Irrational exuberance in the IT industry: CMDB is going nuts

Everybody is piling into the CMDB frenzy now, including dodgy journos and madly re-inventing software companies... and this blog. ITIL may not be a fad but the IT Skeptic thinks CMDB now is one.

I can't believe someone from Gartner said this:

Companies — with the exception of BMC Software, which has been in the market for a few years — began offering commercial CMDB solutions only about six months ago, said Ronni Colville, a vice president at IT research firm Gartner

They may speculate wildly, they may pump up markets, but normally Gartner have their facts pretty straight. I suspect this was a mis-quote. The same article goes on to say

Organizations can invest in CMDB software from companies such as IBM, BMC, Hewlett-Packard, CA or Managed Objects to get started.

which either means that all those others have piled into the market in the last few months or the reporter has a little to learn about both CMDB and facts checking. But it fits the pattern: everyone is sniffing around CMDB, even those who know very little about it. The very title of the article suggests someone who hasn't a firm grasp:

A prescription for preventing service outages
Configuration management databases can alert managers to dangerous interactions

and this one later on

CMDBs are like the software that alerts pharmacists to dangerous drug interactions.

and, like, you know, no way.

The media keeps pumping ITIL but the public is losing interest, according to one of my favourite barometers. Not so CMDB which is definitely the hot topic.

Some would have us believe this is because the advent of virtualisation is the tipping point driving everyone to discover CMDB. That draws a long bow. If there is any tipping point it is of course the take-off of ITIL in the USA, but even more I think the CMDB hype wave has reached critical mass and is now exploding. The vendor/analyst machine is well experienced in whipping up these frenzies.

I had hoped we got wise to their games after Y2K but IT seems hopelessly vulnerable to promises of silver-bullet technology solutions to process problems. To create solutions, think of people, process and technology in equal parts and in that order. Read the blog entry predicting top data centre trends for 2007 and the associated 10 top stories for 2006 for a beautiful example of technology-centric thinking. Remember my crude linguistic test for process orientation or technology orientation: do they talk about verbs/actions or nouns/things? ITIL is mentioned only in the context of metrics; disaster recovery is fixed by tape transport; big iron; liquid cooling; site selection etc etc etc

So IT gets more complex and unstable every day and the solution is not to look at how we do things and the quality and culture of the people doing them. Oh no, it is to introduce yet more technology: federating, synchronising, reconciling, singing, dancing CMDB.

The vendors are piling in. The established ones are spending millions to re-invent, graft on or acquire anything that can be labelled a CMDB. CMDB startups sprout like mushrooms.

For a moment there I thought there was hope when the first article mentioned above asked "What comes first — setting up a CMDB tool or defining processes for configuration management?". But look at the answers!

“A lot of people get enamored with starting with the tool first,” Lithgo said, adding that “it’s a big mistake.” [Good! Good!] Irvine said he agreed that people can easily misuse the tool. “A CMDB is thought of as something technical when in many ways it shouldn’t be,” he said. [great!] “It should be treated as more of a services-based project that identifies first and foremost all the services you want to map in the CMDB.” [Nooooooo we're back to things! What about the Configuration Management processes?? Shouldn't we start with them first??]
A service-oriented approach is the way to proceed, said Andy Atencio, manager of information and technology for Greenwood Village, Colo. The city will be using a hosted CMDB to supplement its ITIL best practices. [So service-oriented means paying for someone elses CMDB-thing instead of having your own thing] If a technology component isn’t directly connected to the provision of a specific IT service or business process, it doesn’t warrant being tracked in the CMDB, Atencio said. “You have to be smart because otherwise you are tracking everything,” he said. [Things] Other experts offer similar advice. “The big key is process first, and let the tool serve the process,” Lithgo said. [ooh yeah, looking good...] “Historically, IT organizations have been about heroics, secret knowledge and the select few. That’s hard to give up,” he said. But all secrets about how things work must be shared via the CMDB [oh no, we're back to processes as things to be stored!!] if the organization expects to achieve repeatable processes and replace anecdotes [Things] from the IT trenches with tangible metrics [Things] , he added.

Is it just me? Am I being over-sensitive? or are the vast majority of IT people incapable of thinking in terms of process and people as well as stuff/things/artifacts/technology?

[Deep breath] Aaaanyways, they seem to get us all worked up over things like CMDB in ways they can't over far more useful concepts like ITIL (though they tried). ITIL is showing steady growth commensurate with sensible adoption. CMDB is going nuts. Irrational exuberance ends in lost money.


A bit over the top

Okay, Skeptic, you are serving up softballs again.

Ms. Colville knows whereof she speaks. It's not a misquote. She's been very consistent on this message that BMC got to the game first. I agree with her on this. (And I have no financial interest in BMC.) The issue of course is that she has re-defined CMDB to suit her analytical purposes, a re-definition that I've found useful. (I won't risk ticking Gartner off by enumerating her criteria; you'll have to pay them their pound of flesh for that.) I agree that IBM, Hewlett-Packard, CA and Managed Objects are all either partial solutions or have made recent acquisitions.

I find the virtualization argument compelling. Part of the problem is that you are confusing the origination of requirements out of the technological management domain, with "technology-centric thinking." One needs to think very carefully and precisely when applying technology to solve technology's problems. I think of it this way:

The advent of virtualization
has created a REQUIREMENT
for improved impact analysis
due to the re-centralization of processing
upon common computing platforms.

necessitates improved dependency mapping between IT components
to support the necessary analytics.

