How?
Badly and that's best we can hope for.
Nevermind, it will provide endless entertainment for us here.
Hey, it is a Government IT system. It can't be that bad can it?
"Unprecedented challenge" was the phrase which kept being repeated in a recent Treasury Select Committee hearing regarding Brexit, the customs union and HMRC. The department is already undergoing the “largest transformational change in Europe,” with plans to digitise the tax system, while simultaneously untying itself from the …
Hey, it is a Government IT system. It can't be that bad can it?
Oh yes, as it will be moved to Microsoft software, which I reckon to be a disaster of truly gargantuan proportion. I've seen it all before. As a matter of fact, I know of another country where this process has just been set in motion. True, the systems are now so aged they must be replaced, but it may be worth spending some time identifying a platform that may last the same number of years again.
With its weekly downtime for patching to stay up to date, Windows does not strike me as having anywhere near enough oompf and reliability to cope with this load without a shocking overspend on hardware. They don't exactly handle trivial volumes of data there.
Alas, we can but sit back and watch what amounts to a replay of what happened before.
Plus ça change..
No matter what OS you use, you need to restart to update kernel level code (kernel trampoline? Yeah, pretty much still unused / in it's infancy as a revamp), which is why you don't make it run off one machine.
You just put 50 machines on it. Update them at different times, reboot, and services should be ENTIRELY UNAFFECTED. Otherwise you never had any idea how to design a proper nationwide critical service.
You just put 50 machines on it. Update them at different times, reboot, and services should be ENTIRELY UNAFFECTED. Otherwise you never had any idea how to design a proper nationwide critical service.
Time to give the Redmond marketing team something to downvote, then.
That still doesn't address the problem that overload and recovery behaviour of Windows requires you to throw a LOT more hardware at the problem than with other, more "industrial" platforms. This game started quite early on when you would be required to build a multi-machine cluster to get the same stability and uptimes as a stupid single Linux box. I've seen someone successfully log into a Linux machine that was 300x overloaded due to a runaway coding error - oh yes, that reminds me, having a low resource command line to manipulate a whole system is also not something that Windows is very good at.
Want to see machines VERY hard at work? Go to CERN. Yup, they're not running Windows, I wonder why.
Because licensing costs.
I agree it's probably not the same level of performance. Trust me, I'm no Microsoft fan.
But with virtualised architecture nowadays, there's almost nothing in the way of Microsoft OS staying up long enough to allow you to do this on the same hardware. We're not talking "I put Linux on a PC", we're talking dozens of hypervisors in separate datacentres running hundreds of virtual machines. HyperV - and the High Availability clustering tools in Server 2012/2016 - are more than capable of keeping them all up, moving them around, spinning up new instances and isolating runaway processes inside them.
You want to know how you know that? Office 365. Windows Azure. Just about all the web services running on Microsoft OS.
Please stop parroting the arguments against 80's/90's Windows installs. It makes us look out of date and stupid.
Now... why does CERN and the top supercomputer not run it? Because you can't code it to optimise to get 100% out of it. You can literally alter the OS on Unix/Linux if you so desire. That's why they do it - to get every inch out of machines that are constantly at full load. Customisability. Doing things not envisioned by Microsoft with the OS. Pruning every last unwanted line of code out of it. And not paying per-core for MILLIONS of cores. That's why.
Could you do it on MS? Quite likely. It'll take a performance hit but there's no real reason you couldn't. The problem is more that it's not designed for that. There's things like HPC Server and compute-cluster versions of Windows. They don't make those just to keep them in a box at MS. People use them. And pay for them. They've been in the Top100 supercomputer lists. It's perfectly feasible and viable.
So please, stop that nonsense. Yes, if it were me, there would not be a single copy of Windows running anywhere. But to say it's not capable of the above, where Top100 has requirements LUDICROUSLY above even things like "running a national government database", is a nonsense. It's more than adequate, easy to source, easy to maintain, easy to manage etc. to do the latter task. Otherwise people who make those big-end clusters and racks full of servers would never even support it, let alone actual build, sell and use it.
To be honest, for a HMRC IT system, you could set it up today and get yourself as many 9's as you wanted, so long as you paid the price on hardware and infrastructure, not even so much the software.
Mainframe & Mainframe software is the answer here - it's been doing this kind of thing for decades, safely, robustly and fast AND runs cheaper and more reliably than server farms at these kinds of volumes.
That's getting harder and harder to do - at best you'll still be able to get people to write code and support mini's, but I have been personally involved in moving a pension management system from a mainframe to a mini because the supplier end-of-lifed the mainframe version.
