Do they pay attention to temperature with any granularity? "Cold aisle temperature" is not a single number; it varies. If it *doesn't* vary, they're either REALLY good (unlikely) or spending too much on cooling and charging you too much.
Field Guide to Datacenter Temperature Monitoring:
http://www.sensatronics.com/index.php/support/library/265.html
5% off from the Engineering department, use this code: ENG09 (expires Jan 1 2010)
I'm a design engineer at a temperature monitoring company (not a sales guy, I asked 'em not to bug you) and could probably give you a suggestion or two for your wiring etc. (A sales guy saw your post and asked me "how Slashdot worked" so I told him I'd contact you.)
Of course, I think in terms of our devices - we have a 16-channel monitor that's rated for installation down to -40 C (-40 F) and isn't terribly expensive ($499, I think, but since I'm a tech guy I don't know for sure) plus $50 per probe. Max run on a probe is half a mile (2600' or so) over cat5 (or you can buy more expensive long probes.) If you use someone else's hardware, look for industrial-temperature rated hardware if the PCB is going to be inside the cooled area. Mechanical parts (e.g. hard drives) should be avoided at all costs.
If you have any more specific questions I'd be happy to answer, even if you don't use our hardware. Our stuff is all Ethernet interfaced, for an example look at:
Perl scripts are available that grab data from the devices. Note that they don't log data (normally... there's an HTTP POST feature that does log up to 18 hours of data) but are designed to get the data into PC-land as quickly as possible. They run a tiny embedded OS, Linux is a bit heavyweight for cheap 256K flash hardware. (uCOS/2, not that it matters.)
Sad to say, that's 85% bullshit. The goth/industrial clubs I've been to (primarily Manray, may it rest in peace, but there are another 3 or 4 around) play only about 10-15% new content in a given month. It's not standard top 40 by any means, but it's still the goth equivalent of top 40. DJs who won't play the goth favorites are the ones who get handed their ass and asked to leave.
On the other hand, there is a definite expectation that a good DJ will try out that 10% of new stuff, most of which gets canned after a couple of plays because no one likes it, or relegated to early evening when not enough people are ready to dance to make it worth trying to get them onto the dance floor.
The goth crowd is less sheeplike than the regular club crowd... but not by nearly as much as they think they are. (Mind you, I'm one of them.) They're just sheep in a different pasture.
Why try to develop something entirely new, with the resulting time and money requirements? A few solar cells + Peltier coolers + some insulation and an ice tray. Yes, Peltiers are inefficient... but they're solid-state, at least, which I think ought to do for remote areas as far as durability. I would think you could assemble a decent mini-freezer out of things portable enough to carry anywhere:
1) Flexible solar panels (less efficient but more portable than glass) 2) A handful of Peltiers... they're pretty small 3) A couple of cans of "Great Stuff" spray-in insulation, or cans of A-B component expanding insulation
One of my friends went to Peru to assemble a non-electric solar water purifier, and anything they couldn't carry on their backs on 30-mile-a-day hikes for a week didn't go. Now that's a design constraint!
Last I checked, Google didn't give you the opinions of a bunch of geeks. This is exactly the sort of question that there's no real resource for except Ask Slashdot - maybe Google should start a Geek Opinions site.;-)
Interestingly, I believe these are from a college capstone course design, "product-ized" by the professor... of course, this is just what I hear around the office, and they're a competitor... could just be gossip, you should evaluate their quality and tech support yourself.
My company (Disclaimer: I am an engineer, not a sales guy) makes products designed jsut for this! Ethernet-based (or serial-port based) and work with MANY software options including Nagios, or we have a perl script for data collection that you could hack to send you email if you like do-it-yourself. Here's the device I recommend. There are some homebrew-ish solutions out there, but we sell to thousands of very happy customers, and provide everything from the basics to lots of bells and whistles. SMS messages are the most commonly used notification method, though a few prefer pagers. Hey, you can even contact one of the engineers directly for tech support.;-)
Note that this is how email works in "Earthweb" by Marc Steigler. Definitely not a new idea at all. I don't think it'd work for various technical reasons, but I can see how it might work for large companies in conjunction with whitelisting... a second, independent email network, almost. Support/sales email addresses would still have to put up with SPAM though.
