My three- year-old daughter Sophia has difficulty making decisions about what clothes to wear. Sunday morning, before church, she cycled through four fancy dresses in 10 minutes. As we slid from “fashionably late” to “in danger of missing post-mass donuts,” I injected some male leadership. “Pick anything. You’ll look beautiful. We have to go. The jelly donuts will be gone! There is no wrong choice.”
Sophia came downstairs wearing only a t-shirt (on backwards) and underwear, ready to go.
Moral: Sometimes there is a wrong choice.
One of the most common questions I get asked is: How should I back up my VM environment? While I’d like to say “there is no wrong choice, just pick one,” I’ve seen more than one backup administrator wearing only a t-shirt and underwear… metaphorically speaking, of course.
What is the right choice then? The answer for the short-term depends on your current environment, your business pain points, etc. Now, I’m not going to pretend that I can answer this question via a simple blog post (seriously, you’re more likely to get practical advice from Gwyneth Paltrow than from a CTO… someday, I’ll have to post an answer to the question, What does a CTO do all day?). For these, you’re better off working with your friendly local reseller or technical consultant. However, what I can do is help you understand the broader industry trends so you can map out your desired technical destination. In this blog post – and the following two – I will walk you through those trends.
So, let’s start by taking a look at some of the deployment styles I commonly see:
- Treat VMs as regular servers (guest OS backups)
- Run VM Consolidated Backups (VCBs)
- Leverage VMware’s Changed Block Tracking (CBT) to create a versioned replica on backup disk
- Perform storage-based versioned replication
While the first two solutions are traditional approaches, the latter two highlight the shift toward the integration of data owners and data protection.
Backup vendors, bloggers, and experts have spilled more barrels of e-ink on guest OS and VCB backups than analysts did on CDP back in the mid-2000s, hopefully to better effect. Rather than regurgitating all the arguments, I’ll boil it down: If you want to preserve your standard backup and recovery processes, one of these approaches is the right approach. And, of course, every backup vendor will explain that 80+% of their revenue come from customers who protect at least one VM (of course, since almost every company has a few VMs, that isn’t saying much) and that they’ve optimized for VMware (but what company would say, “We’re terrible for this market segment.”?).
Most people agree that, since CPU and I/O tend to be premium resources in a VM environment, you want to leverage the most-efficient clients available. We have found one basic efficiency principle:
Don’t back up the same chunk of backup data twice. Aside from the reducing the network and backup storage system load, deduplication improves the performance on the server itself. That’s right, on most modern processors, an optimized deduplicating backup client will reduce CPU load compared to one that simply transfers backups with 95% redundant data.
While the deduplication principle above can help you find a solid solution today, it has more far-reaching implications for the future. Remember, the focus of this series of posts isn’t about optimizing today’s solution, but to help you set your strategic direction for protecting your VM environment.
In my next post, I’ll dive more deeply into the next-generation backup technologies (versioned replication via VMware CBT vs. storage-based versioned replication), and in the final post in the series, I’ll walk through some customer deployment examples for these leading-edge technologies.
After all, when it comes to protecting your VMs, nobody wants to get caught with their pants down, off or on the ground.