It's not just hardware, it's also skills. If you're not prepared to pay for the fact that the required skills are now so rare it's easier to find unicorns you have a dev and a maintenance problem.
I personally don't know just how robust open source databases are, but from a maintenance perspective that strikes me as a better option than choosing yet another proprietary vendor who will do their level best to lock this up as soon as they can get away with it. Ive seen that game played by many a consultancy and as a taxpayer I'd say it stinks. This has as major advantage that you can shove practically any hardware underneath it and even switch if the vendor decides to play games.
Well, Linux (Oracle for years, Ubuntu since last year) has had the facilities to patch the running kernel without rebooting the system. It also has support for hotplugging even CPUs.
Windows because of it's filesystem semantics requires reboots for most updates, even the ones that do not touch the kernel. (An EXE/DLL being executed is locked, thus even an user level app needs a reboot to update these.)
OTOH, I tend to agree, to idea is to create a virtual setup in a cloud, where one system going down, is not really an issue.
Then how on earth would the Civil Service bods roll into high paid Private Sector jobs? Answer me that.
Plus the UK is completely different to everywhere else as it uses Sea Horses, Mermaids and Lions for goods transport.
The UK did not get to where it is today by replacing perfectly good IT systems every 20 years.
plans to digitise the tax system, while simultaneously untying itself from the £800m per-year mega IT contract. And that’s before Brexit.
Oh look, a Big Bang change. Undoubtedly it will produce a whole new universe - of pain. Probably for us tax payers as well.
Government IT, just say "run away".
Government IT, just say "run away".
God no, the consultancies will be on this like flies onto fresh poo.
I've seen it all before. Someone has a private discussion in which they plan yacht purchases and post-project employment, then comes in with some powerpoint to convince the other saps, confident that the job is already in the bag.
If there is any concern, one of the consultancy's management will "retire" or otherwise become accidentally available for the top job at the National Audit Commission so he/she can ensure that the next audit of the project is done by inexperienced staff who will be no match for fully briefed project consultants.
At some point, they decide that the money vs exposure equation is starting to get too hairy and the project will be allowed to fail, at which point everyone walks away with a huge chunk of tax money and a fresh job to go to (after a waiting period so it's not too blatantly obvious), leaving the tax payer with nothing to show for it. Wait a few years, repeat.
Oh, that never happens? Really? Let's just say you really have no ID..
How will companies cope.
Just pretend for a moment that there's a chance in hell of HMRC getting their systems ready in time. Big companies need to then set up their internal systems to integrate with the new HMRC system. That'll take months at least and can't be done until there's a working integration version of the HMRC system to integrate with.
We don't generally import or export bricks as the shipping cost tends to be more than the cost of manufacturing them. The exception is bricks going across the Irish / Northern Irish border, as if you are a builder, your nearest brick supplier may well be on the other side of the border.
You have two years (well 18months now) to imagine all possible scenarios for trade with Europe from total free market access to a Cuba style blockade of Britain.
Then you implement all of them and the day before Brexit when we are told what the deal is - you just turn on that strategy for the morning.
".....so the deal won't be nailed until the day before Brexit" It is very obvious the majority of the EUSSR has no intention of making any deal with the UK unless it is so punitive as to make the current Nork sanctions a dream option. Which means we need two systems - one for the very unlikely event there actually is a deal, and a second for the much more likely WTO rules and reversion to our pre-EU import/export laws.
"....Juncker, Barnier, Guy vanWhatsisname etc aren't actually supranational dictators." They just believe they are.
On a more positive note (yes, Remoaners, there is one), the problem is not having a system, it's of having a fast and effective one. In the worst case, we will simply revert to the paper systems of pre-EU times and let the European imports stack up, rust and rot on our docks.
That's just the face of it. The whole mechanism for EU VAT and how it will be administered has yet to be decided. Once done, this has got to be rolled out to companies producing Accounts packages e.g., Sage. Then the man in the street who runs a VAT registered business trading with the EU will need to understand what's happening.
You can bet your life there are bespoke software and website shopping carts out there that have EU settings hard-wired into them which will have to be re-coded, or spoofed to ensure Native VAT Registered EU customers no longer appear as EU turnover with no VAT element, and ditto on the supplier side.
Is it feasible to roll out and test this within the alloted time frame? I doubt it,
Relax. With the widespread acceptance of BTIP (Beta Test in Production - known to the yoof as "DevOps") we only care about MTR (Mean Time to Remediate) rather than actually getting things right. So long as there is a continuous delivery pipeline set up, I'm sure HMRC will be relaxed about wildly inaccurate VAT returns for the 18+ months it takes to iron out all the glitches. If an inspector turns up just tell him you're being Agile. He'll understand.
