recommend the salmon?  It’s just in and very fresh.


So what’s the plan for VPLEX and VVOLs?  The short version is we’re working on it.  The slightly longer version is that VPLEX has to work with many other storage arrays, and to do that with VVOLs we need to utilize the VVOL control interfaces of those storage arrays ourselves.  Once they have VVOL support, we can build on top of that foundation.  This means that VVOL support for VPLEX is coming in the future, which means we don’t support it today (bypassing a stupid question there with a stupid answer).  Should the delay in VVOL support cause you sleepless nights, an increase in anxiety, and a constant urge to pop a Valium (thank you Dr. Smith)?  Not really.  The concept of VVOLs is just getting started and VMware is continuing to grow and improve on the VVOL idea.  In the meantime you can continue to use VPLEX as you always have to support your VMware environment with VPLEX continuous availability with VMware High Availability, metro stretched clustering, instant vMotion, and load balancing between sites.


Now that those topics are out of the way, can we talk about the elephant in the room?  I’m going to assume that most of you don’t actually have an elephant in the room, and instead talk about Long Distance (LD) vMotion and VPLEX.  For those of you who haven’t been attending VMware technical conferences, LD vMotion allows vSphere 6 to (wait for it) vMotion individual VMs over extremely long distances.  Previously with vSphere you were limited to metro distances which equate to about 5-10 ms of roundtrip network latency between sites.  That usually means within a city or between relatively geographically close cities.  With LD vMotion you can move a VM across continents!


I imagine those of you who have and love VPLEX are quivering with excitement at this point.  In a Metro environment VPLEX enables instant vMotion of VMs between sites.  VPLEX makes sure all the data for the VMs are read/write accessible at both locations and VMware auto-magically knows that it can just move the VM running state and BAM! Instant vMotion.  So can VPLEX now do the same thing with LD vMotion and VPLEX Geo (or VPLEX Metro)?  Sadly, no.  But don’t worry, since no one else can do it either!


With LD vMotion, VMware has allowed for two separate methods for enabling the vMotion.  The first uses a new file-based version of storage vMotion that runs within ESXi between vSphere clusters.  This system transfers all the base information about the VM, then all the storage of the VM, and finally the running state of the VM between sites.  The long pole here is the transfer of the storage, which is what VPLEX eliminates in a Metro environment.


The second mechanism is for VVOL-enabled Active/Active storage.  This means that if a storage array supports VVOLs, and that array also supports Active/Active storage, then the LD vMotion doesn’t have to move the VM storage.  Amazing!  So amazing in fact, as of today, no vendor or product supports this VVOL based mechanism!  So your ONLY choice today is through the vSphere mechanism, and a full storage vMotion.  Feel free to check the VMware support matrix and see for yourself.


What does LD vMotion mean for VPLEX customers?  If you are truly transferring VMs at long distance, you have no choice but to sit and wait for your VM storage to transfer, just like everyone else.  If you are actually at sync distances, using VPLEX Metro to pre-stage VM data enables instant vMotion.  And in the not-to-distant future, VMware could (in theory) allow for other VVOL-enabled replication technologies like RecoverPoint and SRDF to pre-stage your VM data over long distances enabling virtually instantaneous LD vMotion.  Pun intended.


Like I said before, not much really changes with vSphere 6 and VPLEX.  VVOL support is baking, FT is local only, and LD vMotion support is so limited that your only option is the VMware solution.  For VPLEX the focus continues to be on the benefits gained by using VPLEX Local or VPLEX Metro with your vSphere environment.  Those of you who have VPLEX today know the enormous value of having native VMware HA and VPLEX Continuous Availability protection.  For those of you who don’t…  Isn’t there someone you can talk to about that? I know, stupid question.


Cody Garvin

VPLEX Product Manager

Announcing VPLEX support in Virtual Storage Integrator (VSI) 6.5 available 4/13/2015! 


New with VSI 6.5, VPLEX provisioning and management has been added to the VMware vCenter management plugin.  You can now view, provision and control VPLEX enhanced storage in conjunction with EMC ViPR, VMAX, VNX, VNXe3200, XtremIO and AppSync with this EMC developed free plugin!  If you have VPLEX in your VMware environment, with or without additional EMC arrays, consider using VSI 6.5 to allow VMware integrated management of all your systems from a single touch-point.


Beyond the VPLEX support, VSI 6.5 also adds new capabilities around:

  • Support for the recent vSphere 2015 release
  • Support for VMware NMP and EMC PP/VE Multipathing for VMware
  • Additional platform enhancements for VNXe3200, AppSync, and VMAX3


For the full release announcement and details please head over to:


Or you can download VSI 6.5 today from the EMC Support Site:

Filter Blog

By date:
By tag: