Why would you need a server unless you have broken firewall rules. Your localhost should simply return TCP reset, which is much faster than having to actually service a page request.
Yea, except with Muni you can expect it to get you where you want to go at slightly faster than walking pace. This is why I bike around. I can bike from the outer sunset to SOMA in less than 30min. This tends to beat driving in many cases as well.
I ride to work as well. Your numbers seem a little high in some cases. Brake pads shouldn't cost more than $10 if you get the modern swappable pads so you don't have to replace the whole brake shoe/mount. Even those are only $20 USD.
I also think you can get a lot more distance out of a bike. I have about 9000km on my current bike (only 4 years old). It's a steel frame bike, so I expect it to last a lot longer than 20,000km.
I did have to replace one of my wheels due to a crack that formed at one of the spoke holes. I might try and pick up a replacement rim and re-lace it.
I have a similar tale. Starting with barely working slackware on 486 machines and a pile of floppies. I still smile every time I drive by walnut creek, CA. Those CDs saved so much hassle downloading on a modem.
I also had a 64bit alpha machines. (DecPC 150, multia) It took a long time to get the majority of software developers to stop doing 32-bit only hacks in their code.
Yup, I'll compare 2 buildings. An office in NYC, and an office in Dublin Ireland. The NYC bulding is 10 stories, the Dublin building is only 7.
In Dublin the stair case is nicely decorated. Good hardwood rails, decorative lighting, it looks like any other part of the office in that building, the stairs open up into the main landing on each floor. Right next to the elevators. If the elevator seems busy, you walk a meter to the right/left and you can just take the stairs.
In NYC, the stairs are bleak bare concrete tread with bare white walls and industrial lighting. It's nothing more than a fire escape. Worse yet, the entrances to the stairs are nearly hidden. The building designers seemed like they actively designed it so you can't find them.
It's no wonder that people in Dublin take the stairs just as often as people in NYC take the elevator.
Tesla doesn't use "traditional laptop cells" either. They're the same size and shape, but they picked specific models with different chemistry to normal laptop cells that suit car safety needs more.
So you really don't seem to understand how datacenters are using water.
Most of the cooling they use is Evaporative. They use the thermal property of evaporation to reduce the temperature hot return water. This is how they consume water, they just evaporate millions of gallons into the air.
Most large buildings do this. You will see this type of cooling on any building larger than a small office. When I worked at the university, I would go up to the roof of a 20 story campus building that had huge 5 meter wide/tall cooling towers to evaporate water to cool the whole building complex.
Fluorescent lamp startup power usage is an urban myth. Modern electronic startup ballasts draw a tiny amount of extra current for a fraction of a second.
If it were true, your other lights in the house would dim any time you turned on a CFL.
As someone who has done a number of those things, I highly recommend it.
The fridge in my apartment was already a fairly efficient model.
I got rid of the dryer completely, and use a high-efficiency washer and hang-dry my clothing.
I have a rear-projection TV, but I watch maybe one or two movies a week.
I replaced my dual athlon server with a nice dual core that supports frequency scaling.
I have been replacing the bulbs around the house with CFL (various color temps depending on the room) as bulbs have gone bad. I had one batch of defective ones that didn't hold up to the moisture in the bathroom like they were supposed to. They have a 7 year warranty, I just need to send them back.
Overall, I cut about 30-40% off my electricity usage.
That's just a matter of software in distributed computing. You just have 2 (or more) worker shards do the same calculations and then you compare the results. You don't need fancy mainframe hardware to do that. You can do even better by having 3 machines do the same work and vote out the bad one.
You say you can replace 40,000 machines with one mainframe. If google has 1000 machines in a container, each with 8G of ram, that's about 40 containers worth of PC servers. Or about 312T of ram.
The IBM z10 maxes out at 1.5T of ram. How does that compare again?
When I worked for a University, we bought a few of the largest IBM pSeries machines (power4 at the time). These were powerhouse machines 5 years ago. Each one had a dedicated 24" oversized rack cabinet, and then we had a couple racks just for disk. The 4 machines, and about 40T of Fibre channel disk (or was it DASD), I think it was a total of 128 core and 256GB of ram. I think we paid about a million for that setup.
As was mentioned elsewhere on the webs, the machine shown off by Google was based on Nocona CPUs.. those are atleast 4 years old now. Not likely what they're buying new now.
I bet you could get a base z10 for a few hundred thousand, but a fully loaded one? With a disk array of 750 drives? I bet 4 racks of disk from IBM would cost most of that 950k budget.
This completely ignores the fact that electricity is cheaper per mile driven than gas. Based on comparing a 25mpg car to a Tesla (55kwh/220mi=250Wh/mile) and CA 0.12/kwh and $2.50/gallon of gas. I also removed the battery replacement since the Tesla's battery is good for 100k miles.