I appreciate the streak of sarcasm in your post, but many will take your comment at face value. Here's some more fuel for the inferno:-
Even if my mind could stop boggling over HMRC being relaxed about anything that could pull wool over their eyes, there is still the issue that, having been freed from the constraints currently imposed on VAT by the EU, the guvment will no doubt want to tinker with those settings. We live in highly volatile times, why make things even more complex?
Tom Paine's comment is a fine analogy, but will no doubt be scoffed at, "as Y2K was a storm in a teacup". Yes it was a non-issue in the end, but purely because of the effort that everyone put into making sure that it was a non-issue. (Admission: I speak from first-hand experience as I was one of the developers who put offending code into some of London Underground's signal control systems in the 1970's, but we had no choice at the time due to hardware constraints).
Anyone reading my original comment and taking it literally is going to be in a whole world of pain when HMRC send in the proctologists and consequently shouldn't be allowed anywhere near such systems unless directly supervised by a qualified Accountant. There is more to system design than Node+React+Mongo (yes, I really have seen a 'clown shoes' trading system using a non-ACID DB).
Then if for example you are a car dealer, you have a separate piece of software from another supplier that links into Sage, and deals with the stuff that is specific to car dealers such as the VAT rules on second hand cars. That will also need to be updated because it only works with a specific version of Sage.
Boom boom.
Hmm. Fujitsu. I'm guessing big ass ICL mainframe? Proprietary 4GL almost no one has heard of?
And didn't anyone realize that Brexit would mean that while CDS is scoped to handle 3x what CHIEF is in fact (if) Brexit happend (y'know, worst case planning) in fact it would need to cope with a 6x rise?
Plus all those favorite ways to f**k up a project. Using IT to change an organization, rather than the other way round. Big bang. Undersized for expected work load and of course failure (and I'm sure it is a failure) to make the system flexible enough to cope with ongoing changes (IE Kerching. That's another £x000 for that mod).
Which is apparently a Codasyl compliant DB extended from IDMS, originally from "Culinet "
WTF that is. Now part of CA, but ICL got the source code, so no dealing with the sharky ones.
Probably big in the 70's.
If any old codgers out there know what any of that means HMRC or the Aspire consortium would probably like to hear from you. *
*Although you'll probably be paid in cat food. Stingy ba***ds
I was a member of the IDMSX development team at the time the final release was shuffled out of the door circa 1994/5.
Prior to that I was part of the 3rd line support team '85 to '92, although I cut my teeth on the 1900 version in the early '80's.
Still got an old diagnostic guide somewhere...
We had a bunch of people from "the revenue" seconded to ICL working on the systems integration phase of the end to end validationof the entire "Open TP[1]" feature set as they were planning on exploiting the new functionalities for developments scheduled for, i assume, '96 or '97. It might have been the system in question, but a lot of those central government applications were late '80's vintage.
Mine's the one with the SFL reference card in the pocket.
1) Distributed transaction processing for heterogeneous systems, All good fun until someone loses an eye.
No, that's not a typo. I'm sure they will illustrate exactly what happens when you combine (at least) two major, rapidly changing unknowns into a single project, that nobody has a clue how to implement within the required, unspecified but shortening timeframe.
Come come...
z14 (everything encrypted in flight & NO interserver network latency) + CICS (Transactional, where reqd + true multilanguage) + MQ(&IIB) + Liberty + distributed for serving UI (UI should be API based for agnostic front end) = robust/scaleable/fast
(just 'cus I work for BB doesn't mean I'm wrong)
Everyone is way overthinking this problem and its solution.
Create a generic text tax document online, have each vendor and customer print it out and fill it out by hand and then FAX it in to the HMRC.
Good way to employ lots of people post BREXIT.
That will be 10 million pounds, I can have it done in 30 days.
Could it possibly be that someone in the Treasury whispered in Hammonds ear.
"You do realise if we 'crash out' a la Boris; we won't be able to collect the revenue due to us, and you will have the mother and father of all financial crises, and it will take at least three years from now to fix it."
“CHIEF could carry on for another 30 years; it can be scaled. The problem is, it is very old and there are few people alive that can use it still.”
Well, when there is no one alive who can use it (is this because it was used by very few people or were the effect of using it so traumatic as to end lives early enquiring minds might like to speculate) then scaling is no problem. I could resurrect my old Science of Cambridge Mk 1(still works after a TTL transplant) and scale it at a vey reasonable cost.