I thought the LSE had bought a linux system + the Indian software house that wrote it? And that it was fast and super and wonderful and would never crash again?
The London Stock Exchange has suffered yet another systems crash, leaving brokers high and dry since 9.30 this morning. The Exchange last went down in September 2008 and took almost the entire day to get back online. That outage, on one of the Exchange's busiest days, was the day after the $200bn bailout of US housing giants …
The LSE are going to move over to a Linux/Solaris platform (see tinyurl.com/yauqcdk), but it takes more than a few days/weeks to transfer a system of that size over to a new platform and adequately test it. I'd give it a year before it's fully up & running, more if the LSE start asking for design changes during the implementation phase :-)
How many times has the "five 9s" MS/.Net platform failed so far? Three times? Or are there more that I'm not aware of? And that in the space of, what, 3 years? Considering MS/Accenture were treating this as a flagship installation to show the world how great their respective platforms are, this is a pretty poor show.
It’s kinda comforting to an old C++ developer to know that there are still applications that really do need careful management of the heap without a garbage collector.. back in the days of yore there was this robot at Imperial College that could play table-tennis.. except it missed every tenth shot because of the Pascal garbage collector.. real-time applications need real-time solutions.
..and Java.. nobody mentioned Java.. come-on SOMEBODY must think Java would have been better..
Instead of thinking that all Linux is driven by forums, try some training or, horror of horrors, read the damn man pages.
The number of forum posts I've read asking idiot questions that are clearly answered by the man pages is shocking. The number of answers to those questions that are wrong and contrary to the man pages is even worse.
By the way Solaris is not Linux. Solaris is a lot more sensible/stable. One particular Linux flavour changed the location of bash from /bin to /usr/bin or something like that, requiring all scripts to be changed....
Installing software is much easier on Unix. Just download the package, uncompress it if required and run whatever command line package management tool to install it.
As compared to Windows which generally gets halfway through the install and crashes or pops out some sort of generic error about "insufficient memory" when actually you've got too much disk space. Yes it's easier to double click an icon that understand something about your computer, until something goes wrong and double clicking provides NO answers.
The worst windows software I've had installed itself to c:. Which was a shame because my Windows system directory and boot disk was on d:. It's amazing how much stuff gets really confused by that.
You can even install software on Unix as a user into your own area that is completely separate from any other users and doesn't require root privilege. (Some software packages are badly written to make that tricky) Try installing windows software without being administrator and see how far you get. Try installing to windows without putting some dlls in your windows system folder....
A good unix system will allow almost any changes. You can change your root location with chroot, change the root user account, most of them will still trundle along. Try not having "administrator" account in Windows.
The reason businesses use something has nothing to do with forums. If they want something they'll get someone trained on it.
You may as well say businesses don't use Cisco networking because it's complicated. They clearly do.
All our critical systems are cisco/unix. We don't run anything on Windows.
There are no software or hardware problems: they are all 'soft engineering' problems arising in bad management decisions and dysfunctional administration. As in, contracting out your core competence to Accenture and sidelining everyone who offered a negative opinion: that is to say, everyone who knew enough to be able to sort out the specific 'software' problem that they've got today.
Doubtless the problem will be resolved to the satisfaction of everyone who matters by a corporate rebranding exercise.
Anonymous comment, for obvious reasons...
I work for a global market data provider. These outages are not as rare as you think. I can count 10 failures globally on particular bourses for today so far.
These are very complex systems and sometimes even a single customer can take many months to migrate. Anyone that thinks you can move platforms in a matter of months is having a laugh. This migration will take years before it is signed off.
As for .NET hmmmmm..... we are about to purchase a new mainframe for this type of work. 99% of the systems that hang off our exisitng mainframe are Solaris or Suse. Windows is suited to the desktop but is nowhere to be found in the data collection side of things here. Even *nix systems are not failureproof although most of the issues are still human error. Going back on the previous large failure by the LSE the volumes for that day were insane compared to a normal trading day. The level of investment needed for a normal days trade would make your bedroom PC setup look like a VIC-20 in comparison.
"The Exchange restarted continuous trading at 14.00. A spokeswoman was uable to tell us what caused the failure."
The poor excuse for development code on the .Net framework? The utter incompetance and sheer bloody mindedness of the management who refused to listen to reason when the tenders for the platform where put out yonks ago?
"He also told me that a computer science degree was the only way to go if I wanted to get into the industry."
I see why he was smug - that statement alone is utter cr*p. I stuck at university doing a CS degree for 6 weeks before I got bored and wanted to do some real stuff - I left qualification-less and am now a developer responsible for the systems keeping 400 shops up and trading.
The only mission-critical failure we've had in 3 years was for 5 hours when a supplier's network route failed. The failure was caused by human error - some idiot decided to add an old router to their live network as a redundant gateway, but the router still had some old routes on it causing all our traffic to them to get redirected to a black-hole.
3 of the 5 hours downtime was them saying "there's no problem here; nothing's changed." As soon as we got our CEO involved, they'd accepted, identified and fixed the problem.
Badgers - because they can understand routes better than afore-mentioned supplier.
As a DOT NET application developer myself, I think you are all being a bit hard on the platform. As far as we know the problems could be down to hardware issues with the servers or networking equipment.
A well written DOT NET application can be just as reliable as anything running on linux.
I agree - to a point.
We have about 30 .NET applications deployed but the one we have the most trouble with (and I wish I had time to rewrite it into UNIX) is an app that processes in the region of 20GB worth of data in 24 hours - about 16GB comes in between 3 and 6am each day.
It constantly falls over with out of memory errors even though the server has 3.5GB free at the time. What I read up was that .NET will try and allocate some memory that's the maximum size in a single chunk - so a 2GB file will not fit into the server's RAM if there isn't a continuous 2GB chunk of RAM free.
If MS exposed the memory-allocation functionality, you could allocate non-continuous blocks of smaller sizes to fit it in - like I can (and have done in the past) on Linux/UNIX.
But I do agree, 2.7ms to process a transaction is a pretty darned good achievement - even if it isn't good enough for the LSE.
If the LSE is 21% owned by the Dubai Government and a Dubai state corporation announces that it can't pay it's interest, does that mean it's less of a software/hardware problem and more of a sleight of hand problem?
I wonder if they get charged £30 for a letter telling them that they've defaulted on their £48 billion debt repayments.
"If the LSE is 21% owned by the Dubai Government and a Dubai state corporation announces that it can't pay it's interest, does that mean it's less of a software/hardware problem and more of a sleight of hand problem?" .... Seventh of 7th Posted Friday 27th November 2009 02:29 GMT
Seventh of 7th,
Do the other 79% of LSE have a Solution to Offer Dubai to head off Foreclosure and IP Ruin and an Unnecessary Gratuitous Malevolent Humiliation? And if not, why not? Is Intelligence Missing?
Biting the hand that feeds IT © 1998–2019