30th June

We've finalised the spec for the new case (absolutely no more last minute changes, I promise! Well, except for...) and I'm planning how to best fill the insides - scale drawings in Visio are proving very useful for this sort of thing, as always, and it does seem that there will be adequate room for the essentials... I'll have to lose the LCD panel for a while, I expect, but that will reduce my serial port usage by one and that's no bad thing considering the new case's more modern IO shield. I can't see any serious problems with arranging everything, but I don't have any experience of mounting neon lights, the variable-speed fans will need some thinking about, and there will be a whole bunch of drive bezels to dye black - so there will be plenty to keep me occupied there as well as the usual and, indeed, inevitable little hiccups and glitches.

With the three hard disks nestled safely behind a pair of 80mm fans and the CPUs being blasted by a 120mm, my main concern is cooling for the tape drive. Just spacing it out in the bays will help, as will the superior thermal properties of the aluminium chassis, but it's just occurred to me that I could probably modify one of my spare hard disk coolers to fit a tape drive instead. The drive is mounted in a 5 to 3 adaptor frame in the current case, and I think a cooler might slip in there quite neatly - it's certainly worth thinking about if I end up keeping the current case for another few months...

29th June

For a change, it's good to see my prophecies of doom starting to come true - a worm has been created to exploit a recent security flaw in the open-source Apache web server, at this stage primarily in it's Windows and FreeBSD incarnations - although even in it's present form the worm may cause unpleasant side-effects to Apache running on other platforms. The payload seems to be a standard distributed denial-of-service sort of thing, intended to create a squadron of zombie servers that can later be used to flood some poor, unsuspecting dupe with attacks from all over the world. Whilst I have no ill-will towards the unfortunate sysadmins affected, it's about time that they had a wake-up call - the "open-source software doesn't have security holes" attitude that seems so prevalent amongst the Linux weenies was always going to come back to bite them sooner or later, and at least this particular worm doesn't have an especially malicious payload - considering how much e-commerce is reportedly hosted on Apache systems, it could be much, much, much worse...

The long-awaited JPEG2000 format is finally starting to emerge to the end-user. It uses the modern wavelet compression techniques to allow improved image quality in smaller file sizes, as well as a few new bells and whistles - alpha channel transparency, 16bit colour-depth support and a lossless "compression" mode. Application support is patchy at present - Fnord Software have release a PhotoShop plugin to support the new format, but it's currently only available in Mac format. Likely filename extensions will be JP2, JPX, J2K and J2C - keep an eye out for them.

BT have opened their Connected Earth "online telecoms museum". It seems rather short on facts and rather long on gosh isn't it all wonderful, and the navigation and layout are rather peculiar, and all-in-all it's more marketing than museum. Oh, well - at least it's a few pennies of my comms bills that haven't gone to pay the over-inflated salary of some already bloated senior manager or director...

28th June

It seems that my new case will slip in just ahead of the mass-market - Thermaltake, the company that started me thinking about a cases a month ago with their Xaser II chassis, has launched a new suite of windowed cases even before the vanilla form are actually available to buy! These join Lian Li's various offerings, the pre-modded SunFlowers that inspired my design, and various small companies offering windowed versions of the ubiquitous Chieftec/Antec/Chenming chassis.

When I was interested in custom motorcycles, ages and ages and ages ago, we used to call these sort of products "factory customs" and spurn them with derision... Mind you, my oppos and I used to be pretty scathing about people who paid a chop-shop to customise a bike for them, too (however spectacular the result - can you spell 'jealousy', children?) and I'm sure that the same pecking-order exists for customised PCs - the popularity of multi-player first-person shooters like Quake and Half Life has given rise to the idea of "LAN Parties", where everyone drags their PC round to a friend's house and hooks up to play together. The current enthusiasm for case-modding seems to have been largely driven by the desire to show off at these gatherings, and I can certainly understand that: I must admit to several pangs over the last few years that hardly anyone has seen the work I've done on INFINITY in the flesh.

