Protecting and Recovering Virtualized Applications is Challenging
By Zerto, on 22 February, 2012
This is a guest post by Steve Thomsen, Zerto’s Director of Central US Sales.
Protecting and recovering virtualized applications has moved to the forefront of data center strategy. However the tools and methods available are based on technologies that are, well…quite old. Host and storage-based replication solutions have been around since the 90’s. The last big thing to come to either of those two technologies was network-based replication for storage arrays. That was 12 years ago.
Host-based solutions do not scale nor do they do much to address multi-tiered applications such as SAP, SharePoint, or even most Exchange deployments. Storage-based replication solutions scale and can protect multi-tiered applications, but they do so with a terrible impact in terms of infrastructure costs and complexity (this is really where virtualization breaks storage and that’s a discussion for another day).
Zerto’s Hypervisor-based approach to data protection and recovery goes a long way to solving the conundrum of, “How do I best protect my VMs?” Or more accurately, “How do I best protect my virtualizated applications?”
A good example of this method is offering application consistent point-in-time images for recovery. For Microsoft applications Zerto supports VSS through the UI and command line. Any other application you can quiesce, such as Oracle via RMAN/Hot Backup, can be protected in such manner through command line.
What is unique about Zerto’s method for capturing application consistent point-in-time images isn’t what we are doing, it is where we are doing it. To illustrate this, let’s look at taking an application consistent image with legacy technologies (for virtualized or non-virtualized applications). First, you quiesce the application at which point its cache is flushed to disk. Once you have all transactions in flight committed to disk you know you have 100% data integrity. Next you take an application consistent snapshot and finally release the application back to production. This whole process takes 2-3 seconds. For most administrators, or more properly for most organizations, the penalty in terms of time to execute an application consistent snapshot/image is too disruptive to the business. In many cases administrators will only take a very small number of application consistent images/snaps in a day or forgo the process altogether.
So what does Zerto do that it so different?
Zerto’s method of providing application consistent point-in-time images is much the same, the big difference is where we capture the application consistent point-in-time/image. Now if you have read anything about Zerto, you know we use Virtual Replication Appliances (VRAs) to perform replication. These VRAs protect VMs by residing on the host and replicating selected VMs and VDMKs. You do not have to protect all of the VMs on a host.
When you want to take an application consistent image, you basically perform the same process…but faster, much faster. Now when you quiesce the application and all I/O is flushed to disk, the flushed write I/O is copied and sent to the Zerto VRA on that host at memory speed. Zerto already has the rest of the data at the target – all +99.99999…% of it. With the cache dump, we logically have 100% of the data. Before replicating this last bit of change across the WAN, Zerto marks the image as VSS/application consistent image “whatever” and replicates it at the speed of light across the WAN. Because the cache dump and VSS/application consistent image checkpointing is performed in host memory, the entire process takes microseconds vs. the full 2-3 seconds other solutions.
What does that mean to a Zerto customer? Enterprises can have many more application-consistent point-in-time images when compared with any other solution – and there is no production environment performance impact of capturing those images. The data center is not only crash-proof, but allows easy fail-back to an application-consistent point in time in the event of a corruption.
Application-consistent images can be captured at pre-defined points-in-time in addition to Zerto native continuous near-synchronous replication; which provides hundreds and even thousands of crash consistent point-in-time images for recovery.
Is this the number one reason people choose to adopt Zerto? Not by a long shot. But it is an excellent example of the power of being in the hypervisor. Among other great aspects, it is about speed and efficiency.