Virtualization Management Communities 2.0

It’s going to be a good year for virtualization administrators. They’ve spent the last few years consolidating servers, ramping up utilization, and slashing the time it takes to deliver new environments to users. Often, they’ve done it without much corporate support as both pioneers and evangelists. Much of their success has been due to the strong community they’ve built at places like the VMware Technology Network, where any number of peers is standing by to help resolve a complex problem or share a favorite script, just when you need it.

In the early years of the virtualization technology wave, there was a lot of talk about automation, and quite a few platforms were developed then quickly snapped up by the systems management heavyweights. Few of these platforms did much of anything out of the box – they were policy-based, but where were the policies? They were containers without content, and content always wins.

Over the past month, I’ve been excited to see the second generation of virtualization communities coming to life. Each of them is building on the VMTN model in its own way, leveraging the collective experience of the field to create intelligence organically. A few worth checking out:

vKernel’s SearchMyVM is a free download utility – delivered as a virtual appliance – that quickly indexes an entire VMware environment and fronts it with a Google-style search portal, complete with pre-built searches as well as a query builder. Think of it as a mini Splunk for virtual machines;

Vizioncore’s Virtualization EcoShell Initiative is a community portal – accessed through a freeware desktop app – for those who use Windows PowerShell to manage virtualization. Members can improve their PowerShell skills, or share and debug their own scripts;

TripWire’s vWire community leverages the vendor’s configuration management and security expertise, offering free Windows downloads to verify whether vMotion is working properly, for example, and whether your virtual machines pass VMware’s Security Hardening Guidelines.

The market will decide which communities will thrive, but I like the aggressive approach we’re seeing in this space: race to market with a free utility that solves a real problem, cuts a few key strokes, eliminates a manual job, or teaches something new. The challenge will be to invest enough energy and resources to keep the content coming and build strong ties between community members.

The changing data center landscape

Recently, Taneja Group published what has become an annual report reviewing the state of InfiniBand in mainstream IT. Once again, the landscape has evolved in interesting ways this year, with the virtual infrastructure and cloud computing being a couple of the forces that are driving InfiniBand adoption in the enterprise data center. Long story short, InfiniBand has proved itself a capable platform for continued evolution, and vendors with products in this space have long ago figured out how to make the fabric into a platform. While us bleeding edge technologists speculate about what infrastructure as a service is going to look like, the most common names in InfiniBand have long ago turned the infrastructure fabric into a service enabled platform. Slightly different twists, but you need a service enabled platform behind your infrastructure as a service, and with a service enabled platform you can turn your own infrastructure into a well managed service, with granular and comprehensive management. I'll illustrate this in more depth, but first a link.

The Taneja Group InfiniBand publication is available on the InfiniBand Trade Association website - www.infinibandta.org - and will be up for download via the Taneja Group website soon.

Now back to the story - don't stop at this report and think that you have all of the story. How is InfiniBand service enabled? It is a bigger picture than just the mechanics of InfiniBand switching and how data is transferred to host processes. InfiniBand has been engineered for extensibility and can in turn be a platform for innovation. Take for example Voltaire. I've recently been given a tour of the Voltaire Unified Fabric Manager (UFM) solution. UFM builds on the architecture of Voltaire switches in order to extend their capabilities with an even more intelligent management layer. That management layer can provide more intelligent routing in a layer above the fabric, while integrating with and leveraging core fabric routing and management technologies. More importantly, UFM can dive deep into the fabric to give real insight into total infrastructure activities and performance. So far, I haven't seen any other solutions claiming to be a "fabric manager" offer the sophisticated insight, resource management, performance trending, and core fabric function extension that UFM can. UFM is just one example, but it fully illustrates what a well architected fabric should be capable of. The fabric shouldn't be an invisible lower layer of connectivity, managed within a separate operational domain. The fabric should be integrated with all aspects of your infrastructure.

My take, all the hubub about new emerging fabrics is earning the ear of the enterprise customer. Whether those fabrics can deliver as a platform for extending enterprise computing capabilities will be judged in its own time. Meanwhile, these conversations are opening doors for new opportunities, and InfiniBand is poised to deliver, and with the door open, the differences between fabrics are starting to make themselves apparent. If the challenge is big enough to warrant a new approach, that's where we are finding users bringing InfiniBand into the mainstream enterprise.

Before Taking Off For the Cloud, Check Your Virtual Engines

In March, McKinsey published a discussion document, Clearing the Air on Cloud Computing, and I’ve had several colleagues and clients mention its controversial nature. In my view, the only finding that can be considered controversial is the claim that current cloud services offerings aren’t cost competitive for larger enterprises – in other words, large data center total cost of server ownership is actually less than most EC2 pricing options, for example. I can’t argue the numbers, but the cart might be getting in front of the horse.

The more salient topic covered, one I find decidedly non-controversial, is that TCO discussions are mostly premature for any but the smallest of IT shops. The first question should be: “How virtualized are you, and how’s that going?” If the enterprise has only seen modest gains in utilization, or is having trouble sharing servers among business units, or is running into tricky new performance problems with virtual servers, it doesn’t help much to let them know they do, in fact, already have an ‘internal cloud’. This type of retroactive rebranding is all the rage, but I’d encourage vendors to step back a bit.

There’s a large and growing demand in virtualized enterprises to leverage, optimize, and control the virtual estate. Virtualization's capital cost savings - consolidation and utilization - are by now well-proven. The operating cost savings? We’ve only scratched the surface. Until enterprise operations teams have greater confidence in the run-time performance of a fully virtualized environment, they won’t be ready for even a partial lift-out. Smart vendors will focus on the tricky contention and performance issues keeping virtualization teams up at night. And those are the vendors that will be trusted to help deploy into private and public clouds when the time comes.