Welcome to Virtualization Phase 2 – Adventures in VBCA
“Has VMware finally peaked?” is a question that pops up frequently in conversation with various members of the technical and financial community. Typically, the underlying concern is over penetration of virtualization throughout the various enterprises of the world. Is there anything left to virtualize, or have we taken care of all that is possible?
To this I say: Virtual Business Critical Apps (VBCA)! It’s a fancy term for “virtualizing the really, really important stuff” which have typically been avoided in the past.
Why VBCA?
Well, let’s go down that rabbit hole. I view this as two separate paths to go down in the decision making process:
- Scenario 1: The workloads we “can” virtualize
- Scenario 2: The workloads we “should” virtualize
You’ve probably run into many of the scenario 1 workloads. Can you virtualize a 64 way SQL box or an old telephony server that has hummed along fine for a decade? It’s certainly possible, but doesn’t always make sense. Much like like the idea that we “can” drive down the wrong side of the road – just because you can doesn’t always mean you should.
I tend to focus more on scenario 2 workloads.
Let’s take the same 64 way SQL box example. Maybe it’s only running on 40 of those cores, or consuming half the memory, or runs mostly batch workloads. This would be a waste of hardware and would benefit from the “right sizing” possible with virtualization. It isn’t any less critical for being over sized, but some opportunity exists to better allocate the hardware. Additionally, it’s quite common that this critical server would have some offline hardware for disaster recovery, and may need to be very similar (if not identical) in spec. This ensures that restoring a backup would be highly successful and avoid driver problems. Again, standardizing on a virtualized platform eliminates a lot of that headache. And further – what if the database isn’t clustered? Maintenance windows would be quite painful to arrange. Even if it is clustered, there are still opportunities to virtualize using more recent methods such as an AlwaysOn Availability Group (as an example).
How about that old telephony server? Perhaps you are not currently providing disaster recovery, or the application can’t be clustered. It would be a huge pain to restore to a similar physical server. And so I say – we should (probably) virtualize that!
Why probably? Take all this with a grain of salt, as you’d have to do discovery on the current state to determine use case and future state design. Nothing is in stone. But I imagine you get the point.
What is Virtualization Phase 2?
So back to my original point. Has VMware finally peaked? Not from my vantage point – and certainly not from the perspective of those I consult with. While it is definitely true that a large quantity of “low hanging fruit” or “the easy stuff to virtualize” has been taken care of in Virtualization Phase 1, there is still an entire wave of workloads that just a few years ago were untouchable. Phase 2 is all about the adventures in Virtual Business Critical Applications and the partners who can assist (note: there aren’t many listed … yet). The big workloads that generate revenue or are relied upon by organizations to keep critical functions working. Think along the lines of hospitals and infrastructure operations (water, sewer, power, gas, etc.).
Recommended Resources
The good thing is that this is an exciting period to be in. Rather than focusing on that boring consolidation ratio statistic, we’re focusing on ways to deliver higher quality of service via virtualization for the real meaty stuff. I typically think of my colleague Michael Webster, VCDX #066, as one of the more public minds on this topic, as he is a well respected expert on VBCA, specifically around Oracle workloads. You should definitely add him to your RSS feed and check out his blog.
Another suggestion would be Matt Liebowitz over at “The Lowercase W” who has been a great fountain of knowledge on successful VBCA tips for SQL and Exchange. He also hates it when you spell VMware as VMWare (as do I).
Don’t Forget Our Buddy, the Mainframe
And let’s not leave the spotlight off our old pal – big iron mainframes and legacy systems. Those are fully ripe for the pickin’ as they start entering operational durations where they could legally drink in the US (assuming they were alive, of course). With advances in x86 systems and hypervisors, the bold and underlined reason to keep these workload tethered to unique hardware is starting to evaporate. I’ve actually worked with quite a few folks who are in the process of migrating off a mainframe and onto Linux via x86. Again, exciting times.
To quote from Spaceballs – When will then be now?
Not to say that wide spread VBCA adoption will happen over night, nor will it permeate all facets of every business, but it is becoming more commonplace to see and work with.