You don't get 5 nines out of a single server install, sorry. The only way you get that is with HA clustering and automatic failover.
PC hardware, even expensive stuff, is not reliable enough no matter what $VENDOR's sales pitch is.
You might get lucky and get a single reliable box, but if you deploy a non-trivial number of servers you will need to plan for hardware/software failures.
Also explaining that open source/standards started at some of the best universities out there. There is also academic freedom that comes with not being tied to a specific corporation that can help gain you some support from the faculty. Getting a few faculty behind you can turn your idea into a huge group movement.
Or this is a direct attempt to go after what non-aware end users think "looks great!" and not what really is great. Think BOSE. Their speakers are not accurate at all, but they "sound great" to the people who think that the speaker built into their TV sounds good.
Personally I did some research and with with an IPS based 24" screen (HP LP2475w) to replace my crappy old 17" TN panel.
The funny thing was I had been using the crappy old screen for so long that the new screen weirded me out at first.
That's the problem with electronics. If your stuff is old enough that NONE of it supports digital reception, it's not worth $3000 anymore.
And considering the fact that you already spend $3000 on TV crap, I don't feel bad for you spend a few dollars on a receiver.
Personally, I have not had broadcast TV or even cable hooked up to my TV in the last 5 years. I mostly just turned the TV off and did something more useful with my time.
Savings before you factor in transmitter efficiencies: 2.1 GW.
I have no idea what the real efficiency of a TV transmitter is, but if it were 80% input to ERP you get about 4.5 GW of energy used to keep running ATV.
Over the 115 day extension that's 12.3 Terawatt-hours.
Slashdot Mirror
Why would you need a server unless you have broken firewall rules. Your localhost should simply return TCP reset, which is much faster than having to actually service a page request.
So doing things like paying Andrew Morton to work on the kernel, and contributing patches isn't giving back to you?
Google Intern Views?
So completely missing from article summary and article itself is any information about the software.
This guy is just late to the party. HBase was contributed to the hadoop project by Powerset. A startup that microsoft bought.
Yea, except with Muni you can expect it to get you where you want to go at slightly faster than walking pace. This is why I bike around. I can bike from the outer sunset to SOMA in less than 30min. This tends to beat driving in many cases as well.
I ride to work as well. Your numbers seem a little high in some cases. Brake pads shouldn't cost more than $10 if you get the modern swappable pads so you don't have to replace the whole brake shoe/mount. Even those are only $20 USD.
I also think you can get a lot more distance out of a bike. I have about 9000km on my current bike (only 4 years old). It's a steel frame bike, so I expect it to last a lot longer than 20,000km.
I did have to replace one of my wheels due to a crack that formed at one of the spoke holes. I might try and pick up a replacement rim and re-lace it.
I have a similar tale. Starting with barely working slackware on 486 machines and a pile of floppies. I still smile every time I drive by walnut creek, CA. Those CDs saved so much hassle downloading on a modem.
I also had a 64bit alpha machines. (DecPC 150, multia) It took a long time to get the majority of software developers to stop doing 32-bit only hacks in their code.
Yup, I'll compare 2 buildings. An office in NYC, and an office in Dublin Ireland. The NYC bulding is 10 stories, the Dublin building is only 7.
In Dublin the stair case is nicely decorated. Good hardwood rails, decorative lighting, it looks like any other part of the office in that building, the stairs open up into the main landing on each floor. Right next to the elevators. If the elevator seems busy, you walk a meter to the right/left and you can just take the stairs.
In NYC, the stairs are bleak bare concrete tread with bare white walls and industrial lighting. It's nothing more than a fire escape. Worse yet, the entrances to the stairs are nearly hidden. The building designers seemed like they actively designed it so you can't find them.
It's no wonder that people in Dublin take the stairs just as often as people in NYC take the elevator.
Tesla doesn't use "traditional laptop cells" either. They're the same size and shape, but they picked specific models with different chemistry to normal laptop cells that suit car safety needs more.
It's not idiotic, it's very efficient, and as stated by TFA, you can use grey water.. stuff that has been cleaned after you shit in it.
So you really don't seem to understand how datacenters are using water.
Most of the cooling they use is Evaporative. They use the thermal property of evaporation to reduce the temperature hot return water. This is how they consume water, they just evaporate millions of gallons into the air.
http://www.google.com/corporate/green/datacenters/summit.html
Most large buildings do this. You will see this type of cooling on any building larger than a small office. When I worked at the university, I would go up to the roof of a 20 story campus building that had huge 5 meter wide/tall cooling towers to evaporate water to cool the whole building complex.
