Can we have a single unified model to describe all IT "assets", everything IT cares about? The idea is certainly appealing, and excites any IT management geek, this one included. Cote has a post where he articulates the advantages of having one standard model to model everything in IT and argues one of the big 4 open sourcing their CMDB solution may be the fastest way to get there. Can this happen? Stranger things did happen but I'd put likelyhood of this one as very low. It would certainly shake things up if a major CMDB became open source and if (this is a big if as well) the model used by it gained traction. One can only hope, not much more we can do.
Since we're developing an open source CMDB federation solution, I've been investigating and thinking about what this "one model" may look like. Is there a standard that can be used as a starting point? The big standards play in this area seems to be Common Information Model, CIM from DMTF. CIM is a mature standard and has significant adaption in the IT industry, so it certainly seems like an option. It should not be that difficult to create a parser that consumes CIM specifications and generate the model in RapidCMDB. When I started looking at the CIM specifications, my brain started to hurt. It is quite complex (sophisticated?)
As I have been thinking about a unified model for IT and the projects we've worked on in the recent years, an uncomfortable feeling started to take over: even if we had a unified model, it would not have met our needs exactly and we would have needed alternative models, that may be simpler and more finely tuned for the tasks at hand.
In the use cases that would make use of the data in CMDB (that I'm familiar with), provisioning management tools, BSM, operations troubleshooting, etc. a model like CIM is quite an overkill. CIM is probably just fine for the consumption of a software developer, but it is way too complicated for the rest of us mortals. I feel that for the model to be useful in IT, it should not be too abstract, and anyone with basic knowledge of the domain should be able to use it intuitively. CIM is not it. The diagram below is a model to be used for an network operations group. It is not standard by any means, has only a few abstract classes. Model is recognizable by the operations folks, can be implemented rapidly, and can be used for variety of purposes.
So I think we may need to be able to develop models specific to our objectives, at least for the use cases I'm familiar with. Hence in RapidCMDB project, we're leaning towards providing the tools to create models easily and leverage existing data sources and systems, and not attempting to provide "the" model for everyone to use. Sort of multiple views to the same data. If a widely accepted single model were to emerge in the future, we would certainly implement it on RapidCMDB, but until then, we're stuck with the reality anyway. What do you think? We'd love to hear your thoughts ..
I don't think one model is
I don't think one model is possible in the near term. I think that we must focus on smaller models aligned to the functions, tasks, org groups, etc. in a typical company and then extract those into the federated model as needed.
The vendors are all focusing on trying to define what's needed to support their visions of a CMDB and how they'll enable ITIL or service management. If you back up a level, what your initiative here can provide is domain specific models and "mini-CMDBs" initially and then the broader based "enterprise CMDB" at the higher level for positioning against the incumbent CMDB plays.
This would allow you to get grass roots traction with the OSS project and provide potential value quicker. If you look into my SMDB concepts (see prezo) you can hear/see some of my thoughts in this area based on what I built at a previous company. This focused on a model for a "mini-CMDB" focused on BSM and service management via all things IT management and monitoring configuration related for CI's (physical and logical).
You already know a great deal about Tivoli Netcool products (and maybe others). Focus on building a model where you can consume all of the DLA's from Tivoli (study the Tivoli CDM) to get information you need. You'll be miles ahead that way! Even creating an expanded version of the Tivoli CDM instantiated in your OSS CMDB would be light years ahead.
I'd be happy to help out in this area!
Doug
BSM/ITSM Blog: http://dougmcclure.net
Hi Doug, Thanks for the
Hi Doug,
Thanks for the insightful comment. nicely put!
domain specific models or "mini CMDBs" is our natural focus as we mostly solve problems at hand. I had watched your presentation from BarcampESM and SMDB concept strongly resonated with me.
I've seen several ITSM/BSM projects get bogged down due to lack of an SMDB type solution. RapidCMDB is partly born out of this need. It is essentially meant to be a pragmatic solution to enable higher level solutions.
I will certainly take another look at the Tivoli CDM, thanks for the pointer!
Berkay, Coming from the
Berkay,
Coming from the operations side of the house I have been looking for a "practical" CMDB solution for years. Generally, we gather the so called experts from IT development and Quality, draw some abstract diagrams, throw in some industry buzz words, make some power points for management, then some software group in India builds something that no one ever uses...
You are familiar with our environment at Hughes and how we use EM7, SMARTS, RapidInsight, APG, and Perl scripts to attempt to present information to our Operations staff. While all these tools are working pretty well, I have always thought that there "must" be a better way. SMARTS does an excellent job of correlation of network infrastructure, but the BIM tool falls short of modeling fault hierarchies. Fuzzy objects and memberships need to be part of the model which is where the CMDB becomes a key contributer.
For example, physical objects (routers, switches, hosts, etc...) are members of many communities such as the data center they reside in, the IOS version, the "last config change time, or more application specific relationships. Being able to gather and relate this information is where some easy but large gains in operations could take place. Today it is too specific to vertical applications such as our satellite systems instead of the same tools for modeling ALL the relationships.
Lastly, gleaning ad-hoc information out of the CMDB by a member of the operations staff who is not an expert is very important. All the information in th CMDB is useless if the only person who can understand it and create a report is the admin!
Looking forward to putting RapidCMDB to work!
Karl Fosburg
Hi Karl, Thanks for the
Hi Karl,
Thanks for the insight! I think we're on the same wavelength for sure, we're in search of that better way! "gleaning ad-hoc information out of the CMDB by non-expert operations staff" is one of the distinct objectives of the project. Users will be able to:
- use web browser to seamlessly browse the CMDB without having to know which data resides in what system
- search the CMDB easily, using ad-hoc and/or predefined queries
- use a simple scripting commands that simply expresses the business logic to create more sophisticated actions/queries .
RapidCMDB 0.2.1 just has been released which includes lots of bug fixes. Let's talk/chat about how we can get the ball rolling when it's convenient for you.
Thanks!
Berkay