those of us, and I'm showing my age, who've been around for a long time may remember how Sinclair Electronics started up? Basically he'd buy up batches of transistors (remember them?) which had failed batch testing, and retest individual ones so he could keep those which worked sufficiently well. That gave him a better margin on his electronic kits, but also caused him problems when components would fail prematurely.
or at least that's the legend around here in Cambridge.
there are so many laws that it is nearly impossible to live in the USA w/o violating at least one a day
I think that's deliberate. How was Al Capone stopped? Not for being a gangster.
it's been done, there's been ubuntu for the zaurus for quite some time, it got the nickname zubuntu, and is still undergoing casual development, it can also be used to run android on zaurus
I was wondering why Google didn't scan for wifi access points when they did their streetview; this would have seeded their location database. After that, all the people running google maps on various smartphones with GPS capability would allow them to keep it up to date or cover areas street view mapping didn't cover.
Or maybe they did and sold the data to skyhook instead;-)
when I was a student engineer I worked at a division of Philips Electronics where no drinks were provided for free except water, there was a drink machine which wasn't too expensive (presumably to deter waste) - about 4 cents or 2 UKpence a cup. in one of the secure labs where only authorised people were allowed (high voltage and RF power) there was a clandestine kettle and tea/coffee kit, as kettles were not normally allowed due to "safety".
A lot of time was lost due to engineers interrupting each other to change money for the machines, and quite frequently whole groups of people would gather to chat for quite a while. One day when the company passed an important test of quality (ISO9000 IIRC) the machines were set to "free" for a week as a reward. I noticed the number of people chatting dropped and productivity must have risen, far in advance of the cost of the coffee.
I commented to a senior manager about that, and he somewhat agreed, but I don't think anything changed.
I do the same; except I also install thunderbird and remove all the links to outlook express (OE) or windows mail. To stop programs trying to invoke OE, I change the privileges on the file to remove the execute bit!
In TBird I change to only send plain text email, as that removes the temptation to pollute the internet with humongous ugly emails.
In FireFox I install flashblock so they aren't hit by drive-by infections, I also usually remove all the Microsoft originated shortcuts/favourites too, and set their home page to google.
I don't know if it's true, but I was told a long time ago that a particular witchcraft act was still on the statute books, and basically if a man had never seen his wife without makeup until after wedding he could divorce her!
OK, this is slashdot and I shouldn't rise to the bait with some actual facts.
Ok, my language may have been clumsy, but my point is this; I'll simplify since you clearly need it spelled out. We have two different JVMs running two different multi-threaded subsystems of a complex distributed application.
If we run just one or two of those JVMs, we can pretty much max out all the CPUs on the box and achieve N transactions per second. If load the box with other sub-systems by running more JVMs, we find that we cannot actually get to full CPU usage at all and yet there are no "missing" CPU cycles we can account for; instead of N/2 tps we can only get maybe N/4. As far as the OS and user space is concerned those missing CPU cycles effectively didn't exist, effectively the CPU just slows down without it being visible to the code being executed.
My point/question is therefore this: where's the facts about the effective percentage of CPU performance that's lost when the CPUs are synchronising caches and locking the memory buses due to multi-tasking? Can this be made visible to the OS and user space so that you can measure it and tune the system?
I used to set up dnetc on many machines, it caused the machines to run flat-out all the time, using more electricity as well as more cooling. Whilst it did only use up otherwise unused CPU cycles, it definitely had an impact on performance for higher priority processes. Modern CPUs run very fast and depend on their caches to maintain performance, and any context switching and loading other apps, even small ones, eats up memory bandwidth as well as CPU cycles. I think this goes relatively unnoticed as most people naively count CPU cycle ratios between processes.
At work we have a large number of dual CPU/eight-core (16 with HT) machines with 24, 32 or 36GB running java VMs, and we notice there's a very big hit on performance if we try and run more than a few VMs on a machine, almost certainly due to loss of cache efficiency; this performance loss doesn't particularly show up in simply looking for CPU cycles used by the OS!
I agree wholeheartedly. I once spent a lot of time trying to get a virtualised windows machine to run in plain old vmware server without the clock galloping head at 40% faster than wall-clock time; I tried many different things on the linux host side as well as the vm and the vmware tools install.