I've always been struck by how the recent trends in custom computers have so closely followed the trend in custom cars in the 1980s - the first modifications were always to gain performance, and the next to solve the problems created by the first: big-block Chevy engines shoe-horned into a small chassis need lots of sticky-up air scoops etc to feed the carburettors, just as CPUs over-clocked to obtain higher frame rates in 3D games need extra fans and big heatsinks... The next wave of modifications were designed simply to show off the first two - I've often seen Perspex windows set into a car's hood to show off the wiring, for example, just like the windows I'm planning for INFINITY - and those, in turn, encourage one to fit the shiniest, glossiest, prettiest components that can be found: copper and gold heatsinks, coloured spiral-wrapped cables (amazing how much my green IDE cables look like a car's HT ignition leads!), neon lights, chrome grills, flashing LEDs... Soon every new PC will look like a Cadillac El Dorado pimpmobile, at which point the early-adopters among us will probably move back to featureless, modular brick or blade components and relish their plain, business-like, no-frills styling. And so the Cosmic Cycle continues...

27th June

Looks like I've finalised the design of INFINITY's new clothes with Kustom PCs, and I think it's going to be pretty <expletive deleted> neat. I was just eulogising on it to Ros, and she let me enthuse for several minutes before asking the obvious question: "Will it be quieter?". Well, I hope so... In spite of many confident assertions online, case cooling does not seem to be an exact science, and at this stage all I know for sure is that the approach I took with INFINITY's SuperMicro chassis was not successful. The fluid dynamics of turbulence in a complex space is at the heart of the problem, of course, and as a chaotic phenomena PC airflow is obliged by the laws of physics to behave in whatever way will be most annoying and frustrating to the end-user. I've always thought that it would be comparatively easy to tune the airflow in a PC chassis if one could see the movement of air - and if I can find a way of generating cheap, "safe" smoke, the multi-windowed case I'll be buying will at last make that possible! I wonder if one can rent smoke machines in East London?

I think that my major blunder with the first incarnation was in overlooking the effect of the distributed.net client - the heat generated by both CPUs running flat-out 24/7 is certainly not insignificant, and when I was first researching the current state of cooling and PC airflow in the autumn of 1999, distributed computing was only just starting to take off in the mainstream. Most reviewers then seemed more concerned that their cooling system could cope with a few hours of hectic Quake deathmatch [Link not for slow connections], probably running a single (although admittedly likely to be over-clocked) CPU at fifty or sixty percent utilisation only. Any full-tower case will look quite friendly to a system like that, but may well turn out to be less so when exposed to the 52 watts of power dissipated by my CPUs alone...

At that point I threw fans at the problem, almost certainly increasing the turbulence inside the case and thus not returning a significant temperature reduction in exchange for their significant noise. The sound insulation was the final nail in the coffin, though - in spite of the early reviewers' direct statements to the contrary, on my particular system and in the particular way I installed it I saw a rise of between two and three degrees, which for me was fairly catastrophic - I have to turn on the secondary fans to stop the tape drive melting during full backups, and the DNet client has been retired for the duration. Your mileage may vary, but it might not do either, and I certainly won't be using the stuff again. Oh, and once fitted it can't be removed without significant efforts... You have been warned.

26th June

I'm beginning to get a little jaundiced with all of Dell's slim-line servers, actually, as we've also been having a whole bunch of problems with our four PowerEdge 1650 servers too, the screaming 2650's 1U cousin. For a start, we've so far been unable to get PCI parallel port cards to work. At all. The systems shipped with a 64Bit PCI riser card, owing to an unfortunate oversight, but although we eventually managed to prise 32bit replacement risers out of Dell, any attempt to install an actual PC card in them results in the Win2000 driver installation wizard hanging completely. It doesn't seem to be the card (we tried several different types) or the drivers (the manufacturer's own, the OEM's, the chipset reference drivers) and it's not a particularly common problem with Windows 2000 - so that really only leaves the server hardware itself...

