OK, I admit it, I'm a big fan of the Vblock concept, and -- by extension -- UIM, the Unified Infrastructure Manager provided by EMC.
I think that anyone who's looked closely at IT-as-a-service concepts realize that rampant standardization coupled with intelligent automation is where the real action is; the hardware simply is an enabler.
So, it was with great interest that I got a preview a while back on what was coming soon with the next major release of UIM.
The Vblock Proposition
Historical IT practice has been to hand-select individual infrastructure components (server, network, storage, hypervisor, etc.), laboriously integrate them, and then have to deal with an interesting bowl of support spaghetti: patches and updates from all the vendors, ping-ponging between support organizations to fix problems, and the like.
Far too much time gets spent running IT, far too little time gets spent actually using it :)
The Vblock platform turns that proposition on its head: here's pre-qualified and pre-integrated elastic private-cloud-in-a-box from best-of-breed vendors. You buy it like a single product, you deploy it like a single product, and it gets supported like a single product.
Configurations are tightly controlled to maximize benefits and minimize support issues. That flies in the face of the traditional "hey, I should be able to choose any darn thing I want to" perspective of many traditional IT people.
IT groups can now spend more time helping their internal customers to use IT to get work done, vs. mired in a morass of poorly cooperating technical widgets. From a more senior perspective, I believe that one of the forcing functions for considering a Vblock-like approach appears to be IT bow having to compete with external service providers.
If you're going to be delivering attractive and competitive services to your internal customers, you're going to start having to focus on how IT gets used, vs. obsessing endlessly on how it gets built.
The great news is that customer and partner experiences so far have been mostly exceptionally positive: the promised benefits are there, and now more people want to do more with the Vblock platform concept.
And -- best of all -- lots and lots of repeat orders from happy customers and SP partners :)
The UIM Proposition
UIM (sourced from EMC) is actually two products: UIM/P (for provisioning) and UIM/O (for operations).
The provisioning side is almost mandatory; it's not really a Vblock platform unless it has an integrated provisioning capability. While -- technically speaking -- there are some third party alternatives, the VCE folks discourage this as much as humanly possible. Going non-UIM for provisioning makes it far harder to get to the Vblock platform value proposition, as we've learned.
By comparison, the operations module (UIM/O) is strictly optional. Though, with its northbound information feed, it is quickly becoming the Vblock platform domain manager which enterprise management system are integrating with to manage this new environment. More about “why this is" in a moment.
Both products are being heavily featured as previews from EMC and VCE in at VM World in Copenhagen. And I think both products have a few compelling new features now that are worth understanding.
Elastic Provisioning in UIM/P
From a "southbound" perspective, UIM/P has to keep up with all the new goodies from the parent companies: EMC, Cisco and VMware. And, as expected, there's good support for all the underlying new stuff. But that's sort of what you'd expect, right?
And, previously, I hadn't had a strong appreciation for just how useful it was for an integrated management environment to have a predictable configuration underneath. Things like lengthy discovery tasks are a thing of the past, for example. And the gap between what you're running and what the tool is really smart about is much more manageable in this world.
What especially jumped out at me in this preview was the new elastic provisioning feature.
Imagine you've got a cluster or group of VMs doing something important. UIM/P will allow the administrator to dynamically and non-disruptively add or delete supporting infrastructure assigned to virtual machine groups, in effect expanding or contracting that cluster's capacity.
For example, imagine you had an Oracle or SQLserver cluster supporting an important application that occasionally got very busy. Adding or deleting compute and storage resources without disrupting the application would be pretty useful. Just to be clear, CPU and network can be both dynamically added and deleted in an integrated provisioning motion; storage can be added but can't yet be deleted. No, I don't know what the limitation might be here, but -- in practical experience -- storage capacity deletion is an infrequent move at best :)
Kind of cool, especially as more demanding applications are now ending up in VMware, and hence Vblock platforms.
That new feature rests on top of UIM/P's existing capabilities, including the all-important configuration validation and policy enforcement capabilities that take so much of the "ooops!" out of provisioning infrastructure resources.
New! Root Cause Analysis in UIM/O
A bit of a digression here, if you don't mind?
Way back when, EMC acquired SMARTS which did large-scale network root cause analysis. Beneath it all, the SMARTS RCA engine was very cool indeed; it would build "fingerprints" of potential error syndromes that matched real-world diagnostic errors that would show up through all sorts of alerts throughout the stack.
Using the SMARTS RCA engine, a network administrator could filter an alert storm into an authentic root cause problem almost immediately.
That same cool RCA technology has now found its way into UIM/O. Specifically, UIM/O now sports an underlying RCA engine that (a) has predefined error syndromes of potential failure modes, (b) the ability to correlate real-world alerts and alarms against those error syndromes, and (c) help the administration team pinpoint the root cause of a problem almost immediately.
Imagine someone solely responsible for monitoring and managing the infrastructure for several thousand production virtual machines. Yes, some of these larger multi-Vblock platform environments are *amazingly* lean when it comes to admin resources, so it's not that unusual these days.
Somewhere in all that plumbing, there's now an issue: a congested network port, a failed storage port, or similar. The deeper down the issue is in the infrastructure stack, the more likely you're going to get a torrent of alarms and alerts.
Quick -- what's the authentic problem, what's the impact, and -- more importantly -- what should you do about it right now?
All this critical information is correlated upwards to EMC IT Operations Intelligence, as well as existing enterprise management systems, making it relatively easy to augment the "smarts" of existing operational management frameworks.
In a nutshell, that's what the new feature of UIM/O brings to the table. And if it's anything like the SMARTS RCA echnology has shown in previous use cases, it should be pretty cool indeed :)
UIM/P and UIM/O Work Together In Interesting Ways
Tom Hayes was in my office the other day, explaining how the two work together in some rather compelling ways.
Take the whole challenge of having an up-to-date, managed, fully accurate infrastructure configurations. You're looking for that "one version of configuration truth" to power a bunch of workflows and processes.
Well, in a Vblock platform (and using UIM/P), the only (sane!) way that configurations can change is through UIM/P and associated VCE config control processes. There's no viable back door. That means that (a) Vblock platform is fully configuration managed, and (b) it's always 100% accurate.
That also means that UIM/O can always depend on a 100% accurate configuration to get its job done, which is useful for many things -- including of course real-time RCA :)
All I can do is encourage you to avail yourself of the nearest VCE representative (or EMC vSpecialist) to get a preview of what's coming out soon for UIM. It just keeps getting better and better.
And, if you happen to be a VMworld in Copenhagen this week, stop in the EMC, VCE or VMware booths to take a look.
The bottom line -- if you're struggling with large, complex and poorly managed virtualized environments, there's now a proven better way to be doing things :)