Yup, and this is why I wish there was mandatory re-training and testing for drivers every 10 years. And the testing should be progressively harder as you get older.
I learned 8 and 4 "shuffle feed" steering as part of an EVOC class taught by a police academy. It was pretty easy for me having learned driving in snow and ice, but there were sections that still took some work and I've been driving for 20 years. Some of the other people in the class could barely make it through a emergency stop J turn without plowing cones.
You're doing the hot capture correctly, but I'm talking about the fact that you have to send poor techs into the hot isle to do work. Also the hot isle needs to be wide enough to allow techs in to cable things up.
If the cabling was the cold side of the machine, you wouldn't need to make the hot isle wider than a foot or two and it wouldn't even need doors.
This is a design problem with the server industry in general, not your setup.
If you can afford a rack of 1U servers you don't use redundant PSUs. Well, unless you're stupid. And there's a lot of STUPID out there.
You design tired load balancing and failover software so no single component is a SPoF.
In the age where supermicro can spit out cheap 95+% AC conversion with mostly single 12v rail mainboard design we're doing pretty well. The only real thing missing from the "Google" design is the per-machine battery backup.
The real problem with datacenter design is not the AC or DC anymore. Sure we can reduce waste heat by improving conversion, but most public colos are still doing power and cooling delivery like it's 1960.
We're still doing rack cabinets wrong. We still load servers from the cold isle, but connect all the cabling from the hot isle. Many datacenters don't do hot isle capture. Until we switch to wiring servers from the cold isle and ducting the hot isle away we can't get any real heat transfer efficiency.
Part of the problem is Dell, HP, IBM, etc are stuck with the 19" telco rack. This form factor is designed for the power density of a kilowatt or two. A fully populated server rack can eat 10kw easily. They're just not design to deal with that much heat output.
Yup, as a SF resident I can say I don't give a fuck about how much parking meters cost. I use them maybe once a month or two when I get a zipcar to do a few big item errands.
The only people that give a shit about this are people dumb enough to drive from the east bay instead of taking BART. (or caltrain from the peninsula)
Or maybe the people that live in the Marina and haven't figured out that owning a car in a major city is a bad idea.
Yes, because it's your god given right to put your 4000+ pound personal possession anywhere you want, any time you want, for however long as you want on crowded publicly paid for city streets.
Wrong, skin effect is related to frequency. There is almost no skin effect at 60hz, or even that much effect on audible frequencies in the 20khz range.
Yup, you're exactly the target market for the Nissan Leaf. It's 35k before tax rebates, so it should be under 30k after. If you only put 5000 miles on the 2nd car and it gets 28mpg you have about $500/year in fuel costs. I figure you save about $3000 in operating costs over 5 years. Probably still more expensive TCO than a mazda 2 or 3 compact-ish cars. But it's getting there.
Yup, or OEMs (Dell, HP, IBM, Acer, etc) who wanted to keep selling PCs for the same price through the months it would take to recover the supply side. They did this to avoid getting "left behind" and lose out on sales to each other due to having to increase their overall system cost due to drive prices. The truth is that home-built PCs are a tiny tiny fraction of the overall market and they are the ones who got left behind because they have no representative supply chain manager thinking about these kinds of things.
See, you're completely ignoring the point of my question to crow about how good you are at supporting 250 users all by your lonesome.
At a 10,000 employee company you don't have a single support person. Do you count the internal portal developer's time spent debugging a firefox on linux issue with a vendor provided app? Do you count them as well spending time fixing IE on windows? Or is that just part of the job of supporting the vendor provided app. When your're talking about accounting "desktop support" in a large network there are many non-obvious or partial admin tasks.
Even in your tiny research lab I'm sure there are a dozen people who are self sufficient in some way and possibly even help other people in their group. Do you count them?
I guess it depends on what you count as admins. I figure at my company we have an admin ratio of about 300-500:1 for Linux desktop workstations and laptops. But do you count the helpdesk people who answer any question for any OS including email, and mobile phone access to corp resources? What about the user storage admins? They make the NFS/CIFS and backups work.
I guess I don't know how that compares to the number of windows and MacOS admins we have, or the number of deployed machines.
Spent fuel rods with long strongly radioactive times like you are talking about come from LWR reactors. (http://en.wikipedia.org/wiki/Light_water_reactor)
These "spent" rods can be recycled into the perfect fuel for several other kinds of reactors that use the U238 "waste" directly as fuel.
Yup, ham radio is legally granted a PRIMARY license for the lower half of the 2.4ghz ISM band. The ISM users of this spectrum are SECONDARY. This means that the ISM users must ALWAYS give way to the ham users.
Sounds like you have a problem with file formats, not filesystems. Filesystem level compression is a stupid idea since it doesn't have any way to apply appropriate compression methods to the files. Should I apply zlib to uncompressed audio? No, use FLAC. Should I apply zlib to logs files? No, I should probably use something like LZO or Snappy that have block seeking.
Slashdot Mirror
Yup, and this is why I wish there was mandatory re-training and testing for drivers every 10 years. And the testing should be progressively harder as you get older.
I learned 8 and 4 "shuffle feed" steering as part of an EVOC class taught by a police academy. It was pretty easy for me having learned driving in snow and ice, but there were sections that still took some work and I've been driving for 20 years. Some of the other people in the class could barely make it through a emergency stop J turn without plowing cones.
You're doing the hot capture correctly, but I'm talking about the fact that you have to send poor techs into the hot isle to do work. Also the hot isle needs to be wide enough to allow techs in to cable things up.
If the cabling was the cold side of the machine, you wouldn't need to make the hot isle wider than a foot or two and it wouldn't even need doors.
This is a design problem with the server industry in general, not your setup.