Moving on, we've discovered that we can hang the system stone dead by just booting with a keyboard attached to the front port (there are duplicate IO and VGA connectors on front and back, for convenience in a rack environment), a problem that has so far survived a BIOS upgrade, two trips from a Dell engineer with three replacement parts, and much head-scratching all round... And today, as we were racing against time to perform a complete OS and app reinstall (a desperate attempt to fix the aforementioned keyboard problem) the last server decided to throw a small hissy fit, detecting the fan in the secondary power supply (which is not now, nor has it ever been, installed!) as having failed and flashing it's "everything is not Ok" LEDs fit to burst. It's a fascinating testament to the current state of hardware management instrumentation - the DMI app actually reports the fan as present, but at zero RPM! My PFY has been wrestling with this particular server for several days, and it's three brothers for the last month or so, and by this afternoon he had definitely had enough. I sympathise - I've been there...

This made me laugh: the online cartoon strip Angst Technology on "Theatre Defragging".

25th June

Another two hits direct from Google, today, both searching on 'Dell 2650 noise'. Evidently the noise that this server produces is a real cause for concern world-wide, and there's still no sign of the promised BIOS upgrade - although I've yet to be convinced that dropping the rotation speed of the fans won't just replace a noise problem with a heat problem... In fact, I have a gut feeling that this server is going to become one of Dell's lemons - which probably guarantees that my manager will buy at least a couple more: we have three Compaq ProLiant ML350 servers, which always raise an eyebrow when a Compaq engineer is on-site - "Oh, are those still running? Most of them have been written off by now!" Well, they are still running, yes, but they've had so many components replaced that they aren't really the same servers that we originally bought! Their favourite disaster is an unusual type of power supply failure, cunningly timed to occur just after close-of-business on Friday evening - the power comes on for an instant, then off for an instant, then on again, off again, on again... and a weekend's worth of that has so far killed two motherboards, five disk drives, several memory DIMMs, a CPU and a SCSI backplane. It's been a while since this has happened, actually, so presumably just writing about it here will make it happen again real soon now. Ho hum...

24th June

The first commercial version of LCDC has just been released, so I've been fiddling with the LCD display a little. It's been the optical equivalent of a paperweight, recently, and the memory of all that measuring and cutting when I mounted it has been nagging at me. However, I still find myself unable to make it do anything particularly slick - I can't make it produce anything in the way of a bar-graph, I don't have much luck with large fonts, I can't hook it in to the system's performance counters reliably and to make matters worse I think there's some kind of flow-control issue even at a fairly snail-like 19,200 bits/sec! I suspect that I'm missing something somewhere, and having now registered the driver software (a very reasonable 12) maybe I'll finally have the incentive to track it down.

23rd June

Plans for INFINITY's new clothes are progressing and this time I'm planning to take the easy way out, plotting and scheming with Kustom PCs in Ayr to pre-modify a case for me so all I'll have to do is transplant the internal organs - a significant job in itself, with the complexity of the wiring loom. The current plan is to start with a black SuperFlower SF-201, add large windows on both sides and the top, throw in a handful of 120mm fans and a couple of neon lights, and see where it takes me... The end result is likely to look something like this:

only better! It feels very lazy to have someone else do all the hard work, and I'm sure that I'll be muttering over the way they handle the fine details, but I really don't feel like another round of measuring, cutting, filing and worrying - especially as I've never worked with Perspex before and would rather not have to learn as I go along: I could easily mangle a couple of panels before finding the right cutting speed etc.

I'm can't be sure of significant improvements in either the noise levels or cooling capacity, but there are a number of factors that should help - 120mm fans in the side and top panels should provide a significantly better airflow over the CPUs, the improved thermal characteristics of aluminium case itself will help in general, and the use of manually-adjustable fans throughout should bring benefits to the ears... And even if it doesn't actually work any better than the current case, it will certainly look prettier! The airflow from the top blowhole should be enough to lift a couple of light streamers, too, which would be rather nice...

22nd June

The air conditioning in the little UPS room failed mysteriously on Friday morning... It was still switched on and set correctly, but there was a complete absence of cold air - and it's lucky that a colleague noticed when he did, as in the three hours or so since I'd checked first thing in the morning (it must have failed soon afterwards!) the UPS battery temperature had risen by 9C, so by the end of the weekend the batteries would have been ruined.

