Actually some of the large cooling systems in skyscrapers and Vegas casinos do that. They make big ice blocks that are made in the early morning when the power is cheap and then use it the rest of the day.
Most med-large datacenters have a bank of batteries either in an UPS or as a DC power source that they just charge all the time. If you can use an alternate source to charge them, then you just saved some on your bill and if the peak is at the time when the solar cells are at it's peak, then that's a load off you peak charge.
I think we have tussled on this one before DocRuby. I agree a lot of the pie in the sky things like immersion and telepresence things that were promised aren't there yet. Well, neither are flying cars and food in tablets. Just like these, I don't think the total tech picture is there yet. Once the picture is complete, it will happen.
A lot of the things we use in I2 are things that large corporations have used for a while already. We just needed the network to support it. The things we most take for granted now are the things we needed the most in the first place.
For the applications we need on I2, we need low latency. Would you want to do a teleoperation with 100ms ping times? Imagine a surgeon doing a teleoperation and he started slicing and the machine had to wait for the packets to be resent to complete it on a congested network? Would it stop and then cut deeper causing a major blood loss or puncture? We needed big pipes with low latency and fewer hops to do anything of meaning or reliability.
With regard to applications, when we were first hooked up to I2 we were doing 5mb/s video classrooms to three other institutions. That's 15mb/s to each school. No way we could have done that with commodity internet, qos was just a twinkle in somebody's eye then, and with that it would have been choppy at best.
We use I2 for video conferencing, large physics data, multi-university distance education, digital libraries, database replication and disaster recovery, and others too numerous to mention.
So I2/Abilene was not really about the network so much as the applications that run on it. If it were not there, many of the things we take for granted at our university would not be possible.
Internet2 is a production testbed. They were the first to do multicasting, IPv6, speeds around 10GB, etc. Those things are tested, developed, and deployed on I2 and the people that make the equipment use that experience to build it to the rest of the internet. Very few providers do v6 and multicast yet, possibly just a few of the big boys. The big new routers from Cisco and Nortel are being used on Internet2 and the cycle begins again.
That said there are some that wanted a more experimental network to do more cutting edge stuff. That is what the NLR (National Lambda Rail) is about, doing a very highspeed network using dark fiber with different wavelengths and getting 50GB or more. This is the cutting edge stuff.
The key is to move your UPS out of the temperature controlled area, since that's the biggest source of heat in most server rooms.
Problem with that is your going to have to swap out your batteries in the UPS something like every 6 months. The leading causes of battery failure is change of temperature and sustained above room temperature. We've had to replace at least half the wiring closet UPS batteries because they were getting in to the 80's or 90's in temp for a sustained period. That's a lot of money right there. We keep the UPS's in the room so the batteries will last for more than a year.
- "Sure you can buy that, but did you check to see if it is on contract?" Applies to state schools mostly I would guess, but frustrating beyond words.
This is not as bad as it used to be. With states doing everything they can to stretch a dollar, they are going on contract with major players like SHI and CDW that have very completitive pricing. Sure there are going to be some contracts that are still like that, but they are getting fewer and fewer in my experience.
Unfortunately, our school is not setup for multicast yet to I2. One of the things I've been poking our network guys to get finished. I've seen successful use on our own LAN for things such as SLP and multicasting drive images.
Some of the video streams that Internet2 has setup (especially for thier big conferences) are multicast. Since I can't do it, I don't know how well it works, but I have heard from others universities that have had good luck with it in video applications.
I would guess that many of the big networks probably have multicast on their backbones. Their clients though are not able to grasp it and implement it to thier WAN connections so it just works locally. Multicast is tricky, it takes some real planning and knowhow to make it work across networks effectively.
Oh yeah, lots of it. One of the things is IPv6 and multicast. The Abilene backbone (one of the I2's biggest) is entirely v6. The knowledge there on how it works on a grander scale is helping to tune and shape the works that come out of places like Cisco and Nortel. Thier code gets production tested first on Abilene and then to the big networks. We also get the new big routers to test with usually before anybody else does. If you go look at Abilene's website, you can see from the network graphic that it's pretty busy.
Interestingly enough there seems to be a moving away from expensive ATM connectors to cheaper 10GigE connections. Our state network has just converted the backbone to GigE, and I expect that our connection to Abilene will change to that soon as well. I think ATM for medium length hauls will die out, only to be used on extra long hauls like across contries and oceans. I can see the big networks doing this to to cut down on costs and brainpower. ATM is just too complicated.