Disclaimer: Yes, I work for a company that makes temperature, humidity, and leak sensing equipment for IT. Sensatronics
I think trying to convince server/CPU vendors to produce cooler products is futile... so long as the server room cooling is available, it makes sense to take advantage of it if it gets you more computing power per cubic foot/per U. If it were economically THAT much cheaper not to provide so much cooling, then hosting providers/rackspace sellers would probably offer reduced-price space for low-power servers. Does anyone know of a provider that gives discounts (or conversely charges fees on the other end) for low-heat colocation space? I haven't seen it.
On the disaster story side, I still remember watching the IT guys at a previous employer wander by the window to the server room periodically and check the thermometer in the window... now that I work for a company that sells cheap networked devices to do the checking automagically (and alarms on overtemp) I wonder how any IT department can justify something like that... it seems potentially disastrous, yet it was a large company I worked for, with (theoretically) experienced IT guys. Does anyone else see silliness like this where they work?
> Therefore: > * The modified version may not be distributed at all. > Incorrect. According to the GPL, it may be distributed with the source and a copy of the GPL. If it's for use inside the company, no huhu, no problem. If it's for distribution or sale outside the company.... it's GPL.
There are some large companies that DO this... use cheap nighttime electricity to compress (and cool, required or tank melts) air during the night, then use this energy to supplement during the day.
That's a self-solving problem, actually... it's the PAID advertising that's the issue, I believe. For instance, one of my company's competitors pays to have their product show up if someone searches for our company name... if it were in fact illegal to buy search terms that included another company's trademarks, then WE would be the ones to report it...all the courts would have to do is respond to complaints - same way it works now.
Unfortunately, Google won a similar case in the US, so our competitors can continue this practice that I, at least, see as a bit unethical.
Maybe I'm blind, but I don't see a table discussing what the different DVD-R samples were made of. There's a nice table of which technology each CD-R was (S1 - S7) but nothing for D1-D3, which is the information I want...so I can buy a stack of D2's.
> But too conscious control over processes which should be >random or out of conscious control has definite social and >political dangers, even if you don't mix humans with other animals. > Aieeeeee... it sounds like you're saying random chance has a better chance of producing a positive outcome than directed scientific research does. OF COURSE scientists should be careful... and guess what, they're TRAINED to be careful, and to spread their discoveries around so others can comment, but... these anti-research comments sound a LOT like the clips I just heard on NPR about how organ transplants were a terrible idea, unsafe, and we needed to do something about these self-glorifying surgeons who would even consider such a thing! Grrrr.
As I recall, the purpose of the second amendment isn't just self-defense; it's a provision included to keep the government honest. If (entirely hypothetical) the government became totalitarian with the trappings of democracy, it would still be possible to revolt. This requires that individuals be allowed to own whatever weapons they want. Using them, of course, is another matter.:-)
What kind of test setup did you use? I worked at [large Ethernet switch/router company] for 6 years, and one of our standard tests was learned unicast traffic between port-pairs. (Simulating two machines talking at full speed, since 90% of the network cards out there won't actually do 100% traffic.) We typically had no trouble doing linerate traffic on 30+ ports on a blade, or between blades. The marketing numbers for backplane bandwidth were based on what we could actually get (don't think that wasn't a fight, though.) The only reasons you'd drop from that in a real-world scenario would be:
1) Shitty NICs. Those $8 D-Link NICs really can't handle line-rate 100 Mbit sustained. Even the pricey ones have troubles, though as often as not that's a PC issue not just the card. 2) Broadcast traffic. Obviously, all broadcast traffic eats bandwidth on every port, not just one. Similar with multicast, but not as bad. 3) New connections - unlearned paths on a switch are broadcast to all ports until the receiver replies, and the path is learned. With a well-chosen aging time, only unused PC's will be dropped from the path table. They'll be relearned from the first reply packet, though depending on the switch architecture they might still be broadcast for some number of packets if they're sent a huge burst of traffic. 4) Other things I forgot - I've covered the major stuff, but I've probably forgotten a corner case somewhere. Seriously, if you use a good switch, you get the advertised bandwidth. Many switch manufacturers weasel-word their marketing, though, and seriously oversubscribe their switches.