Will Intel and AMD please sit down like adults and come up with a standardised mechanism that virtualises and copes with dynamic clocking, multiple cores with/without hyperthreading and all the idle and sleep states.
x86 is with us because of backwards compatibility. even Intel were unable to shrug it off with Itanium and various other things. Intel even threw away Xscale Arm devices (sold off to Marvell, where Silicon IP goes to die) and did Atom instead.
there are loads of very good compilers; way back when I tried second guessing a C compiler for 68000 series processors and found it was very very hard to improve on what it could do. These days it's hard to write an x86 compiler because the x86 instructions are decoded and parallel-ised and executed out of order to keep the pipeline clear, and so you almost need a different compiler for each generation of CPUs (core2 is very different from P4 for example).
I was very lucky my parents recognised that I would only achieve if I had a private school education. The school I was in didn't make me feel there was any other way of life other than to work hard and achieve, even so I was a bit of an underachiever in comparison when I did exams at 15 & 16, but still managed to surprise my teachers, and went on to be a high achiever at "A" level.
The main component of the success of the school: parents who cared sufficiently about their children's futures to pay for their education and make the sacrifices that entailed. The rest is a side effect - kids work hard because they expect to, teachers work hard because the parents won't pay for 2nd class service, uncaring/disinterested parents won't even sign up, and finally the school can eject disruptive children.
Disruptive children *steal* the education of their classmates, and sadly for many parents there's not much they can do about it.
I would happily be a teacher in a private school, but not one in a state-funded school where being a baby-sitter is as much part of the job as teaching. I saw this in my mother who gave up teaching to being a nurse, as at least the recipients of her efforts were grateful.
* N+N redundancy in main supply *and* UPS *and* generators *and* cooling
* security - access to the site, protection of the installed equipment, storing equipment being delivered
* truly diverse fibre as well as sufficient bandwidth with low/no contention
* good peering policies to avoid poor latency to your customers' providers
* good value remote hands - some providers are very rigid about charging for every second you ask them for support, most are helpful with free basic cover
* hidden charges for telephone line installations and cable management
* extra charges for dual WAN uplink
* cost of committed AND "overage" charges
* protection from DoS attacks
* rack depth - will your servers fit?!
* patching infrastructure available for your own use to cable between racks if you expand, what cost?
not all of the following are yes/no decisions BUT should be considered because they can make a big impact to the time taken to get your service up and running
* power arrangements - who provides the PDUs - if buying yourself, will they fit, will you have the right brackets?
* parking - this is not trivial in london and can be a major bug-bear if you're delivering equipment
* traffic issues - if you've an emergency and you or the Dell or HP engineer can't get there to fix things without inordinate delays, you might have a growing headache
* onsite facilities when working - drinks and food prep areas, bathrooms, showers even (not needed these but some sites do, if you're doing a major roll-out they could be significant)
* rubbish disposal
* local facilities - eateries, hotels or B&Bs
* level floors - no steps - from loading bay to server floor, lifts etc. don't want to be lugging servers up/down stairs!
I visited six significant colo facilities before choosing the one I picked, most had merits, some had minor (to them) problems which would have been a minor but consistent headache to me.
We were recently forced to pull our entire server farm out of an inadequate colo, which cost us big time as we had to duplicate significant amount of equipment in order to maintain service instead of shutting down and relocating, as well as having to pay two hosting bills in parallel for a while. This came about because we originally chose the cheapest most convenient provider, which would have been ok for a while, but stayed with them far too long - moral, well before renewing or expanding, review whether your colo service is up to scratch.
UPS batteries are sealed lead-acid and they definitely benefit from being kept cooler, it's also good to keep them in a separate room, usually close to your main power switching. As far as servers are concerned, I've always been happy with ab ambient room temp of about 22 or 23, provided air-flow is good so you don't get hot-spots, and it makes for a more pleasant working environment (although with remote management I generally don't need to actually work in them for long periods of time).
Slashdot Mirror
those of us, and I'm showing my age, who've been around for a long time may remember how Sinclair Electronics started up? Basically he'd buy up batches of transistors (remember them?) which had failed batch testing, and retest individual ones so he could keep those which worked sufficiently well. That gave him a better margin on his electronic kits, but also caused him problems when components would fail prematurely.
or at least that's the legend around here in Cambridge.
there are so many laws that it is nearly impossible to live in the USA w/o violating at least one a day
I think that's deliberate. How was Al Capone stopped? Not for being a gangster.
it's been done, there's been ubuntu for the zaurus for quite some time, it got the nickname zubuntu, and is still undergoing casual development, it can also be used to run android on zaurus
http://www.google.co.uk/search?q=zubuntu+zaurus
I was wondering why Google didn't scan for wifi access points when they did their streetview; this would have seeded their location database. After that, all the people running google maps on various smartphones with GPS capability would allow them to keep it up to date or cover areas street view mapping didn't cover.
