At last – at last – there’s now a ‘framework for enterprise-architecture’ that’s actually worthy of that title.
The framework is called TRAK, (The Rail Architecture Framework) originally developed in 2009 for the London Underground, but now published as an open-source project on Sourceforge:
- TRAK metamodel [download is .PDF] (and metamodel-project website on Sourceforge)
- TRAK viewpoints [download is .PDF] (and viewpoints-project website on Sourceforge), in accordance with IEEE1471 / ISO42010
- TRAK UML profile [download is .XML] (and UML-profile project website on Sourceforge)
- TRAK MDG add-in for Sparx Enterprise Architect toolset [download is .ZIP] (and MDG add-in project website on Sourceforge)
There’s more detail on TRAK’s background and the MDG add-in on the website for Eclectica Systems – the originators of TRAK – and also a brief project-description from IET (The Institution for Engineering and Technology).
The framework is based on MoDAF, the architecture framework used by Britain’s Ministry of Defence (which in turn was adapted from DoDAF, the [US] Department of Defense Architecture Framework). As the TRAK Wikipedia page explains:
The TRAK Metamodel is a simplification of the MODAF 1.2 metamodel. It has removed and redefined stereotypes and any defence-specific constructs have been removed. … Significant changes include:
- System is central to TRAK and can represent hard systems and soft systems (in MODAF 1.2 System is an artefact and part of the Physical Architecture and cannot include non-physical parts)
- TRAK can represent any type of interface / flow – information, energy or materiel
- TRAK can represent exchange characteristics associated with human resources – Organisations, Jobs and Roles
- other types of dependency and associations can be represented – physical, membership, responsibility extent
- addition of ISO/IEC 42010 concepts to represent the architectural task
- addition of consistency rules that constrain how and in what order relationships can be made
With MoDAF’s defence-specific items removed, the TRAK metamodel covers the full scope of an enterprise – for example, clear distinctions are made between information-resources, machines and people:
Although in a few places this does show its physical-engineering heritage, in effect this is a generic high-level metamodel suitable for use in enterprise-architectures for any type of organisation. (This is a very important contrast to the metamodels in TOGAF or even in Archimate [Business, Application and Technology layers], which at present are effectively usable only for IT-centric architectures in information-oriented enterprises such as banks, finance, insurance and the like.)
Note also that, as I’d argued in my post on ‘A question of Who‘, the Human Resources section of the metamodel does not include any reference to real-people. It describes what people do (Job, Role), their relational location (Organisation) and their capability (Competence), but – correctly, to my mind – it does not attempt to include people as themselves.
There’s a lot of detail in this framework: 90+ pages in the metamodel description, another hundred pages or more in the description of the viewpoints. I’ve haven’t read all of it as yet, but so far feels solid, comprehensive and, well, just right, really. (Part of that, I suppose, may be because I’ve spent so much time in engineering and logistics environments – where the information is often more about physical things than merely information-about-other-information – and in other contexts where people need to be viewed as people, rather than solely in terms of information-about-people.)
What’s missing? Well, it’s based on DoDAF and MoDAF, and hence, like them, it only covers the information-description side of an enterprise-architecture framework: it doesn’t provide any methodology. One option for a methodology would be to use the TOGAF ADM, as described in an Open Group white-paper by Terry Blevins et al on bridging TOGAF and DoDAF: the catch there is that the ADM’s structure is inherently IT-centric, which would rather work against the whole point of TRAK as covering a broader scope. Another (and arguably better) alternative would be to combine that with the modified version of the ADM that I use in my own architecture-work, because – like TRAK – it’s explicitly designed to cover any architecture scope. And another good option – as Craig in fact suggests in his Tweet – would be to link it to the methodology in PEAF, Kevin Smith’s ‘Pragmatic Enterprise Architecture Framework’.
Anyway, definitely worth a detailed look. Even from a fairly cursory review to date, my own strong opinion is that for TOGAF-type architecture-developments that could touch any space beyond IT, we should all standardise on something like this as a base-metamodel, rather than the as-yet unusably-incomplete one provided in the TOGAF 9 specification.
Recommended – very recommended.
[Update: 15Aug10] Philip Allega of Gartner tells me via Twitter that even though it was only developed and released last year, TRAK is already considered obsolete: “EA leads at TfL [Transport for London] and LU [London Underground] presented at [Gartner's] London EA Summit, noting that TRAK retired” – kind of embarrassing, considering how much I’ve gushed about it above… I still feel it’s of real value, though, because it provides a good example of what’s needed in an enterprise-architecture metamodel once we finally break out of the IT-centric deadlock. No information yet as to what (if anything) has replaced TRAK, but will post another update here as soon as I found out.
[Update 2: 15aug10] And have now had a detailed set of replies from Nic Plum of Eclectica Systems (see Comments section below) explaining that TRAK definitely is very much still in use, and also does have a methodology derived from ISO 42101. More details from Nic below, anyway. (Many thanks.)