I didn't include the name of the company I worked for because I'm not with them anymore, and their newest stuff sucks (IMHO).
Hardware designers use all caps on schematics, generally (except for comments.) SIGNAL_NAME is much easier to read on an 11x17 schematic with tons of "wires" crossing your text than Signal_name is... especially on a B&W hardcopy in the middle of a schematic review.
I own a 2003 Toyota Prius, so I thought I'd correct a misconception and make a comment. In the Prius, the electric motor *is* frequently used at highway speeds... the better-than-a-manual transmission design practically requires it, and it allows the car to trade unneeded engine torque for speed. For stop-and-go traffic, provided you're not driving binary (two speeds, stop and go), the ICE hardly ever starts (unless you've got the AC on.)
If you're an engineer, or can handle some gear ratios/math, do a google search for "prius transmission" - it's a damned nice design. Less than half the complexity of a typical automatic transmission, and more efficient than a manual transmission (though not by much.) Why? No clutch...indeed, no shifting! It varies the speed by varying the speed of the ICE between 1000 and 4500 RPM, and by varying the speed and sometimes the direction of the two electric motors. It's basically a standard planetary gear arrangement, with the small electric motor as the sun, the larger one driving the ring, and the ICE attached to the planetary carrier. The ring is also driving the wheels (through some gearing etc. as in regular gas vehicles.)
The Prius ECVT transmission is VERY different from previous CVT's. See discussion: Prius ECVT discussion
Read "Earthweb" by Marc Steigler... I like the ideas for SPAM reduction he presents there. Here's my brief impression of how it works:
A user has one or more "brands" which are forgery-proof digital identities. This could be approximated with public key crypto until quantum computing comes along, I'm sure there are many other technical methods for doing it. A "brand" may be anonymous or linked to the user's real-world identity, but anyone can attach a (digitally signed) endorsement or flame to a brand on one of several public sites.
If someone sends you an email, your email client can, of course, filter based on brands. Nothing new here. What is interesting is that it can refuse email unless a fee is paid... and there are various levels possible (obviously.) A typical user setup would be:
Don't charge people I specifically add to my "A" list. Charge people on my "S" list, whom I don't wish to speak to, $100 if they want me to read the mail. Charge brands linked to a verified real-world identity 10 cents. (Obviously doesn't include people on other lists.) Charge anonymous brands 50 cents to accept their mail.
Of course all this is user-configurable, I'm sure that if a system like this were in place that there would be all sorts of additions, like a version of the RBL for brands... check with the RBL and if the brand is listed there, tell the mailsender it'll be $50.
Obvious problems with this scheme compared to current email: Little compatibility with current mailreaders. Sender must be able to determine recipient's charge when mail is sent so they can agree when they send. -- This could be handled at the ISP, but still presents problems compared to current mail-handling, where you only need to be able to reach one hop to start an email on its way.
Slashdot Mirror
Do they pay attention to temperature with any granularity? "Cold aisle temperature" is not a single number; it varies. If it *doesn't* vary, they're either REALLY good (unlikely) or spending too much on cooling and charging you too much. Field Guide to Datacenter Temperature Monitoring: http://www.sensatronics.com/index.php/support/library/265.html 5% off from the Engineering department, use this code: ENG09 (expires Jan 1 2010)
This is clearly bad software design. The access card identifies you, and likely knows your weight within 10%.