We both poked at the controls for a few minutes without result, and then while I was off trying to track down the building services manager to report the failure my colleague had the bright idea of just switching it off and on again. [Lesson one - always reboot the system] At that point it burst into life as if nothing had happened, which doubtless will produce much scratching of heads and random replacement of components on Monday when the engineers come back in. None of this inspires much confidence in what is definitely a mission-critical installation, I have to say - the commissioning engineer was reassuring me last week that he only seen three failures in this type of unit over the last twenty years; well, here's number four... Tsk!

I've been re-reading an old article at Dan's Data, tonight, on his experiments with some rather unusual magnets. Especially interesting are the pictures of a "colloidal suspension of ten nanometre magnetite particles in a liquid base" - better known to the world as ferrofluid - but since I was there last the article has been updated to include a colloidal polymer film that displays magnetic fields visually - pretty indeed!

21st June

This site has had an unprecedented bump of hits over the last few days, thanks to linkage from a more prominent 'blog... Most only stayed for a moment to look at the pretty picture of the Connection Machine, but several lingered to browse - and that's always nice to see, so whoever you were, I hope that you found something amusing or interesting! What has made me smile most, though, is that (to date) three of the recent visitors have come direct from Google, with search terms of "cooling fans on 2650 dell", "noise poweredge 2650" and "dell 2650 noise"... Evidently I'm not alone in suffering from Dell's latest opus, but at least we can all hope that the promised BIOS upgrade arrives before our eardrums start to bleed rather than afterwards!

I can't understand how this design ever escaped from the R&D lab, though - the idea of a full-height cabinet populated with twenty of those screaming bastards (and that's exactly what Dell's marketing is currently proposing!) fills me with horror... Well, as the hardware is obviously unusable in it's current form, I think we should form a 2650 Owner's Club, collect our various servers together into a spare rack, and escort it to Dell's UK corporate headquarters in Bracknell for a short audition... Hah!

19th June

The first UPS maintenance went smoothly, today, with no unpleasant side-effects on switching the inverter in and out of the loop. However, once opened up it turned out that the plaster dust from the recent building work had thoroughly gummed up one of the three intake fans (what is it with me and fans? They seem to have become a central part of my life!)... Fortunately it hasn't caused any over-heating, as the aircon in that room is massively over-specified and the UPS is operating down around absolute zero. The supplier is even willing to replace it as a warranty repair, which I think is more than reasonable of them as it's quite obviously an environmental issue and so "our fault".

The rest of the day was a rush, arranging for and replacing a failed disk drive in the mail server (Oh, I love RAID!), stuffing some more memory into a couple of the other servers, and beginning to evaluate McAfee's new ThreatScan vulnerability detection software. I started to do some work on the screaming PowerEdge, destined to become a SQL server, but was soon deterred by the noise... However, my PFY reports that Dell are on the point of releasing a BIOS update which allows control of the fan speed, so if we can keep the computer room temperature down far enough to permit a lower airflow then we can clock it down to a mere 4000RPM or so, at which point hopefully it will disappear into the constant drone of the other servers.

Ooh, look - pages of photographs taken at the Moffet Field Historic Computer Collection - everything from the guidance computer of a Minuteman missile to a classic PDP-8, via the fabulously doomed Xerox Star... Coo! This is the collection's curator, The Computer History Museum in Mountain View, California.

Here's a neat PC cooling gadget - a little microprocessor that monitors the internal air temperature and varies the speed of a whole bunch of fans in proportion. There are many other ways of doing this sort of thing, of course, but as an emergency measure when a newly built system's noise levels are unexpectedly high, it could be just the thing - I wish they'd been available a year ago...

18th June

Working with Dell's new PowerEdge 2650 server is rather like being beaten with a club - it's really good when it stops... They're a slimline 2U design, and the fans required to pull air through the small surface area of the front panel vents in quantities sufficient to cool five SCSI hard disks, two P4 CPUs and a pair of power supplies are unusual to say the least. There are seven of them in the current configuration (with room for a further two!), all of them ultra-high-speed 60mm Delta "screamers", and the noise has to be heard to be believed - a rushing, wailing sort of squeal that sets even the nerves of hardened techies on edge. Interestingly, the exhaust fans are actually pairs of contra-rotating fans in neat little integrated housings - a plausible design, but one that has been talked to death and ultimately rejected on most of the hardcore case and cooling forums. My big-iron ProLiants have stacked fans for redundancy, but I have the distinct impression from the Dell's contra-rotating blades that the aim here is to increase the volume of air moved - by repute, it's the only way in which this design is worth implementing.

