All ‘Change

Back in 2006 and 2007, Microsoft ran a lot of advertisements, and published case studies, on the Windows-based trading system that had been installed at the London Stock Exchange [LSE] in September, 2005:

Using the Microsoft® .NET Framework in Windows Server® 2003 and the Microsoft SQL Server™ 2000 database, the new Infolect® system has been built to achieve unprecedented levels of performance, availability, and business agility. … Its successful implementation, with support from Microsoft and Accenture, shows the London Stock Exchange’s leadership in developing next-generation trading systems.

I have no special inside track on the doings at Microsoft’s Redmond WA campus, but I think this had to be a pretty big deal from their point of view.  Historically, Sun Microsystems, with its UNIX workstations, had pretty much owned the “Wall Street” market, with HP, IBM, and DEC (as it then was) making occasional sales, also with UNIX workstations.  This workstation-based technology made huge inroads with financial market firms beginning in the late 1980s.  It was usually an easy choice; the only realistic alternative at that time would have been a mainframe-based system.  PC-based systems in general, and Windows in particular, could not come close to delivering the performance and reliability required for a (nearly) real-time trading platform.  Your obedient servant was quoted in the London financial press in 1992, saying that “anyone who installs a trading-floor system based on Windows has rocks in his head.”

So I think Microsoft valued the LSE deal, not least because they felt it showed that Windows had “grown up”, and was ready to compete with the big kids on the block.  The primary system that the LSE had been using was quite an old one, called SEAQ (Stock Exchange Automated Quotation system).  It was a modified version of a system called  NASDAQ (National Association of Security Dealers Automated Quotation system), originally developed for the US over-the-counter equity market.  LSE had adopted the system, which ran on Stratus fault-tolerant hardware, because it was a dealer market, like the US OTC market.  (It did not have a centralized trading “floor” like, for example, the New York Stock Exchange had.)

Thus, it was considerably embarrassing when, in September 2008, the LSE suffered a trading systems outage of six hours and forty-five minutes, at a particularly inauspicious time:

It should have been a great day on the London Stock Exchange. The U.S. government had announced on the Sunday before that it was coming to the rescue of Freddie Mac and Fannie Mae. Trading would have been extremely brisk, but then, at 9:15 AM GMT, the Exchange’s software failed due to “connectivity issues.”

No explanation beyond vague statements about “connectivity issues” was ever given, as far as I know, but people I know in the City have told me that the Windows-based TradElect system was to blame.

Beginning this past summer, reports began to surface that the LSE was planning to ditch the TradElect system.  It appeared that the system was never able to meet its own performance goals, nor could it match the performance of alternative systems.

It’s not often that you see a major company dump its infrastructure software the way the LSE is about to do. But, then, it’s not often you see enterprise software fail quite so badly and publicly as was the case with the LSE.

Now it is more or less official: the LSE is going to replace the TradElect system with a new, Linux-based trading system, built by a Sri Lankan company called MillenniumIT.

The question of whether London Stock Exchange (LSE) will replace its £40 million ($65 million) TradElect platform, supplied by Accenture, has finally been answered: yes, it will.

In fact, the LSE has bought the company, and will use its primary development center, located near Columbo, which employs 451 technical staff.  The acquisition price was only $30 million, less that half what the LSE has already spent on TradElect to get less than satisfactory results.  With this new acquisition, the LSE will obviously have much greater control over the future direction of development.  They also expect to achieve more frequent software updates, and to save approximately £ 10 million in annual ongoing costs.

This is not only a black eye for Microsoft on this particular project; there are not many places in the trading world now where Microsoft has any presence at all:

With LSE and its Italian subsidiary, Borsa Italiana, converting to Linux, Microsoft’s .Net offering is left with virtually no takers – the only remaining one being Johannesburg Stock Exchange (JSE).

Oddly enough, there hasn’t been much talk about stock exchanges coming out of Redmond lately.

Windows was designed as an operating system for desktop personal computers, and it has its place in that environment (although even there, Linux today is a very viable choice).  But when people ask me if I think that they should install Microsoft Windows servers and infrastructure, I always ask them the same question:  What do you think you know about running servers that Google, Amazon, and Yahoo! (as well as, now, the LSE) don’t know?

Splashy Booting

A couple of days ago, I wrote about efforts by PC and BIOS vendors to speed up the process by which PCs are started, initialized, and booted with an operating system.   Now Ars Technica has an article on an alternative approach to the start-up problem, Splashtop®, produced by a company called DeviceVM.

As I mentioned in my earlier post, the total time it takes to start up your PC includes both the time required for the BIOS firmware to run, and the time to load and initialize the operating system  What Splashtop does is to combine some speed improvements in the BIOS with a slimmed-down version of the Linux operating system (plus selected applications), in order to present the user with a start-up options screen within a few seconds.  In the existing Splashtop offering, for example, one can choose to start up a Web browser, a music player, or a photo manager, running in the Linux environment; or one can choose to go ahead and boot the primary operating system.  In essence, a Splashtop system is a dual-boot system, with one OS being the specialized Linux version, and the other being a more conventional OS (either Windows or Linux is possible, for example).  The Splashtop system may have part or all of the minimal Linux kernel in flash memory, to speed up booting.

As a point of reference, the Acer laptop on which I am writing this is set up as a dual-boot system, with Ubuntu Linux and Windows Vista (ugh!).  The OS choice is made from a menu presented by the GRUB boot loader, which is displayed as soon as the BIOS is done.  On this machine, in 10 tries starting with the power off, the mean time from pressing the power-on switch to getting the GRUB menu was eight seconds, with a standard deviation of less than one second.   Getting from the GRUB menu to the graphical (KDE) Linux login screen took an average of 56 seconds, with a standard deviation of about 3.5 seconds; most of that time is used in starting the graphical environment (the X Window system and KDE).  As for Vista … let me just say that, even in my advanced state of decrepitude,  I can probably run at least half a mile in the time it takes to start.  (I can definitely get and drink a cup of coffee.)

