Zerto, one of the sponsors and presenters at Gestalt IT’s Virtualization Field Day 2, pitched a rather unique product that handles replication of virtual machines at the hypervisor layer (as opposed to the more “traditional” storage layer). It’s unfortunate that I’m not able to link a video to their presentation, but Zerto opted out of a recorded session so that they could offer a higher amount of disclosure on their latest product build (2.0 beta) and also go into some depth on their customer portfolio. If you’ve had to deal with disaster recovery (DR) from a perspective of log shipping, LUN replication, or even VMware’s Site Recovery Manager (SRM) product using traditional storage replication, you’re familiar with some of the headaches that go along with these approaches. Zerto saw this as an opportunity to enter the market, and built a product that strives to resolve these issues by harnessing some strong personnel with technical DNA from the DR space (Ahem, Kashya) . While Zerto isn’t exactly a new player to the space, having won the “Best of Show” award at VMworld 2011 (Las Vegas), I was not that informed on their product and was very impressed with their offering and future road map.
The Big Guns
When you’re coming to any Tech Field Day event, go big or go home. Zerto stepped up to bat with a pair of heavy hitters: Gil Levonai, VP of Marketing and Products, and Oded Kedem, CTO and Co-Founder of the company. Gil lead much of the discussion on the product from a high level (who we are, what we do) and was open to any shots from the audience. It’s obvious that he is not “just a marketing guy” as he was quick to respond to any question with a detailed answer, and would state facts in an absolute manner. Oded fielded many questions of the “under the hood” variety and configured a remote session into one of their environments for a live demo. The presentation was very light on PowerPoint and included a detailed walk through the product’s actual interface, along with a test failover of their demo environment. This is a great recipe for success when your room is filled with thought leaders and technical subject matter experts.
Gil Levonai discussing Zerto with the delegates at Virtualization Field Day 2
Why Not VMware Site Recovery Manager (SRM)?
In all reality, SRM was not brought up that often during the presentation because it’s such an apples to oranges comparison in many respects. Although SRM does have vSphere Replication in the version 5.0 release, the main selling point for enterprise customers is still storage array replication – just look at any marketing documentation on SRM and you’ll get the hint that vSphere Replication is for ROBO or small deployments. And indeed, with Mike Laverick in the room (the guy who literally wrote the book on SRM), the delegates were well covered on that angle.
This is roughly how Zerto feels about Storage Array replication. Poor truck.
Protecting the Application
I think the major thrust behind Zerto’s offering is the idea of protecting applications, not LUNs or a pile of VMs. With the Zerto product, two pieces are inserted into the virtual environment: a Zerto Virtual Manager (ZVM) that ties into vCenter as a plugin and manages the replication stack, and Virtual Replication Appliances (VRA) that are deployed by the ZVM onto each vSphere host. Each VRA is able to see the writes performed by a virtual machine on the host it is sitting on, which enables it to then ship the changes over the wire (be that LAN or WAN) to the recovery site’s environment. Because Zerto is aware of the virtual environment, it can protect application groups, such as folders of VMs, and vApps in both vSphere and vCloud Director. The product doesn’t care if VMs vMotion to another host, storage vMotion to another datastore, or even participate in SDRS – VM location data is available to the ZVM, which is in turn sent by each host’s VRA, and protection continues as the VM moves around.
Zerto uses the idea of Virtual Protection Groups (VPGs) to protect the VMs that comprise an application
There are some pretty giant advantages of not sitting low in the stack (such as with SRM’s Storage Replication Adapters or SRAs) in that the product can be storage agnostic (huge win for most companies and a requirement for future cloud use cases) and also storage configuration agnostic (LUN arrangement no longer matters).
There’s an entire laundry list of bullet points covering the presentation written by Rodney over at the Musings of Rodos for a complete overview of the presentation, along with a great blog post created by Edward that discusses the backup of virtual applications on The Virtualization Practice.
There were a number of advantages to using Zerto that I found to be truly Enterprise Class:
Per-Virtual Disk Replication Settings
Depending on your environment, there many be several VMs within an application group that contain virtual disks that you don’t want to replicate, such as tempdb for SQL, a disk dedicated to the page file, or some sort of near-line archival repository. If you choose not to replicate that particular hard disk with other products, you’re left with a VM that recovers and screams “Where is my disk!?” in a pissed off tone. Sure, you script around this, or hand edit the VMX files, or do one of several other options to fix this by hand. But isn’t the point of buying a product to avoid this situation? Zerto allows you to replicate the disk once, then turn off replication for future writes. This ensures that the disk is at the recovery site and loads into the VM properly, but doesn’t suck up bandwidth replicating what I’ll term as “garbage” data.
CDP-Like Journal System
Another feature I appreciated after watching the demo was the granularity of configuration. You could literally sit down at a disaster recovery / business continuity meeting, figure out your RTO and RPO objectives, and then feed them into Zerto. This includes setting the size of journal files (based on number of hours) to figure out exactly how far back you want to recover to. Need to recover an application from an hour ago? Just move the slider to 60 minutes in the past and click the red button. That’s pretty spiffy!
Testing and Reporting
Additionally, the product can offer up reports on your DR tests, and allows you to set an “age” value on how often a DR test should be performed. This seems like a great thing to send up the chain to the C-level guys who want to check the box on DR.
I was very impressed with both the presentation given by Gil and Oded and the Zerto product itself. They are working on a number of improvements above and beyond what I’ve laid out here that should be available once 2.0 comes out to the masses. While Zerto is definitely setting the trend right now, it will be interesting to see when and how VMware’s SRM responds as future builds are released.
If you’ve used their product, I’d be interested to hear comments on what you think about it below.