One of our clients, SourceLabs, announced a very interesting system today. I want to narrow down on the systems management aspect of it, which is just one part of the broader Continuous Support Solution diagramed above:
SourceLabs' Continuous Support System features what Sebastian calls "adaptive diagnostic probes" that are fully integrated and configured for customer environments. The probes identify production issues and begin to gather diagnostic information to help get to the root of the problem, he said. Indeed, the probes can be configured so that as soon as a problem occurs, the SourceLabs support team extracts system information to find and resolve the problem. And the system includes a database of more than 200,000 signatures of problems that might occur, Sebastian said.
Steve and I were pre-briefed on this last week, and it was very intreging. It's another approach to what I call Collaborative Systems Management. As Steve says in the above article:
The most interesting aspect of the SourceLabs announcement to me is the way it leverages a literally immense of amount of information that is currently scattered around the Internet, putting it to work for engineers.
Crowdsourcing Sysadmins
This is a crowdsourcing approach to problem solving in systems management. It's handy to think of how the idea works by analogizing to fingerprints:
A problem occurs on your system.
The "adaptive diagnostic probe" takes a fingerprint of that problem.
The fingerprint is sent up to SourceLab's fingerprint database.
SourceLabs compares that fingerprint to all the other fingerprints in the database, trying to find past occurrences of the problem.
If it finds a match, then it can bring in the "profile" of that problem, including how to fix it, what other people are saying, etc.
Once you identifying the problem you know how to fix it. What's exciting here is the idea that systems management is spanning silos and firewalls. That's may seem small, but you can compare it's potential impact to using Google instead of your own bookshelf. That is, collaborative systems management enables Metcalfe's Law where it wasn't possible before. The more general term for this line of thinking is crowdsourcing or the longer moniker, "the wisdom of crowds." See Anne for another good tangent off the crowd-* idea.
As with all things that have hints of AI -- even if they're Mechanical Turks -- milage may vary. Usually, the output of such systems isn't a complete solution, but more so advice for the people who end-up do the real work.
But, as anyone who's admin'ed a machine or network knows, the more targeted information during problem solving, the better.
Open Source
While it may be obvious to the more enlightened out there, it's probably worth pointing out that open source enables all of this. If the software wasn't open, SourceLabs would have had a much more difficult time creating it's probes. More importantly, due to Conway's Law, open source software tends to be highly modular and loosely coupling -- matching the geo-dispursed and time-shifted organization that created it. Such a system us much easier to instrument than a mono-stack...even for people in the organization.
Partnering
In talking with SourceLabs last week, they're not interested in becoming a systems management (or applications management, whatever you want to call it) company. They want to remain an open source support company (my phrase for it), which is still a great idea in my book.
As such, the door is open for other people and organizations to work with them. As I told SourceLabs, I was taken off guard by the Splunk/CA partnership and it's opened my mind to other such Big 4 + Spunky Youngster combos.
Identify
Interestingly, BMC's Identify acquisition comes to mind as well. I finally got a chance to see Identify's AppSight in action at TechEd 2006, and though it was a booth-demo, it still looked quite useful. AppSight does similar, though not quite the same, analysis and "stack dumping" of problems at runtime. To be fair, the booth-man was quick to correct me when I said, "so it instruments code at run-time," expanding AppSight's function out much wider.
Interop Standards
In the rainbows and sandals world, I'd envision the SourceLabs "adaptive diagnostic probes" and BMC/Identify working together or, at least, standardizing the way they package up and provide access to the underlying data: an ATOM (feeds/data and REST APIs over that data) for systems management.
Niche vs. Gorrilas
Why all this fuss about partnering and matching up? In the software world of today, where small is better, you don't want to bite off more than you can chew. Compare Vista to Eclipse. It's better to niche and collaborate than to gorilla and lock-out.
Splunk, of course, is an obvious match-up.
Of course, that's just are just two pairing. If SourceLabs' probes and the "fingerprint" database backing it prove to be an effective approach to collaborative systems management, any systems management vendor or project would be wise to look into it.
Disclaimer: as mentioned, SourceLabs is a client, as is BMC.
Previous Topic
