In an interesting report from the Cambridge MPhil for Public Policy student group, on the largest IT project failure ever, so far, a lot of historical detail is dredged up, and a number of useful conclusions drawn under headings like Haste, Design, Culture & Skills, which are pretty hard to disagree with.
Some stuff goes unsaid, but reading between the lines, one might ask
1. were no "references taken up" for the people who were shortlisted for the contract? I mean I know BT pretty well, and I'd never hire them for a systems integration job like this - sure, for the spine/N2 (which worked fine, and afaik, were loosely modelled on SuperJANET), but not to bring together an unknown outfit with a dodgy sounding name (iSoft - wannabe Microsoft:) and Fujitsu (ok, make decent clone PCs once then, but large databases etc? Not really).
2. So I realize it was a while ago, but even then, we had rapid growth internet outfits with many services (like google and apple and onlone gaming) who had built customer relation databases (with ok security) for 100s Millions of users, which managed a lot of stuff that scaled out, and could address the srts of needs the NHS requirements said (oh, ok - the requirement capture was one of the biggest failings in the whole thing, of course) - one thing that open source, internet/cloud companies do is to stay agile, so they can interface to legacy systems (as they get big enough to acquire them - e.g. google buying youtube or the maps system, or microsoft buying skype) - in the same way, rather tan imposing a central design on an heterogeneous set of health services, interfacing between them woould have been just fine...
3. How bad was Lorenzo? Didn't anyone do a code inspection ever? It sounds a bit like the Obamacare sign-on system fiasco - but why? I mean lashing up a bunch of patient recor databases into mongodb, mysql or whatever (let alone Oracle or Ingres or some big iron dbase) is not rocket science. Its done by Internet startups 100s of times a year and is hardly ever the reason they fail.
Of course, the other component that work (Picture Archiving / Comms) worked because there is much less heterogeneity in those systems - image standards exist and you just need web/email access to glue systems together....so lucky no-one managed to balls up that part, eh!
Anyhow, nice report, but would be nice to hear more about the contract failures (lack of penalty clauses with enough teeth)....
Some stuff goes unsaid, but reading between the lines, one might ask
1. were no "references taken up" for the people who were shortlisted for the contract? I mean I know BT pretty well, and I'd never hire them for a systems integration job like this - sure, for the spine/N2 (which worked fine, and afaik, were loosely modelled on SuperJANET), but not to bring together an unknown outfit with a dodgy sounding name (iSoft - wannabe Microsoft:) and Fujitsu (ok, make decent clone PCs once then, but large databases etc? Not really).
2. So I realize it was a while ago, but even then, we had rapid growth internet outfits with many services (like google and apple and onlone gaming) who had built customer relation databases (with ok security) for 100s Millions of users, which managed a lot of stuff that scaled out, and could address the srts of needs the NHS requirements said (oh, ok - the requirement capture was one of the biggest failings in the whole thing, of course) - one thing that open source, internet/cloud companies do is to stay agile, so they can interface to legacy systems (as they get big enough to acquire them - e.g. google buying youtube or the maps system, or microsoft buying skype) - in the same way, rather tan imposing a central design on an heterogeneous set of health services, interfacing between them woould have been just fine...
3. How bad was Lorenzo? Didn't anyone do a code inspection ever? It sounds a bit like the Obamacare sign-on system fiasco - but why? I mean lashing up a bunch of patient recor databases into mongodb, mysql or whatever (let alone Oracle or Ingres or some big iron dbase) is not rocket science. Its done by Internet startups 100s of times a year and is hardly ever the reason they fail.
Of course, the other component that work (Picture Archiving / Comms) worked because there is much less heterogeneity in those systems - image standards exist and you just need web/email access to glue systems together....so lucky no-one managed to balls up that part, eh!
Anyhow, nice report, but would be nice to hear more about the contract failures (lack of penalty clauses with enough teeth)....
No comments:
Post a Comment