Dependencies are INFORMATION
which require PROCESS to create and maintain.

Unlike many others, I don't always bow down before the altar of process as the first and foremost concern. Usually, formal process is implemented so that some service is of high quality. Often, that service is one of "accurate information on demand." That is the true goal. The process is never an end in itself. In my world, often the data comes first.

Focusing on nouns may not be indicative of a technical bias - but rather a bias towards information and its underpinning data - the flip side of the process coin. As a member of DAMA, I'll arm wrestle process bigots any day. What was the value-added end state of your process? Updated, accurate data available for some other purpose? I rest my case...

I actually have concerns about thinking about Configuration Management as a process. By accepted BPM theory, it is not. It is a capability, service or function (take your pick, depending on your personal ontology). See

(But Charlie, you can't go redefining what Configuration Management is...!)

Watch me. If ITIL can call John Zachman a "Configuration Management guru" (much to his bemusement), then I can commit the heresy of questioning whether we have a true process here at all. What is a Configuration and when are you done managing it? Can you count the number you have managed today?

Whoever came up with the holy shibboleths of Configuration Management "process" needs to be hung out to dry. (Identify, Control, Account/Report, Audit). I don't care if these are MIL-SPEC things in existence for forty years, with long and distinguished aerospace lineages. I say they are at best showing significant signs of age and irrelevance - or at least, non-value-add obviousness.

The so-called "activity" of Configuration Status Accounting in particular I find ridiculous. It boils down to "maintain the configuration data, and make it available on demand for a variety of purposes." That is NOT a process. It is a service, and a rather obvious one at that, implicit in any decent data management architecture. "Control" is likewise incoherent. It is a broad and obvious requirement, not an activity or process.

Most IT systems have some form of reporting. Most systems are concerned with data integrity. Tracking the status of entities throughout a lifecycle is a core analysis exercise for any data-centric system. Again, just obvious to any competent architect.

The interesting activities to my mind are identification and verification. They are where I find I spend the majority of my time, because CIs may have type-specific workflows, e.g.:

Establish Application Service
Provision Server
Configure Database

and the corresponding verifications/audits of those. But these workflows are so heterogeneous that shoehorning them into some totalizing "Identify CI" or "Audit CI" process I have found to be an academic exercise at best.

I think I'm actually more skeptical than you in some ways; you are doubting the tool, while I have more concerns around the so-called "process." But I can harbor that doubt, while also believing that CMDBs are an essential tool. Are they going to ride the hype cycle? Of course, and with a vengeance. The trough will look something like the crash and burn of metadata repositories 20 years ago.

What has been is what will be,
and what has been done is what will be done;
there is nothing new under the sun.
(Ecc 1:9)

Do I care? No. The value proposition is plain at least in my world.


Charles T. Betz

a trifle ... er ... kind to BMC I feel

One other thought. the wording "began offering commercial CMDB solutions only about six months ago" is a trifle ... er ... kind to BMC I feel. the implication is everyone else is johnny-come-lately.

To slate all the other vendors' solutions as not being CMDBs because they don't meet a particular standard for what is, is like saying Toyota only entered the car market once they offered the Lexus.

there is a difference between being a player and having an excellent solution. In fact in IT there is a difference between being successful and having an excellent solution too. But that is another blog topic...

occasionally I get a bit off the edge

You have copped a bit of flak on this site Charlie but I really think you are a great thinker in this intersection between operations and architecture. You always punish me when I go out there :-)

We have agreed in the past that CM should not really be a process but rather a function like Service Desk. Being immersed in ITIL all day, and being a less rigorous thinker, I still slip back into treating it as process though. Oops.

I think the data/process, technology/process, objects/activities, nouns/verbs arguements are like the nature/nurture one: the reality is somewhere in the middle as both are important. But I believe the culture of IT as the first decade of the millenium heads to a close [already!] is off-balance, object-centric, so this gadfly hovers on the other end of the spectrum to try to restore some balance. And occasionally I get a bit off the edge. But if I do I can count on Charlie and a few others to nudge me back ;-)

It is a bit like Greenpeace. I cancelled my membership years ago because as a skeptic I could no longer tolerate their misinformation and abuse of science, yet I value their existence and acknowledge the necessity of what they do. I wouldn't have them moderate their stance. Democracy is the balance of opposites, and if nobody is on that end of the seesaw it don't work.

So, coming back to earth, your points are good.

I have an open mind on BMC having such a lead on the CMDB market. It is news to me, and I welcome views from other readers on the matter, especially those familiar with some of the niche tools like Infra and Marval. I'd also have thought that all the work CA did a couple of years ago may not be perfect but would have got a C pass.

I still can't see virtualisation as a tipping point for CMDB frenzy. It simplifies configuration by centralising servers under better management tools with much higher visibility. A friend is consolidating many hundreds of physical boxes onto virtual machines: the first challenge for the metal servers was just knowing they were there, and exactly where. Even if it is yet another confounding factor, I can't see it as a bigger one than many of the others we have wrestled with for years: server pools, dynamic everything, n-tier architectures, mobile devices, end-user computing...

Just to add another dimension to it I have lately come to see People as the first priority, then process and technology/data. IT is starting to get its process house in order, but we still have no idea about the management, development and assessment of people, nor of dealing with cultural issues in change. But yes, we can argue the process/data thing for ever.

Be warned about CMDB: IT'IL end in tears!

Be warned about CMDB: IT'IL end in tears!

Syndicate content