Or maybe they did and sold the data to skyhook instead ;-)
Can a single bottle of soda decimate your company? Absolutely.
when I was a student engineer I worked at a division of Philips Electronics where no drinks were provided for free except water, there was a drink machine which wasn't too expensive (presumably to deter waste) - about 4 cents or 2 UKpence a cup. in one of the secure labs where only authorised people were allowed (high voltage and RF power) there was a clandestine kettle and tea/coffee kit, as kettles were not normally allowed due to "safety".
A lot of time was lost due to engineers interrupting each other to change money for the machines, and quite frequently whole groups of people would gather to chat for quite a while. One day when the company passed an important test of quality (ISO9000 IIRC) the machines were set to "free" for a week as a reward. I noticed the number of people chatting dropped and productivity must have risen, far in advance of the cost of the coffee.
I commented to a senior manager about that, and he somewhat agreed, but I don't think anything changed.
In TBird I change to only send plain text email, as that removes the temptation to pollute the internet with humongous ugly emails.
In FireFox I install flashblock so they aren't hit by drive-by infections, I also usually remove all the Microsoft originated shortcuts/favourites too, and set their home page to google.
"false advertise" for the wedding.
I don't know if it's true, but I was told a long time ago that a particular witchcraft act was still on the statute books, and basically if a man had never seen his wife without makeup until after wedding he could divorce her!
I decided to reply to the original casual dismissal of my points, admittedly not well made, with a more detailed post:
http://slashdot.org/comments.pl?sid=1464550&cid=30305274
OK, this is slashdot and I shouldn't rise to the bait with some actual facts.
Ok, my language may have been clumsy, but my point is this; I'll simplify since you clearly need it spelled out. We have two different JVMs running two different multi-threaded subsystems of a complex distributed application.
If we run just one or two of those JVMs, we can pretty much max out all the CPUs on the box and achieve N transactions per second. If load the box with other sub-systems by running more JVMs, we find that we cannot actually get to full CPU usage at all and yet there are no "missing" CPU cycles we can account for; instead of N/2 tps we can only get maybe N/4. As far as the OS and user space is concerned those missing CPU cycles effectively didn't exist, effectively the CPU just slows down without it being visible to the code being executed.
My point/question is therefore this: where's the facts about the effective percentage of CPU performance that's lost when the CPUs are synchronising caches and locking the memory buses due to multi-tasking? Can this be made visible to the OS and user space so that you can measure it and tune the system?
I used to set up dnetc on many machines, it caused the machines to run flat-out all the time, using more electricity as well as more cooling. Whilst it did only use up otherwise unused CPU cycles, it definitely had an impact on performance for higher priority processes. Modern CPUs run very fast and depend on their caches to maintain performance, and any context switching and loading other apps, even small ones, eats up memory bandwidth as well as CPU cycles. I think this goes relatively unnoticed as most people naively count CPU cycle ratios between processes.
At work we have a large number of dual CPU/eight-core (16 with HT) machines with 24, 32 or 36GB running java VMs, and we notice there's a very big hit on performance if we try and run more than a few VMs on a machine, almost certainly due to loss of cache efficiency; this performance loss doesn't particularly show up in simply looking for CPU cycles used by the OS!
I agree wholeheartedly. I once spent a lot of time trying to get a virtualised windows machine to run in plain old vmware server without the clock galloping head at 40% faster than wall-clock time; I tried many different things on the linux host side as well as the vm and the vmware tools install.
Will Intel and AMD please sit down like adults and come up with a standardised mechanism that virtualises and copes with dynamic clocking, multiple cores with/without hyperthreading and all the idle and sleep states.
x86 is with us because of backwards compatibility. even Intel were unable to shrug it off with Itanium and various other things. Intel even threw away Xscale Arm devices (sold off to Marvell, where Silicon IP goes to die) and did Atom instead.
there are loads of very good compilers; way back when I tried second guessing a C compiler for 68000 series processors and found it was very very hard to improve on what it could do. These days it's hard to write an x86 compiler because the x86 instructions are decoded and parallel-ised and executed out of order to keep the pipeline clear, and so you almost need a different compiler for each generation of CPUs (core2 is very different from P4 for example).
actually, Halt-and-catch-fire might be more useful if you needed to light a fire.
magsafe power is pointless without magsafe network cable! admittedly, it's probably the primary cable in use that needs it
I was very lucky my parents recognised that I would only achieve if I had a private school education. The school I was in didn't make me feel there was any other way of life other than to work hard and achieve, even so I was a bit of an underachiever in comparison when I did exams at 15 & 16, but still managed to surprise my teachers, and went on to be a high achiever at "A" level.