I'm fairly sure from the fan specs (they're 8000rpm, dude!) that the volume and pitch of the noise is completely expected, but as I was leaving today my PFY was about to get Dell's technical support on the phone so that they could have a listen... And if it is a standard feature and not a collapsed bearing or similar, the server is going straight back as unfit for use - it makes more noise, and a more intrusive and annoying noise at that, than the other twenty-odd servers put together!

17th June

A very old engineering joke: "The turbo-encabulator in industry" was written by J. H. Quick of the Institution of Electrical Engineers in London, and first published in the Institution's Quarterly Journal in December 1944. It has been around a bit since then, but as far as I can tell this is the classical form:

The turbo-encabulator in industry.

For a number of years now, work has been proceeding to bring perfection to the crudely conceived idea of a machine that would not only supply inverse reactive current for use in unilateral phase detractors, but would also be capable of automatically synchronizing cardinal grammeters. Such a machine is the "turbo-encabulator." Basically, the only new principle involved is that instead of power being generated by the relative motion of conductors and fluxes, it is produced by the medial interaction of magneto-reluctance and capacitive directance.

The original machine had a base plate of prefabulated amulite, surmounted by a malleable logarithmic casing in such a way that the two spurving bearings were in direct line with the pentametric fan. The latter consisted simply of six hydrocoptic marzelvanes, so fitted to the ambifacient lunar waneshaft that side fumbline was effectively prevented. The main winding was of the normal lotus-0-delta type placed in panendermic semiboiloid slots in the stator, every seventh conductor being connected by a nonreversible tremie pipe to the differential gridlespring on the "up" end of the grammeters.

Forty-one manestically spaced grouting brushes were arranged to feed into the rotor slipstream a mixture of high S-value phenylhydrobenzamine and 5% remanative tetryliodohexamine. Both of these liquids have specific pericosities given by P=2.5C.nexp6.7 where n is the diathetical evolute of retrograde temperature phase disposition and C is Chlomondeley's annular grillage coefficient. Initially, n was measured with the aid of metaploar refractive pilfrometer (for a description of this ingenious instrument, see Reference 1), but up to the present, nothing has been found to equal the transcendental hopper dadoscope (2).

Electrical engineers will appreciate the difficulty of nubing together a regurgitative purwell and a supramitive wennelsprock. Indeed, this proved to be a stumbling block to further development until, in 1942, it was found that the use of anhydrous nangling pins enabled a kryptonastic boiling shim to the tankered.

The early attempts to construct a sufficiently robust spiral decommutator failed largely because of a lack of appreciation of the large quasi-piestic stresses in the gremlin studs; the latter were specially designed to hold the roffit bars to the spamshaft. When, however, it was discovered that wending could be prevented by a simple addition to the living sockets, almost perfect running was secured.

The operating point is maintained as near as possible to the h.f. rem peak by constantly fromaging the bitumogenous spandrels. This is a distinct advance on the standard nivel-sheave in that no dramcock oil is required after the phase detractors have been remissed.

Undoubtedly, the turbo-encabulator has now reached a very high level of technical development. It has been successfully used for operating nofer trunnions. In addition, whenever a barescent skor motion is required, it may be employed in conjunction with a drawn reciprocating dingle arm to reduce sinusoidal depleneration.

16th June

I've found a useful online resource, the PC Guide. Some of it is obviously more appropriate for newbies than hardened techies, but it's definitely of use to both: this page confirmed, without sifting through the manufacturer's obsolete system documentation, that INFINITY's Supermicro motherboard is in the 12" x 9.6" ATX form factor rather than the larger Extended ATX and so will fit even the smallest of the cases I'm currently pondering over. Well worth a look next time you need a fact at your fingertips.

A gratuitous image of a Connection Machine, courtesy of the NCSA. <sighs with envy> Oh, the flashing lights...



Click here for 1st June to 15th June


Weblog Archive