If you can afford a rack of 1U servers you don't use redundant PSUs. Well, unless you're stupid. And there's a lot of STUPID out there.
You design tired load balancing and failover software so no single component is a SPoF.
In the age where supermicro can spit out cheap 95+% AC conversion with mostly single 12v rail mainboard design we're doing pretty well. The only real thing missing from the "Google" design is the per-machine battery backup.
The real problem with datacenter design is not the AC or DC anymore. Sure we can reduce waste heat by improving conversion, but most public colos are still doing power and cooling delivery like it's 1960.
We're still doing rack cabinets wrong. We still load servers from the cold isle, but connect all the cabling from the hot isle. Many datacenters don't do hot isle capture. Until we switch to wiring servers from the cold isle and ducting the hot isle away we can't get any real heat transfer efficiency.
Part of the problem is Dell, HP, IBM, etc are stuck with the 19" telco rack. This form factor is designed for the power density of a kilowatt or two. A fully populated server rack can eat 10kw easily. They're just not design to deal with that much heat output.
Drive all you want, parking's gonna cost you extra.
Those aren't "sketchy characters", they're software engineers. That's just how people dress in SF.
Even better is to just take BART into town and walk/cab.
Oh, and if you want Asian, goto Nombe. Yum.
Seriously, this needs to happen yesterday. Fucking Slower Than Walking Muni is why I ride a bike around town.
Yup, as a SF resident I can say I don't give a fuck about how much parking meters cost. I use them maybe once a month or two when I get a zipcar to do a few big item errands.
The only people that give a shit about this are people dumb enough to drive from the east bay instead of taking BART. (or caltrain from the peninsula)
Or maybe the people that live in the Marina and haven't figured out that owning a car in a major city is a bad idea.
And SF needs this money. Right now people living here are going to be paying off hundreds of millions for road repairs due to all the car traffic.
I wish they would extend parking meter hours from 6pm to midnight.
6 hours * 6 days a week * 52 weeks * 8200 spaces * $2.50/hour = only $38M/year. That's not going to pay back the bond measure any time soon.
Yes, because it's your god given right to put your 4000+ pound personal possession anywhere you want, any time you want, for however long as you want on crowded publicly paid for city streets.
Sweet! Where on Google can I buy (your) personal information?
I have some of my money go through a non-FDIC credit union. They have a non-federal deposit share insurance provided by http://www.americanshare.com./
Yea, I really need to plan a BWCAW trip this year. No cell service for a week is great.
Wrong, skin effect is related to frequency. There is almost no skin effect at 60hz, or even that much effect on audible frequencies in the 20khz range.
Based on some quick searching, about $100 per person under 18 in the US.
You think that's bad, try Peanut Butter.
Yup, you're exactly the target market for the Nissan Leaf. It's 35k before tax rebates, so it should be under 30k after. If you only put 5000 miles on the 2nd car and it gets 28mpg you have about $500/year in fuel costs. I figure you save about $3000 in operating costs over 5 years. Probably still more expensive TCO than a mazda 2 or 3 compact-ish cars. But it's getting there.
Yea, and the average family has 2 cars. And when they do a trip of over XXX miles they only take one of them.
Yup, or OEMs (Dell, HP, IBM, Acer, etc) who wanted to keep selling PCs for the same price through the months it would take to recover the supply side. They did this to avoid getting "left behind" and lose out on sales to each other due to having to increase their overall system cost due to drive prices. The truth is that home-built PCs are a tiny tiny fraction of the overall market and they are the ones who got left behind because they have no representative supply chain manager thinking about these kinds of things.
See, you're completely ignoring the point of my question to crow about how good you are at supporting 250 users all by your lonesome.
At a 10,000 employee company you don't have a single support person. Do you count the internal portal developer's time spent debugging a firefox on linux issue with a vendor provided app? Do you count them as well spending time fixing IE on windows? Or is that just part of the job of supporting the vendor provided app. When your're talking about accounting "desktop support" in a large network there are many non-obvious or partial admin tasks.
Even in your tiny research lab I'm sure there are a dozen people who are self sufficient in some way and possibly even help other people in their group. Do you count them?
I guess it depends on what you count as admins. I figure at my company we have an admin ratio of about 300-500:1 for Linux desktop workstations and laptops. But do you count the helpdesk people who answer any question for any OS including email, and mobile phone access to corp resources? What about the user storage admins? They make the NFS/CIFS and backups work.
I guess I don't know how that compares to the number of windows and MacOS admins we have, or the number of deployed machines.
You can get equally good AC/DC on small PSUs and big central PSUs these days. 6% on 1000 servers is the same if you have 1000 PSUs or 1.
The Parent post is correct and your reply didn't answer it.
Spent fuel rods with long strongly radioactive times like you are talking about come from LWR reactors. (http://en.wikipedia.org/wiki/Light_water_reactor)
These "spent" rods can be recycled into the perfect fuel for several other kinds of reactors that use the U238 "waste" directly as fuel.
See:
http://en.wikipedia.org/wiki/Liquid_fluoride_thorium_reactor
http://en.wikipedia.org/wiki/CANDU_reactor
Yup, ham radio is legally granted a PRIMARY license for the lower half of the 2.4ghz ISM band. The ISM users of this spectrum are SECONDARY. This means that the ISM users must ALWAYS give way to the ham users.
Sounds like you have a problem with file formats, not filesystems. Filesystem level compression is a stupid idea since it doesn't have any way to apply appropriate compression methods to the files. Should I apply zlib to uncompressed audio? No, use FLAC. Should I apply zlib to logs files? No, I should probably use something like LZO or Snappy that have block seeking.
This isn't 1995. Nobody cares about filesystem level compression anymore. Go buy a 2T drive.