The main component of the success of the school: parents who cared sufficiently about their children's futures to pay for their education and make the sacrifices that entailed. The rest is a side effect - kids work hard because they expect to, teachers work hard because the parents won't pay for 2nd class service, uncaring/disinterested parents won't even sign up, and finally the school can eject disruptive children.
Disruptive children *steal* the education of their classmates, and sadly for many parents there's not much they can do about it.
I would happily be a teacher in a private school, but not one in a state-funded school where being a baby-sitter is as much part of the job as teaching. I saw this in my mother who gave up teaching to being a nurse, as at least the recipients of her efforts were grateful.
why don't twitter and others scan tweets and replace third party shorteners with their own shortener services? it's really not that hard to do!
damn, html formatting FAIL!!!
* N+N redundancy in main supply *and* UPS *and* generators *and* cooling * security - access to the site, protection of the installed equipment, storing equipment being delivered * truly diverse fibre as well as sufficient bandwidth with low/no contention * good peering policies to avoid poor latency to your customers' providers * good value remote hands - some providers are very rigid about charging for every second you ask them for support, most are helpful with free basic cover * hidden charges for telephone line installations and cable management * extra charges for dual WAN uplink * cost of committed AND "overage" charges * protection from DoS attacks * rack depth - will your servers fit?! * patching infrastructure available for your own use to cable between racks if you expand, what cost? not all of the following are yes/no decisions BUT should be considered because they can make a big impact to the time taken to get your service up and running * power arrangements - who provides the PDUs - if buying yourself, will they fit, will you have the right brackets? * parking - this is not trivial in london and can be a major bug-bear if you're delivering equipment * traffic issues - if you've an emergency and you or the Dell or HP engineer can't get there to fix things without inordinate delays, you might have a growing headache * onsite facilities when working - drinks and food prep areas, bathrooms, showers even (not needed these but some sites do, if you're doing a major roll-out they could be significant) * rubbish disposal * local facilities - eateries, hotels or B&Bs * level floors - no steps - from loading bay to server floor, lifts etc. don't want to be lugging servers up/down stairs! I visited six significant colo facilities before choosing the one I picked, most had merits, some had minor (to them) problems which would have been a minor but consistent headache to me. We were recently forced to pull our entire server farm out of an inadequate colo, which cost us big time as we had to duplicate significant amount of equipment in order to maintain service instead of shutting down and relocating, as well as having to pay two hosting bills in parallel for a while. This came about because we originally chose the cheapest most convenient provider, which would have been ok for a while, but stayed with them far too long - moral, well before renewing or expanding, review whether your colo service is up to scratch.
UPS batteries are sealed lead-acid and they definitely benefit from being kept cooler, it's also good to keep them in a separate room, usually close to your main power switching. As far as servers are concerned, I've always been happy with ab ambient room temp of about 22 or 23, provided air-flow is good so you don't get hot-spots, and it makes for a more pleasant working environment (although with remote management I generally don't need to actually work in them for long periods of time).
when Iceland was going bankrupt: Google should buy the whole place....Icelandic women are really hot
Ah, that's why you never see Icelandic women working in data centres, they overload the air-con!!!
Imagine how backwards computers would be if you had to write a new kernel, window system, and libraries every time you wanted to write an application.
:-P
I don't have to imagine it... you can get pretty close if you go with Gentoo linux
>> Tchernobyl hasn't done all of its damage yet.
> It hasn't? Feel free to provide more info. After all, I live quite close.
you know that second head growing out of your shoulder? well, it has a hairy wart, and that wart is caused by living a near nucular reakta.
I worked for "the adult industry".... My wife didn't even really give me a hard time
:-)
/., please explain what a wife is!
you mean you didn't bring your work home? didn't even score some free vi@gr@? man, what a wasted opportunity
meanwhile, this is
another 1/3 is wasted powering computers used to read slashdot
satan is reported to start funding his own superconductivity laboratory