Fluorescent lamp startup power usage is an urban myth. Modern electronic startup ballasts draw a tiny amount of extra current for a fraction of a second.
If it were true, your other lights in the house would dim any time you turned on a CFL.
As someone who has done a number of those things, I highly recommend it.
The fridge in my apartment was already a fairly efficient model.
I got rid of the dryer completely, and use a high-efficiency washer and hang-dry my clothing.
I have a rear-projection TV, but I watch maybe one or two movies a week.
I replaced my dual athlon server with a nice dual core that supports frequency scaling.
I have been replacing the bulbs around the house with CFL (various color temps depending on the room) as bulbs have gone bad. I had one batch of defective ones that didn't hold up to the moisture in the bathroom like they were supposed to. They have a 7 year warranty, I just need to send them back.
Overall, I cut about 30-40% off my electricity usage.
That's just a matter of software in distributed computing. You just have 2 (or more) worker shards do the same calculations and then you compare the results. You don't need fancy mainframe hardware to do that. You can do even better by having 3 machines do the same work and vote out the bad one.
So you seem to know a lot about mainframes.
You say you can replace 40,000 machines with one mainframe. If google has 1000 machines in a container, each with 8G of ram, that's about 40 containers worth of PC servers. Or about 312T of ram.
The IBM z10 maxes out at 1.5T of ram. How does that compare again?
When I worked for a University, we bought a few of the largest IBM pSeries machines (power4 at the time). These were powerhouse machines 5 years ago. Each one had a dedicated 24" oversized rack cabinet, and then we had a couple racks just for disk. The 4 machines, and about 40T of Fibre channel disk (or was it DASD), I think it was a total of 128 core and 256GB of ram. I think we paid about a million for that setup.
As was mentioned elsewhere on the webs, the machine shown off by Google was based on Nocona CPUs.. those are atleast 4 years old now. Not likely what they're buying new now.
I bet you could get a base z10 for a few hundred thousand, but a fully loaded one? With a disk array of 750 drives? I bet 4 racks of disk from IBM would cost most of that 950k budget.
Ubuntu desktop CD installs don't pull down the updates during install. The alternate install CD does.
Applications -> Add/Remove. There already is a simpler interface for complete newbies.
This completely ignores the fact that electricity is cheaper per mile driven than gas. Based on comparing a 25mpg car to a Tesla (55kwh/220mi=250Wh/mile) and CA 0.12/kwh and $2.50/gallon of gas. I also removed the battery replacement since the Tesla's battery is good for 100k miles.
http://www.ccds.charlotte.nc.us/~jarrett/EV/cost.php
Total came out to $0.126/mile per mile for gas, and electricity is $0.049/mile
Over 8000 miles a year, you save $600.
You don't get 5 nines out of a single server install, sorry. The only way you get that is with HA clustering and automatic failover.
PC hardware, even expensive stuff, is not reliable enough no matter what $VENDOR's sales pitch is.
You might get lucky and get a single reliable box, but if you deploy a non-trivial number of servers you will need to plan for hardware/software failures.
Also explaining that open source/standards started at some of the best universities out there. There is also academic freedom that comes with not being tied to a specific corporation that can help gain you some support from the faculty. Getting a few faculty behind you can turn your idea into a huge group movement.
Or this is a direct attempt to go after what non-aware end users think "looks great!" and not what really is great. Think BOSE. Their speakers are not accurate at all, but they "sound great" to the people who think that the speaker built into their TV sounds good.
Personally I did some research and with with an IPS based 24" screen (HP LP2475w) to replace my crappy old 17" TN panel.
The funny thing was I had been using the crappy old screen for so long that the new screen weirded me out at first.
Yea, seriously, I have no use for coupons for TV. I'm sure there are a hundred people on slashdot that would have shared theirs.
That's the problem with electronics. If your stuff is old enough that NONE of it supports digital reception, it's not worth $3000 anymore.
And considering the fact that you already spend $3000 on TV crap, I don't feel bad for you spend a few dollars on a receiver.
Personally, I have not had broadcast TV or even cable hooked up to my TV in the last 5 years. I mostly just turned the TV off and did something more useful with my time.
It is worthwhile. For this one reason. Gigawatts.
Using data from the FCC, http://www.fcc.gov/mb/video/tvq.html I calculated the sum total effective radiated power of all TV stations in the US.
Total for ATV: 3.6 GW
Total for DTV: 1.5 GW
Savings before you factor in transmitter efficiencies: 2.1 GW.
I have no idea what the real efficiency of a TV transmitter is, but if it were 80% input to ERP you get about 4.5 GW of energy used to keep running ATV.
Over the 115 day extension that's 12.3 Terawatt-hours.