Virtualized availability: Small is beautiful and perfectly formed
I’ve worked with a lot of “enterprise” products in my time, which may make it sound like I work mostly
with big multi-nationals. But the truth is in my business I hold the roles of CEO, CTO, CIO, CFO and tea-boy. That’s right — a company of one! I’ve had the SMB in mind throughout my career because I’ve shared the same constraints as my SMB customers: time, skills, and of course, resources.
Most of my time in IT has been spent as a consultant and trainer. I’ve seen an awful lot of different businesses over the past 20 years, and I can feel for those who have to wear half-a-dozen different admin hats — I do too. In my collocation I have eight servers and a couple of storage boxes, and if anything goes wrong it’s up to me to fix it. This means I’ve got a long drive to the office ahead of me should something go so terribly awry that I need to be physically present. Increasingly, making sure the lights are on is critical, but at the same time I can’t spend all of my time doing that. If I did, I’d never get any other work done.
Since 2003, when I first got into virtualization, everything I do is virtual (including Active Directory, Databases and even vCenter itself). For me, virtualization is more than just server consolidation — it has the power to transform everything we do in the IT environment. If we restrict virtualization to only enabling server consolidation, we miss out on the other big benefits it brings. Namely, virtualization can transform how we prevent and recover from disasters, offering a simple way to protect our businesses from acts of God or acts of great stupidity (i.e., human error).
The center of my world isn’t server consolidation. It’s actually the virtual machine (VM), and the fact that the entirety of an operating system, its applications and data, can all be contained in a series of files — namely the virtual disk. That’s the biggest single change that virtualization brings. Understanding this “encapsulation” as it’s sometimes called, means you can empower your business to do things that simply weren’t possible in the bad old days of “physicalization.” With virtualization, moving VMs from machine to machine is a trivial task making it easier to back them up, avoid failures or even scale services to meet increasing demands.
The old faithful backup
Regardless of all the funky new technologies on the market, backup remains essential in most recovery strategies. It protects from the loss of individual VMs as well as loss of data within the VMs. If not generating a great deal of data change in any given day, many SMBs can recover from a backup set well within their Recovery Time Objectives (RTO) and Recovery Point Objectives (RPO) — handy terms for setting your organization’s constraints on how long it takes you to recover the data and what state the data is in once it is recovered.
Virtualization simplifies backup by doing away with the pain of having to install and manage agents on every OS instance. It also reduces costs by eliminating the need to pay vendors for backup agents installed on each guest OS. After all, a VM is just a file and can be backed up via the server that runs the VM — a relatively simple process.
At the start of the backup job a VM snapshot is applied. This process flushes the file system cache and unlocks the virtual disks that make up a VM, allowing the files to be archived elsewhere. Modern backup systems from vendors like VMware ship with features that can detect what data has been modified since the last backup, so only files that have changed get backed up. VMware calls theirs “Change Block Tracking” (CBT) and it results in much shorter backup job times. Those of us who have been around in the virtualization community for some time regard this approach as being essential. It’s been shown that running a backup job using the conventional agent method increases the CPU load on the server and can flush the network with excessive I/O.
Of course, the need to restore entire VMs is pretty rare, and no one really wants to restore gigabytes of data. The vast majority of restores are of relatively small amounts of data. For this, VMware has a self-service utility called the “VMware Data Recovery Client,” part of the VMware Data Recovery (vDR) Appliance and integrated directly into vSphere. vDR adds views to vCenter that allow you to backup and restore VMs. The Data Recovery Client can be copied and extracted into the operating system, and it allows for the mounting of the backed up virtual disk. This way you can copy the files from the mounted virtual disk directly back into the system where the files were lost.
Murphy’s Law is a law for a reason
Protecting yourself from data loss doesn’t begin or end with backup alone. Consider the possibility of a major disaster wiping out an entire site. In the past many organizations had to live with this threat and hope that a backup strategy on its own would protect them. As the quantity of data grows year by year, many SMBs are beginning to realise that the restore process might take too long to complete, putting their business in jeopardy.
In the early ’90s I worked for a business that was affected by an act of terrorism. In 1996 at the height of the Irish “Troubles,” the Irish Republican Army (IRA) detonated a bomb near the Arndale Centre. The business I was working for was affected. Fortunately no members of staff were injured, but it did cause us to shut down our Manchester office, which was our head office central to operations. As a relatively junior staff member at the time, I was only involved at high level with the recovery process, but I did get insight into how unforeseen events can rock employees’ faith in the company. “Will we get paid?” some people asked. “Will the business survive this impact?”
Back then, there was no mainstream virtualization, but if there had been, I can only imagine how much simpler it would have made things.
Virtualization allows us to replicate VMs from one site to another at a fraction of the traditional cost and complexity. Traditional site-to-site DR required equal physical servers in both locations as well as expensive storage array replication. This came with a hefty price tag, including the upfront cost of having to buy a second storage array at another location.
Virtual replication leverages a virtual appliance sitting on a host at each site — the protected site and the recovery site. The virtual appliance captures changes taking place inside your protected VMs and replicates them to storage in the recovery site. This “host-based” replication, as it’s sometimes referred to, sits so high in the layers of the stack that you really don’t have to care what storage is in use — VMware calls theirs “vSphere Replication.” This means you can replicate a VM residing on any type of storage to any other. For example, you could be using arrays from the likes of NetApp or EMC, and replicate to the recovery site to storage from an entry-level vendor. Virtualization is now doing for replication what it did in the last decade for servers — making it easier to manage and more cost effective.
Virtualized replication offers a more granular way of protecting your servers. Rather than having to replicate entire datastores it allows you to pick out and prioritize the VMs that really matter to you.
Buddy up for site recovery
Beyond backup and site replication, virtualization has enabled customers to look at availability in new and creative ways. Increasingly, I’m seeing SMB customers “partnering” to offer DR resources to each other. I’ve heard stories of two small U.S. colleges using each other as DR locations. The great thing is that they don’t even share a storage vendor — replication gives them a solution that works. In another case, two companies buddied up. This worked easily since the IT manager for one company had previously held the same position at the other. In fact, the two different companies had exactly the same setup from a storage vendor perspective — after all, what’s the point in re-inventing the wheel?
This partnering approach is a good way to drive down the cost of doing DR while sharing best practices with other businesses. For an SMB of course, the recovery site could be just some rack space you hire from a collocation facility, but I think the future is really DR being offered as a service. In this model the SMB signs up to a deal with a service provider who offers replication as part of their package. That way, all you have to do is worry about the recovery plans, and the provider takes care of the complexity of configuring the hardware resources required to make it all happen. This removes the concerns about warranties and maintenance contracts at the recovery site — that’s the provider’s problem.
For the service provider, replication enables DR-as-a-Service to as many folks as possible without having to worry about the hardware differences from one customer to another. Hopefully this will bring economies of scale to DR that will benefit everyone. This new approach will be ideal for SMBs who have a single site and lack “buddies” they can team up with. In this way the service provider takes over the role of the buddy.
Bye bye, physicalization
Virtualization goes way beyond the narrow confines of just consolidating servers. It gives you an opportunity to do what you used to do — better, while replication empowers you to offer availability to the business that previously was regarded as too costly. To maximize the value of your server consolidation project you need to look beyond the low-hanging fruit and think about how virtualization opens up new opportunities to streamline the burden of your daily administration tasks.
How has virtualization improved availability in your environment? What can you do today that you weren’t able to do before virtualization?
Mike Laverick is a VMware vExpert who writes, instructs and otherwise communicates about virtualization.
Get to know Mike – Read 10 Questions With… Mike Laverick