I know El Reg likes to have fun with their headlines but I actually thought it was the millennium bug at fault here until I got half-way through the article.
The London Stock Exchange suffered a second day of problems yesterday as its new trading platform struggled to function again. On Tuesday the market failed to close correctly, causing confusion over closing prices for brokers and traders. Yesterday the MillenniumIT system displayed zeroes against some bid and ask prices, …
Since when has SUSE Linux Enterprise Server been equivalent to Microsoft .NET framework? Will El Reg be telling us that a new car based on the Internal Combustion Engine is the equivalent to a new car based on the VW Golf platform next?
You are supposed to be an IT rag after all...
*nix based systems are not real-time systems. Well, not if you've actually used a real-time system. Yes you can make them *look* like real-time if you throw enough horsepower at them but it's rather like watching that guy with the spinning plates ... as the number of plates rises ... he's going to start dropping them.
Royal Bank of Scotland is about to replace a whole load of its Infrastructure engineers with guys in India. Over the next year or so 10s of thousands of man-years of experience will be made redundant in the UK and picked up in India.
It'll be interesting to see how that experience goes for them...
People who bad-mouth money don't get what it is. You can't have lives without it. It is how resources get where they are needed and how injustices are healed. There are also a lot of sharp types who are constantly looking for a way to divert some of the flow into their own pockets- that's how we got flush toilets and mobile phones.
You think you could be happy without thinking about money? Stop thinking about it then. Let us know how you get on.
Lighten up a bit!
I think the original poster was making a thing called a JOKE. You see a JOKE is a funny thing people say to provoke a reaction to a situation that is becoming tense. JOKEs or HUMOUR can relieve the stress and tension and allow people to relax a little!
Now why don't you relax a little and take the comment in the vein it was so obviously intended you muppet!
Google it...Money is not the root of all evil...of course...but the antagonistic competition for profit is...it pits me against you...money is only the oil...still, we can, as a species, live without either...You lot have just not worked it out yet..one day :)
Time for the brightest minds on the planet to waken up I'd say...
If it was necessary a workaround would be found just as they did in this case. Critical bits would continue. 'Keep Calm and Carry on"!
Don't even consider talking about real time systems and their relative structure and merits until you've considered something 'mission critical' like train or EMU control, nuclear station or explosive chemical reactor plant control.... That is real 'real time' control. Everything else is just control. I suppose timeliness of information in the stock exchange starts to become important were you have automatic trades happening - but that is just automated greed...
I seem to recall loads of people slagging off Windows, when the previous system was having problems, saying how Windows should never be used for mission critical etc. etc. etc. Now I presume that we'll see similar people saying the same about Linux...
In reality, hows about this:
Both OSes are perfectly capable of running high up time/high load systems provided:
1) The software is designed and written properly
2) The OS builds are competently designed and implemented
3) The infrastructure is designed and implemented properly
4) The operators are competent
And finally as a tedious PS, to show I've not left anyone out:
Why aren't they using mainframe?
Trading systems require real time features - that's the "low latency" bit. .NET runs on top of the CLR, which isolates it from the critical timing facilities it needs. Java would have been another awful choice for similar reasons.
The SUSE system will be based on the Linux kernel's real time extensions, and therefore makes much more sense as a platform.
so presumably you can point us in the direction of some metrics to prove your point?
The JIT compiler only compiles CLR code into native code ONCE, then simply does an address lookup from then on in. It may be very, very slightly slower on the first run of stuff, but I very much doubt there's a critical difference in the timings (unless there's a difference in resolution of the timings, .NET is limited to 10ms IIRC, but I've run tests on it before and it has performed very well). If you have some figures to back it up, instead of making assumptions, I'll happily apologise and retract. I also can't speak for Java, but I would expect that it does suffer from some slowdown on real time-critical systems because that extra layer does actually exist.
This post has been deleted by a moderator
This is a common misconception. A real time system is not about speed. It's about predictability - more specifically, being able to be 100% sure that a given task will complete in a known amount of time. Usually, this is a very short time, but that's not the point.
.NET is almost as fast as C++ 99% of the time - except when it decides to run garbage collection, then a task that MUST complete within 10msec suddenly takes 150msec for no apparent reason, and you have a problem.
Not that simply using C++ would fix the problem: if you run it on a non-real-time OS, then the OS might still decide to preempt you for whatever reason every now and then, and you're still in trouble.
"The new Linux/C++ platform in it's already short life has already demonstrated as many and as serious problems as the Microsoft/.NET platform did through it's life", AC
"Due to a technical problem, that system sent out the daily automatic message to clients – stating that the closing auction was beginning – 42 seconds late"
> so presumably you can point us in the direction of some metrics to prove your point?
It doesn't require metrics as such, just a better understanding. The issue isn't about a "slowdown" it's about a "slowdown just when you don't want it." You're not quite right about the JIT - although it does only do it's thing once, it doesn't do it all in advance. It does it when it first encounters a new code path, which can happen at arbitrary times. Same with garbage collection - as described here:
"When a collection starts, the collector suspends all application threads..." If this happens just when you don't want it to, you get a delay and the trade happens outside the required time window. Things like memory paging also cause little hiccups. You won't notice them except in timing critical applications.
In Java-land there's the Realtime Specification for Java modification which solves some of these problems (ahead of time JIT, controllable GC, pinned memory, etc). Maybe there's something similar in .NET-land?
This post has been deleted by a moderator
Trading systems don't need real time facilities, what they need is to be sized correctly to process X volume of transactions in Y secs as general load, allowing for peaks, which will also be specified. This is not realtime processing. The actual response times don't seem to have been the problem with the old or new systems, rather the availability.
"But that was blamed on "human error", which the exchange claimed was likely sabotage." ..... Is there any other kind of malfunction if we are not to assume that there is mechanical life too?
*And their Stealthy TakeOver of Systems and MakeOver of Humanity is Sublimely Supplied by your Thinking that such Thinking, and therefore Life Phorm, is Preposterous and Most Unlikely even whenever IT Floats the Program in front of you.
"Wow .. I actually got to the second paragraph before I realised this comment was from amanfromMars 1" .... Anonymous Coward Posted Thursday 17th February 2011 12:54 GMT
I think that then positively proves the second paragraph rather nicely, AC. Thanks for the confirmation of the efficacy of the methodology.
I like to think that you have been advancing, AC, rather than suggest that things are being made much easier for more general understanding and virtual comprehension, although, as is quite natural and therefore to be fully expected in the quantum communications world, is it probably a bit of one and a lot of the other, for travel towards the Singularity of Omniscience Shared for a Better Beta Operating System in Big AI Societies and SMARTer Enabled Civilizations.
Why are we suddenly blaming SuSE or Windows for this? Do we know that this is an underlying OS fault? The article doesn't say. Just because I can't get OpenOffice to run on my own main Linux system, I'm not throwing out the OS, any more than I would throw out my Windows 7 system because it doesn't run Pinnacle Studio 10. Neither is the fault of the OS (in one case it's a setting problem, in the other it's the age of the application).
I'd want to see a lot more information about the cause and effect of this fault before I'd pass judgement.
When Windows was the OS hosting the platform lots of "the usual suspects" were queuing up to blame Windows/MS. Now there are a few people blaming Linux, but really it seems to be apparent that there is nothing like enough information available to blame either. In reality it's the infrastructure, support and software that are far more likely to blame.
Biting the hand that feeds IT © 1998–2022