One of the biggest things that we've used it for is something that needs low latency and big pipes, videoconferencing. We have had classes that have students at 3-4 different universities with the profs at each contributing to the class, even in the same session. These were the high quality 5MB/s streams times the number of universities. That's ~20MB going back and forth with all the overhead that that would have. We needed Internet2's pipes to do that.
It's also used to do regular Polycom conferences without the latency you see with busy Internet1 connections. Our I1 pipes get pretty clogged in the afternoon, and it's a mess to try to keep the connection, even with QoS. We've had conferences with others from Hawaii to New York over I2 with minimal dropouts or frame rate problems. Without the I2 pipe, these meetings would not have been possible, or at the least uncomfortable to be in. And it saves so much in travel time and money.
Heck, even the MCU that we use (to allow more than 2 units to talk) is in another state.
Another thing that I2 has done is start linking K-12 schools together as well. Many states use the same network for higher ed and K-12. So not only are the universities getting the benefits, but so are regular elementary and secondary schools that are using the same pipes.
We have had at least 6 MB serial ports die on us, and even the second one on the same motherboard. Where I worked before we had like a dozen Palm V deployed and it just keep happening, maybe once a month. We even did some tests and sometimes we would get overvoltages when the palm was cradled. There just seems to be no protection for the serial ports.
Like every new chipset enhancement, Intel always made the drivers. USB2 will be the same. They will just release them when XP gold comes out, like they have when all the other windows flavors came out, abiet a little later.
DHCP in v6 will not exist in the form we think of now in v4. The equivalent is the adding of a MAC address to the end of your local network upper address. Since v6 is tied to an actual geographical route, you just have to tack on the MAC address to the end to get a unique IP.
Of course I could be wrong on this but this is my interpretation of the RFC's.
Well, the first problem is the QoS is VERY propritary and few operating systems have the hooks to talk QoS. Those that do are limited. So to get around it, all you would have to do is turn off QoS on the workstation or server. All traffic would get pumped through the first pipe anyway. Seems about as effective as port blocking.
I work at a I2 institution and have seen a major change in the available bandwidth. It helps me get my work done faster. Here is an application of I2 from last fall at a few instituions.
Another benefit of I2 is as a testbed for new protocols and applications. Multicasting is already up on most of the network and IPv6 is being tested and will probably be in production in a year or so.
Slashdot Mirror
As for automated rendezvous, the Russians have been doing this for years. Just buy it from them.
Problem is the Russian one sucks bad. Seems like 8/10 times the ISS crew has to bring the Progress ships in manually anyway.
From what it looks like, the peering is back up. Internet Health Report
Actually some of the large cooling systems in skyscrapers and Vegas casinos do that. They make big ice blocks that are made in the early morning when the power is cheap and then use it the rest of the day.
http://www.cnn.com/2007/TECH/science/07/24/ice.cooling.ap/
Actually I've heard it's Level 3. Much cheaper rate and better locations for POP's.
Most med-large datacenters have a bank of batteries either in an UPS or as a DC power source that they just charge all the time. If you can use an alternate source to charge them, then you just saved some on your bill and if the peak is at the time when the solar cells are at it's peak, then that's a load off you peak charge.
I think we have tussled on this one before DocRuby. I agree a lot of the pie in the sky things like immersion and telepresence things that were promised aren't there yet. Well, neither are flying cars and food in tablets. Just like these, I don't think the total tech picture is there yet. Once the picture is complete, it will happen.
A lot of the things we use in I2 are things that large corporations have used for a while already. We just needed the network to support it. The things we most take for granted now are the things we needed the most in the first place.
For the applications we need on I2, we need low latency. Would you want to do a teleoperation with 100ms ping times? Imagine a surgeon doing a teleoperation and he started slicing and the machine had to wait for the packets to be resent to complete it on a congested network? Would it stop and then cut deeper causing a major blood loss or puncture? We needed big pipes with low latency and fewer hops to do anything of meaning or reliability.
With regard to applications, when we were first hooked up to I2 we were doing 5mb/s video classrooms to three other institutions. That's 15mb/s to each school. No way we could have done that with commodity internet, qos was just a twinkle in somebody's eye then, and with that it would have been choppy at best.
We use I2 for video conferencing, large physics data, multi-university distance education, digital libraries, database replication and disaster recovery, and others too numerous to mention.
So I2/Abilene was not really about the network so much as the applications that run on it. If it were not there, many of the things we take for granted at our university would not be possible.
Internet2 is a production testbed. They were the first to do multicasting, IPv6, speeds around 10GB, etc. Those things are tested, developed, and deployed on I2 and the people that make the equipment use that experience to build it to the rest of the internet. Very few providers do v6 and multicast yet, possibly just a few of the big boys. The big new routers from Cisco and Nortel are being used on Internet2 and the cycle begins again.