Field Guide to Datacenter Temperature Monitoring:
http://www.sensatronics.com/index.php/support/library/265.html
I'm a design engineer at a temperature monitoring company (not a sales guy, I asked 'em not to bug you) and could probably give you a suggestion or two for your wiring etc. (A sales guy saw your post and asked me "how Slashdot worked" so I told him I'd contact you.)
Of course, I think in terms of our devices - we have a 16-channel monitor that's rated for installation down to -40 C (-40 F) and isn't terribly expensive ($499, I think, but since I'm a tech guy I don't know for sure) plus $50 per probe. Max run on a probe is half a mile (2600' or so) over cat5 (or you can buy more expensive long probes.) If you use someone else's hardware, look for industrial-temperature rated hardware if the PCB is going to be inside the cooled area. Mechanical parts (e.g. hard drives) should be avoided at all costs.
If you have any more specific questions I'd be happy to answer, even if you don't use our hardware. Our stuff is all Ethernet interfaced, for an example look at:
For humans:
http://em1.sensatronics.com/
For computers:
http://em1.sensatronics.com/config/
http://em1.sensatronics.com/data/
XML for computers:
http://em1.sensatronics.com/xmlconfig/
http://em1.sensatronics.com/xmldata/
Perl scripts are available that grab data from the devices. Note that they don't log data (normally... there's an HTTP POST feature that does log up to 18 hours of data) but are designed to get the data into PC-land as quickly as possible. They run a tiny embedded OS, Linux is a bit heavyweight for cheap 256K flash hardware. (uCOS/2, not that it matters.)
--Drew Van Zandt
Hardware Engineer
Sensatronics LLC
http://sensatronics.com/
http://www.librarything.com/
Does it all, I do believe.
Sad to say, that's 85% bullshit. The goth/industrial clubs I've been to (primarily Manray, may it rest in peace, but there are another 3 or 4 around) play only about 10-15% new content in a given month. It's not standard top 40 by any means, but it's still the goth equivalent of top 40. DJs who won't play the goth favorites are the ones who get handed their ass and asked to leave.
On the other hand, there is a definite expectation that a good DJ will try out that 10% of new stuff, most of which gets canned after a couple of plays because no one likes it, or relegated to early evening when not enough people are ready to dance to make it worth trying to get them onto the dance floor.
The goth crowd is less sheeplike than the regular club crowd... but not by nearly as much as they think they are. (Mind you, I'm one of them.) They're just sheep in a different pasture.
... so if someone with actual knowledge in the field contributes to a Wikipedia article, it's ruined? That's sofa king wee todd did.
...when I show you how to construct: ... from a Pringles can.
- A working fusion reactor
- A 3" mortar
- A simple teleportation device
AND
- A cat
Why try to develop something entirely new, with the resulting time and money requirements? A few solar cells + Peltier coolers + some insulation and an ice tray. Yes, Peltiers are inefficient... but they're solid-state, at least, which I think ought to do for remote areas as far as durability. I would think you could assemble a decent mini-freezer out of things portable enough to carry anywhere:
1) Flexible solar panels (less efficient but more portable than glass)
2) A handful of Peltiers... they're pretty small
3) A couple of cans of "Great Stuff" spray-in insulation, or cans of A-B component expanding insulation
One of my friends went to Peru to assemble a non-electric solar water purifier, and anything they couldn't carry on their backs on 30-mile-a-day hikes for a week didn't go. Now that's a design constraint!
Last I checked, Google didn't give you the opinions of a bunch of geeks. This is exactly the sort of question that there's no real resource for except Ask Slashdot - maybe Google should start a Geek Opinions site. ;-)
And I should preview... -also supports SNMP with or without traps, another model adds humidity and leak detection, etc.
Interestingly, I believe these are from a college capstone course design, "product-ized" by the professor... of course, this is just what I hear around the office, and they're a competitor... could just be gossip, you should evaluate their quality and tech support yourself.