Thoughts
Phase 2 will typically take an organization from 30% to 70% virtualized, depending on the application structure. Obviously there are companies today that are 100% virtualized (or close to it) due to the way they’ve embraced SaaS or converged infrastructure in their data center.
These next few years should be interesting for virtualization, and I don’t mean just VMware. IT organizations have a choice of where to take their VBCA workloads – continue with physical, migrate to virtual, or go further and stick it in a cloud built for these exact workloads (such as via Virtustream)? As long as pure workload consolidation is off the table (usually a losing game for VBCA), I can see a good chunk of these types of apps getting virtualized to realize other benefits over the next 3-5 years.
Your thoughts?
Great article, Chris!
I think it is important for people to realize that tackling Business Critical Applications requires a high degree of organizational/operational maturity. Just as bringing applications into virtual infrastructure for the first time required careful planning and preparation to ensure that appropriate levels of *performance* could be provided, BCA’s have an additional requirement for uptime.
Typically, these workloads have SLAs at or exceeding 5 “nines” and are consequently subject to more stringent change control, testing, and validation processes than I commonly encounter in environments that currently handle Windows or Linux (x86) virtualization. Those workloads are a little more forgiving of the middle-of-the-day reboot than, say, an application that handles bank teller access to customer accounts, balances and transactions.
As administrators of the virtual infrastructure, we need to be cognizant of these facts and determine the best ways to provide the appropriate levels of service to these new workloads. We have the additional challenge in that migration from mainframe or traditional UNIX platforms to x86 virtualization brings with it an OS replatforming onto Linux (or Windows?). While ESX has a reputation for “mainframe-like” availability, the differences between AIX or HP-UX and Linux are still rather great and there is typically somewhat of a learning curve for existing UNIX administrators. I’m not saying Linux is bad; it’s just different.
Great article Chris. Unix to VMware (Linux or Windows) and VBCA has been my life since 2007. In this new phase of virtualization consolidation ration is probably the least important metric. The name of the game is resource efficiencies under blended peak workloads without any negative impacts to SLA’s and guaranteeing and improving quality of service, while at the same time automating and simplifying management, business continuity and security. Like Doug mentioned the change controls and management disciplines might be different to your traditional Windows and Linux workloads. This drives certain architectural and operational decisions, and can impact how the virutalization team does things. Operational readiness is critical. Linux is different, but a lot of Unix admins have been needing to use Linux for a while and the transition over isn’t that much of a jump. Virtualization phase 2 will be very exciting and it is a massive opportunity for customers, partners and VMware.
Nice article, Chris. Most people having issues with virtualizing BC apps usually have treated them as if they were any other VM. It’s important to insure sufficient resources to BC apps. Customers might not see the same VM density on hosts running BC apps than they see elsewhere, but the benefits of virtualizing those workloads more than pay for themselves in the long run. Virtualizing Business Critical applications can result in better uptime, recoverability and portability while delivering performance as good as, and in some cases better, than physical servers. However, if not done properly, it can lead to ongoing woes for a SysAdmin.
As the others have said – great article Chris. I think the time has come for VBCA more so than in the past. The applications are changing (AAGs in SQL 2012 are a great example as are DAGs in Exchange 2010/2013), the vSphere platform is maturing, and storage related products that can use flash/SSD to accelerate reads/writes are helping to ease performance concerns.
I think folks always assume that all business critical applications are fully utilized on all of their processor cores, consume lots of memory, etc. There is nothing that says a business critical application has to be a heavy consumer of resources. That’s where, as Doug, Michael, and Tim have all said, the other key considerations and benefits come into play. Consolidation ratios aren’t important – providing better up time, better availability, etc., all come into play. It isn’t always about performance, though it often is for workloads like SQL.
Virtualizing the app is really just one part of the VBCA story. The next part is having all of the other cloud management/monitoring/automation pieces in place to help ensure performance, monitor for errors, ongoing sizing, etc. That’s where there is also a ton of opportunity in the VBCA space.
I’m totally with you – the next 2-3 years are going to be very busy with organizations virtualizing these applications. I can’t wait!
Matt
Appreciate everyone’s feedback on this – great to get other expert opinions as well. It sounds like a general consensus of agreement, although level of virtualization penetration seems to vary greatly among different geographical locations.