That said there are some that wanted a more experimental network to do more cutting edge stuff. That is what the NLR (National Lambda Rail) is about, doing a very highspeed network using dark fiber with different wavelengths and getting 50GB or more. This is the cutting edge stuff.
Problem with that is your going to have to swap out your batteries in the UPS something like every 6 months. The leading causes of battery failure is change of temperature and sustained above room temperature. We've had to replace at least half the wiring closet UPS batteries because they were getting in to the 80's or 90's in temp for a sustained period. That's a lot of money right there. We keep the UPS's in the room so the batteries will last for more than a year.
Ok, is it just me, or would this have been a really good April fools post? Just doesn't seem believable.
Unfortunately, our school is not setup for multicast yet to I2. One of the things I've been poking our network guys to get finished. I've seen successful use on our own LAN for things such as SLP and multicasting drive images.
Some of the video streams that Internet2 has setup (especially for thier big conferences) are multicast. Since I can't do it, I don't know how well it works, but I have heard from others universities that have had good luck with it in video applications.
I would guess that many of the big networks probably have multicast on their backbones. Their clients though are not able to grasp it and implement it to thier WAN connections so it just works locally. Multicast is tricky, it takes some real planning and knowhow to make it work across networks effectively.
Oh yeah, lots of it. One of the things is IPv6 and multicast. The Abilene backbone (one of the I2's biggest) is entirely v6. The knowledge there on how it works on a grander scale is helping to tune and shape the works that come out of places like Cisco and Nortel. Thier code gets production tested first on Abilene and then to the big networks. We also get the new big routers to test with usually before anybody else does. If you go look at Abilene's website, you can see from the network graphic that it's pretty busy.
Interestingly enough there seems to be a moving away from expensive ATM connectors to cheaper 10GigE connections. Our state network has just converted the backbone to GigE, and I expect that our connection to Abilene will change to that soon as well. I think ATM for medium length hauls will die out, only to be used on extra long hauls like across contries and oceans. I can see the big networks doing this to to cut down on costs and brainpower. ATM is just too complicated.
One of the biggest things that we've used it for is something that needs low latency and big pipes, videoconferencing. We have had classes that have students at 3-4 different universities with the profs at each contributing to the class, even in the same session. These were the high quality 5MB/s streams times the number of universities. That's ~20MB going back and forth with all the overhead that that would have. We needed Internet2's pipes to do that.
It's also used to do regular Polycom conferences without the latency you see with busy Internet1 connections. Our I1 pipes get pretty clogged in the afternoon, and it's a mess to try to keep the connection, even with QoS. We've had conferences with others from Hawaii to New York over I2 with minimal dropouts or frame rate problems. Without the I2 pipe, these meetings would not have been possible, or at the least uncomfortable to be in. And it saves so much in travel time and money.
Heck, even the MCU that we use (to allow more than 2 units to talk) is in another state.
Another thing that I2 has done is start linking K-12 schools together as well. Many states use the same network for higher ed and K-12. So not only are the universities getting the benefits, but so are regular elementary and secondary schools that are using the same pipes.
We have had at least 6 MB serial ports die on us, and even the second one on the same motherboard. Where I worked before we had like a dozen Palm V deployed and it just keep happening, maybe once a month. We even did some tests and sometimes we would get overvoltages when the palm was cradled. There just seems to be no protection for the serial ports.
Like every new chipset enhancement, Intel always made the drivers. USB2 will be the same. They will just release them when XP gold comes out, like they have when all the other windows flavors came out, abiet a little later.
DHCP in v6 will not exist in the form we think of now in v4. The equivalent is the adding of a MAC address to the end of your local network upper address. Since v6 is tied to an actual geographical route, you just have to tack on the MAC address to the end to get a unique IP.
Of course I could be wrong on this but this is my interpretation of the RFC's.
See the link here It's much tougher to do it over longer distances. Quest learned a lot of this from the I2 project.
Well, the first problem is the QoS is VERY propritary and few operating systems have the hooks to talk QoS. Those that do are limited. So to get around it, all you would have to do is turn off QoS on the workstation or server. All traffic would get pumped through the first pipe anyway. Seems about as effective as port blocking.
I work at a I2 institution and have seen a major change in the available bandwidth. It helps me get my work done faster. Here is an application of I2 from last fall at a few instituions.
Another benefit of I2 is as a testbed for new protocols and applications. Multicasting is already up on most of the network and IPv6 is being tested and will probably be in production in a year or so.