My company (Disclaimer: I am an engineer, not a sales guy) makes products designed jsut for this! Ethernet-based (or serial-port based) and work with MANY software options including Nagios, or we have a perl script for data collection that you could hack to send you email if you like do-it-yourself. Here's the device I recommend. There are some homebrew-ish solutions out there, but we sell to thousands of very happy customers, and provide everything from the basics to lots of bells and whistles. SMS messages are the most commonly used notification method, though a few prefer pagers. Hey, you can even contact one of the engineers directly for tech support. ;-)
Note that this is how email works in "Earthweb" by Marc Steigler. Definitely not a new idea at all. I don't think it'd work for various technical reasons, but I can see how it might work for large companies in conjunction with whitelisting... a second, independent email network, almost. Support/sales email addresses would still have to put up with SPAM though.
Disclaimer: Yes, I work for a company that makes temperature, humidity, and leak sensing equipment for IT. Sensatronics
I think trying to convince server/CPU vendors to produce cooler products is futile... so long as the server room cooling is available, it makes sense to take advantage of it if it gets you more computing power per cubic foot/per U. If it were economically THAT much cheaper not to provide so much cooling, then hosting providers/rackspace sellers would probably offer reduced-price space for low-power servers. Does anyone know of a provider that gives discounts (or conversely charges fees on the other end) for low-heat colocation space? I haven't seen it.
On the disaster story side, I still remember watching the IT guys at a previous employer wander by the window to the server room periodically and check the thermometer in the window... now that I work for a company that sells cheap networked devices to do the checking automagically (and alarms on overtemp) I wonder how any IT department can justify something like that... it seems potentially disastrous, yet it was a large company I worked for, with (theoretically) experienced IT guys. Does anyone else see silliness like this where they work?
> Therefore:
> * The modified version may not be distributed at all.
>
Incorrect. According to the GPL, it may be distributed with the source and a copy of the GPL. If it's for use inside the company, no huhu, no problem. If it's for distribution or sale outside the company.... it's GPL.
There are some large companies that DO this ... use cheap nighttime electricity to compress (and cool, required or tank melts) air during the night, then use this energy to supplement during the day.
That's a self-solving problem, actually... it's the PAID advertising that's the issue, I believe. For instance, one of my company's competitors pays to have their product show up if someone searches for our company name... if it were in fact illegal to buy search terms that included another company's trademarks, then WE would be the ones to report it...all the courts would have to do is respond to complaints - same way it works now.
Unfortunately, Google won a similar case in the US, so our competitors can continue this practice that I, at least, see as a bit unethical.
Maybe I'm blind, but I don't see a table discussing what the different DVD-R samples were made of. There's a nice table of which technology each CD-R was (S1 - S7) but nothing for D1-D3, which is the information I want...so I can buy a stack of D2's.
> But too conscious control over processes which should be
>random or out of conscious control has definite social and
>political dangers, even if you don't mix humans with other animals.
>
Aieeeeee... it sounds like you're saying random chance has a better chance of producing a positive outcome than directed scientific research does. OF COURSE scientists should be careful... and guess what, they're TRAINED to be careful, and to spread their discoveries around so others can comment, but... these anti-research comments sound a LOT like the clips I just heard on NPR about how organ transplants were a terrible idea, unsafe, and we needed to do something about these self-glorifying surgeons who would even consider such a thing! Grrrr.
As I recall, the purpose of the second amendment isn't just self-defense; it's a provision included to keep the government honest. If (entirely hypothetical) the government became totalitarian with the trappings of democracy, it would still be possible to revolt. This requires that individuals be allowed to own whatever weapons they want. Using them, of course, is another matter. :-)
George W. Bush?
