For the next barcampESM session, with some a customers wish-list laid out, we rolled into the next session from Zenoss‘ Erik Dahl outlining the need for an “open agent.” Here, what we mean is a piece of software and standards that would be shared by all of the IT management vendors, to varying degrees of course. As mentioned above, there’s not much sharing, there’s even dozens of ways to monitor all that IT. It’d be kind of like if there were 12 different ways to check how much oil your car had instead of the universal system of using a dip-stick.
The group spent a long, long time trying to figure out what was up with the standards that the DMTF has been putting up over the years. Ostensibly, the DMTF has been creating specifications for this shared knowledge and work. But, as Erik said, and many people agreed with, people just weren’t using DMTF standards enough.
On a related note, I’ve actually discussed the DMTF with Winston Bumpus recently:
And see the discussion on the topic in the IT Management Podcast #02.
There was much discussion of OpenPegasus, and open source agent for DMTF standards (CIM). Written in C++, OpenPegasus probably appeals to hardware makers, but probably not too much to the folks writing IT management tools where C/C++ isn’t usually invited to the dance.
Here’s What Needs to Happens
From the intense interests participants had in the DMTF’s work, and the interest of others recently, it seems like the DMTF has a window to start helping and getting the help of the IT Management open source world. The folks at barcampESM really, really wanted something – almost anything – to serve as the independent standard for IT management models and protocols. But, for the usual standards bodies reasons, the DMTF wasn’t doing it for them.
Interestingly, there was no mention of Eclipse COSMOS either, and only faint mention of the CMDBf (now at the DMTF, it should be noted). It should be noted that there’s some interesting looking stuff going on between COSMOS and CMDBf. But, not to detract from those efforts, what most people I talked with at barcampESM wanted from CMDB efforts was a schema. They didn’t want federation between proprietary schemas of how IT data (CI’s, for you super-nerds out there) are stored in each CMDB, they wanted the one schema. If things are going that way, but the barcampESM folks haven’t gotten the memo.
From where I sit, it doesn’t seem like these standards efforts have trickled down from the architecture levels they tend to be born and live in. The people writing the code, esp. in the open source world, aren’t getting involved.
In follow-up conversations with various people, I mentioned that Microsoft was going nuts over the general idea of an IT management modeling specification with their standardizing of SDM to SML. Also, it should be noted, the DMTF’s CIM found wild success in Redmond: it’d be interesting to track how CIM evolved to SDM and is (or is not) evolving into SML.
The issue is that, as always, we’ve got several islands of standardization going on. The “scrappy” open source guys – mostly people not working at the Big 4 or Microsoft – are either ill-informed or are otherwise not involved. Adding in the open source crowd is not only “the right thing to do,” but would bring all the benefits of rapid development, innovation, and reality checking that open source brings to other areas. Look at what open source has done for the Java world: Java wouldn’t be what it is today without Eclipse, Apache, Spring, and all the other open source efforts that keep Java, as a community, from getting fat and happy.
The standards efforts – God bless them everyone – in the IT management world need that same benefit. Sure, and I’ll be honest here, the open source people can cause a lot of “problems” and be damned annoying: just look at the muted wrastlin’ Sun is doing with OSGi. But, from a user’s and customer’s perspective the end-result are good.
Now, not to point the finger at just one side: the open source must take on the responsibility of actually participating and implementing these standards, protocols, etc. They have to resist the urge to conceptually fork from the “industry standards.” Otherwise, we’re just back where we started. Dare I say “compromise,” coupled with the even worse “boring”?
Reaching back to the Java world: sure, it’s a dysfunctional ecosystem made up of absurd turf battles, the need to make money, and pressures from hype-driven competing technologies. It’s like any large family. But, at the end of the day, the Java world gets the job done like a well maintained highway system: it ain’t sexy, and you might end up in a traffic jam now and then, but you’ll get from point A to point B, safely and cheaply. It’d be great to see that same boring effectiveness in IT Management, and I’m dead-sure the actual users and customers would love it, if we can just make it for them.
Disclaimer: Zenoss is a client, as are Eclipse and Microsoft.