Now DeviceVM is introducing a new version of Splashtop that is intended for business customers.

The business version introduces a new e-mail client that is designed to sync with an instance of Microsoft Outlook that is installed on the computer’s Windows partition. It supports synchronization of e-mail, tasks, calendaring, and contacts. Another key feature in the business version is support for connecting to remote desktops via RDP or with client software for Citrix and VMware.

The business version also incorporates an interface to the Web-based Zoho productivity tools, so that users of the Splashtop environment can work with Microsoft Office® documents, for example.  DeviceVM also says that the Splashtop environment has additional reliability advantages:

DeviceVM says Splashtop has other advantages too, in addition to swift startup. For example, the company touts it as a fully functional fallback environment that users can rely on in the event that their regular Windows installation becomes unusable.

The company’s Web site also has a front-page blurb that says, “With Splashtop, you surf the Web safely, immune from malware that  targets Windows.”   DeviceVM CEO Mark Lee said in a statement,  “Even if Windows crashes or is infected by a virus or malware, Splashtop for Business gets users back online quickly to access corporate email, data, and applications.”

The idea of this is intriguing, and seems to have definite appeal.  The PC maker Dell, which has committed to shipping systems with Splashtop for Business, also has a new notebook offering, the Latitude Z600, which uses a similar approach to quick start-up:

The first idea, Latitude ON, announced over a year ago but fully debuting on the Z600, is basically a instant-boot OS combined with a self-contained system-on-a-daughter-card. The OS is Linux-based with a bunch of Dell customizations to both the OS and the applications (browser, mail, contacts, calendar, VPN client, and Citrix client)

I like to see companies trying to do a bit of innovative thinking, rather than just producing another “me too” laptop or other PC.  But I’m a bit bemused, too.  I have a laptop that is “immune from malware that targets Windows”, is extremely reliable, and that restarts from sleep mode in a handful of seconds.  Even though Windows is installed on the hard disk, it really is pretty easy to live with, as long as I don’t try to do anything crazy — like run it,

Bomb Scares

Over the weekend, the New York Times carried a rather disturbing report that the International Atomic Energy Agency [IAEA] thinks Iran has amassed the critical mass of knowledge necessary to build an atomic bomb;

Senior staff members of the United Nations nuclear agency have concluded in a confidential analysis that Iran has acquired “sufficient information to be able to design and produce a workable” atom bomb.

The IAEA report is hedged about, as reports from governmental agencies are wont to be, with cautions that the information is tentative and subject to further investigation and confirmation.  The details that have surfaced so far, however, are troubling.

Readers will recall that the United States dropped two atomic bombs on Japan, at Hiroshima and Nagasaki, helping to bring about the end of World War II in the Pacific.  What is not so well known is that the two bombs were of different designs, and used different fissile materials.  The Hiroshima bomb, dubbed “Little Boy”, was a gun-type device using enriched uranium as the fissile material.   The Nagasaki bomb, “Fat Man”, used plutonium in an implosion-type design.

A gun-type design is the more primitive of the two types.  Essentially, two pieces of enriched uranium, each of less than critical mass, are placed at opposite ends of a tube (the “gun barrel”); typically, one piece will be shaped like a hollow cylinder that fits over the other piece like a sleeve.  The two are rapidly rammed together by detonation of an explosive charge behind one of the uranium masses, resulting in a super-critical mass, a chain reaction, and an explosion.  A successful bomb requires ~25 kg of highly-enriched uranium (80% U-235); this type of design generally is not usable with plutonium.  This type of device tends to be inefficient, in terms of the energy produced for the amount of fissile material used.

An implosion-type bomb, such as Fat Man, is more difficult to design and build.  It uses a hollow sphere of fissile material, usually plutonium, which is compressed to super-critical mass by the detonation of a surrounding layer of shaped-charge explosives.  It is necessary to compress the plutonium as a sphere; if  it becomes misshapen, some material is likely to be ejected, resulting in what is termed a “fizzle”, which may spread a quantity of radioactive (and extremely toxic) plutonium around, without producing a very large explosion.

Gun-type designs have usually been the principal concern as far as terrorism is concerned: they require less expertise to put together. A crude gun-type bomb, which could easily fit into a truck (the barrel of the Hiroshima device was ~2 meters long), might not be efficient, but might be attractive as a terrorist weapon.  This is why some experts place such emphasis on securing available stockpiles of fissile material.

Countries with nuclear weapons programs universally use implosion-type designs, because of their greater efficiency, and because that efficiency allows smaller and lighter weapons to be built:

Implosion designs, compact by nature, are considered necessary for making nuclear warheads small and powerful enough to fit atop a missile.

Both the Times article, and a report in Wired magazine, suggest that the IAEA has evidence that Iran has been working on high-explosive technology, as a prelude to building implosion-type warheads that could be used on a missile.

Mastering the design of high-explosive lenses is crucial to the design of an implosion device like the Trinity gadget or the Nagasaki atomic bomb. In addition, the IAEA has also probed Iran’s interest in ballistic missile design. In the latest board report, the agency reiterated “the need to hold discussions with Iran on the engineering and modelling studies associated with the re-design of the payload chamber [for a new ballistic missile] … to exclude the possibility that they were for a nuclear payload.”

This does not sound good.  A country run by religious lunatics is not one that should be allowed to have nuclear weapons.