What kind of test setup did you use? I worked at [large Ethernet switch/router company] for 6 years, and one of our standard tests was learned unicast traffic between port-pairs. (Simulating two machines talking at full speed, since 90% of the network cards out there won't actually do 100% traffic.) We typically had no trouble doing linerate traffic on 30+ ports on a blade, or between blades. The marketing numbers for backplane bandwidth were based on what we could actually get (don't think that wasn't a fight, though.) The only reasons you'd drop from that in a real-world scenario would be:
1) Shitty NICs. Those $8 D-Link NICs really can't handle line-rate 100 Mbit sustained. Even the pricey ones have troubles, though as often as not that's a PC issue not just the card.
2) Broadcast traffic. Obviously, all broadcast traffic eats bandwidth on every port, not just one. Similar with multicast, but not as bad.
3) New connections - unlearned paths on a switch are broadcast to all ports until the receiver replies, and the path is learned. With a well-chosen aging time, only unused PC's will be dropped from the path table. They'll be relearned from the first reply packet, though depending on the switch architecture they might still be broadcast for some number of packets if they're sent a huge burst of traffic.
4) Other things I forgot - I've covered the major stuff, but I've probably forgotten a corner case somewhere. Seriously, if you use a good switch, you get the advertised bandwidth. Many switch manufacturers weasel-word their marketing, though, and seriously oversubscribe their switches.
I didn't include the name of the company I worked for because I'm not with them anymore, and their newest stuff sucks (IMHO).
Hardware designers use all caps on schematics, generally (except for comments.) SIGNAL_NAME is much easier to read on an 11x17 schematic with tons of "wires" crossing your text than Signal_name is... especially on a B&W hardcopy in the middle of a schematic review.
I own a 2003 Toyota Prius, so I thought I'd correct a misconception and make a comment. In the Prius, the electric motor *is* frequently used at highway speeds... the better-than-a-manual transmission design practically requires it, and it allows the car to trade unneeded engine torque for speed. For stop-and-go traffic, provided you're not driving binary (two speeds, stop and go), the ICE hardly ever starts (unless you've got the AC on.)
If you're an engineer, or can handle some gear ratios/math, do a google search for "prius transmission" - it's a damned nice design. Less than half the complexity of a typical automatic transmission, and more efficient than a manual transmission (though not by much.) Why? No clutch...indeed, no shifting! It varies the speed by varying the speed of the ICE between 1000 and 4500 RPM, and by varying the speed and sometimes the direction of the two electric motors. It's basically a standard planetary gear arrangement, with the small electric motor as the sun, the larger one driving the ring, and the ICE attached to the planetary carrier. The ring is also driving the wheels (through some gearing etc. as in regular gas vehicles.)
The Prius ECVT transmission is VERY different from previous CVT's. See discussion:
Prius ECVT discussion
Read "Earthweb" by Marc Steigler... I like the ideas for SPAM reduction he presents there. Here's my brief impression of how it works:
A user has one or more "brands" which are forgery-proof digital identities. This could be approximated with public key crypto until quantum computing comes along, I'm sure there are many other technical methods for doing it. A "brand" may be anonymous or linked to the user's real-world identity, but anyone can attach a (digitally signed) endorsement or flame to a brand on one of several public sites.
If someone sends you an email, your email client can, of course, filter based on brands. Nothing new here. What is interesting is that it can refuse email unless a fee is paid... and there are various levels possible (obviously.) A typical user setup would be:
Don't charge people I specifically add to my "A" list.
Charge people on my "S" list, whom I don't wish to speak to, $100 if they want me to read the mail.
Charge brands linked to a verified real-world identity 10 cents. (Obviously doesn't include people on other lists.)
Charge anonymous brands 50 cents to accept their mail.
Of course all this is user-configurable, I'm sure that if a system like this were in place that there would be all sorts of additions, like a version of the RBL for brands... check with the RBL and if the brand is listed there, tell the mailsender it'll be $50.
Obvious problems with this scheme compared to current email:
Little compatibility with current mailreaders.
Sender must be able to determine recipient's charge when mail is sent so they can agree when they send. -- This could be handled at the ISP, but still presents problems compared to current mail-handling, where you only need to be able to reach one hop to start an email on its way.