On top of that, their "2 hour window" is bull. I checked before I left home, and there wasn't anything there about reconfirming it. When I got to work (15 minutes later) it's already closed.
From another ISP, it sounds like they are talking about the Cisco NetFlow software. This software is used to have more fine-grained control of a large network. It allows you to see what areas of the Internet are sending/receiving most of your data, and allows you to change the routing to what looks like sub-optimal, but is actually better. Shaping traffic on a per-flow basis is actually a very processor intensive process, so most companies would rather use controlled congestion...
You might want to consider offering access to the 1 meg network in an open manner, but surcharge bandwidth usage (K transfered) above a point. I assume you could use IP accounting to manage this. You could also set it up so that when a new user brings up a new machine, a program watching the ARP tables could see the dynamic ARP entry, add a firewall rule to block all requests but port 80 and DNS. Then redirect port 80 to a local page where they can "log in." After they log in, create a static ARP entry for them and tie the accounting for their IP. If you do a search, MIT has done something like this on their network.
AT&T has made it very clear that they intend to become a monopoly again. They plan to start offering cable telephone access (have tried it in CA) and are going to fight tooth and nail to keep any outsiders out... If you want cable telephone, you will HAVE to use AT&T Long Distance. If you want Net access, you HAVE to use AT&T's @HOME.
AT&T is acting like a phone company, but doesn't want to be regulated like a phone company. The Telecomunications Deregulation Act of 1996 stated that ALL phone companies have to open their networks to competition.
The exemption that you are describing was put in place in the late 80's or early 90's. At the time, it was felt that modems and other data communications services needed to be protected. Since then, the telcos have been working to remove this restriction. Since Reed Hunt left the FCC, along with 5 other members of the FCC board, the FCC has been more than happy to do what ever the telco's want. There is NO leadership at the FCC at this time. They are only reacting to what is placed before them.
I realise that before the per-minute charges go into effect, congress will probably pass a law, but with all of the money the telcos can spread, do you think it will be able to get through congress?
We have run into the results of this already with GTE. GTE has declared us to be an IXC (Inter-eXchange-Carrier) and has prohibited us from purchasing any services that are local-tarrifs, including partial T1 circuits. I have 25 partial T1's already, but I can't purchase any more period. I can't move them, I can't change them, I can't touch them at all except to pull them out.
This ruling says that you can't be charged long distance on your PHONE BILL. However, the telco can go to court now and claim that since the FCC declared ISPs to be IXC's, they should be charged the same as Sprint, MCI, or other long distance carriers for each minute of access that the ISP uses on the TELCO network. Based on an ISP with 100 phone lines:
100 lines * 60 minutes= 6,000 line-minutes per hour
6,000 line-minutes * 5 hours of peak time (6pm-11pm) = 30,000 line minutes for peak
50% usage for the rest of the day, 3,000*19=57,000 line minutes non-peak
57,000+30,000=87,000 line minutes *.03/minute = $2,610.00/day * 30 days per month = $78,300 per month in usage. No ISP in existance can afford to absorb those charges and still provide flat rate. Assuming a 10:1 user ratio, the ISP should have 1000 customers. The average customer would have a $79 usage portion of their bill along with the standard $20 base rate.
This ruling doesn't do the above, it just opens to door for a future COURT case to open it. This is a form of incrementalism. These rulings are only the opening moves on a chess board. Once the pieces are in place, one move and checkmate.
I'm assuming that you will ignore that bit of advice, so here's what I'd do if I was starting again.
#1, have $50,000 in capital available. Not necessarily cash in the bank, but liquid that you can draw on.
#2, choose your provider. Negotiate the best deal you can get with them. If you know Telco stuff, purchase your own T1 to the provider. If they have frame relay available, make use of it. Full T1 for resale should run between $1500 and $3000/month total (Port from ISP and T1 cost) depending on whether you have to cross LATA's or not.
#3, shop for your equipment.
For your core router, choose Cisco period. No flames please. I've used Livingston, Engage, Cisco, Bay, Emerging Technologies, and Sangoma. From what I've found:
Livingston is good for general use, but the IRX series doesn't support full duplex on the 10BaseT port or have support for bridge interfaces over frame relay. Almost a requirement if you want to support ADSL.
Engage is cheap, and includes a CSU with the router, but it's limited on what you can configure.
Not many Bay users out there to draw on for support.
The guy that builds Emerging Technologies boards for PC's has poor support for Linux, and doesn't want to talk to anything not running Cisco HDLC. (It will work with PPP over HDLC, but not stable.)
Sangoma has good support for Linux, but they are difficult to configure if you've never done it before.
Cisco seems to be the most flexible for different configurations.
For access equipment, either 3Com/US Robotics or Livingston Portmaster 3. Both are about $10,000. You bring in a digital T1 to them and you're done. If available, ask your Telco for ISDN PRI lines rather than DCS. PRI's are 23 lines, where DCS gives you 24 but PRI's allow your equipment to talk to the phone switch with out-of-band signaling rather than having to use the same channel for data and phone call. Just easier to troubleshoot.
I personally like the US Robotics equipment. V.90/X2 I think is a more reliable technology. If you're looking to connect a lot of low-end modems, you might want to go with a 56Kflex based modems in the Livingston box. Personal preference.
You might also want to look at Ascend equipment. They seem to have the best VoIP implementation right now if that's important to you.
For servers, you want at least 3 to start with.
#1, Primary Name Server/Backup Authentication server
#2, Secondary Name Server/Primary authentication server
#3, User E-Mail/Web Server
#4, Buy your billing software. If you don't have an automated billing system, you're GOING to fall behind, and start loosing money. Check out usenet "info.inet.access" for information on different billing packages. Don't scrimp on this step. You want online signup, automatic account creation, etc.
#5, Advertise. You've got the basics, so now you need to get customers to pay for the equipment.
Cost Breakdown:
Router and CSU/DSU $2500
Access Server $10,000
10MB Switched Ethernet hub $500
Servers w/Linux 3@$1500 each $4500
Install on T1 and Phone Lines $7500-$10,000
Billing software and server $4000
Total Startup cost: $29,000-$31,500
Monthly costs:
T1 to ISP $2500
Telco T1 $900
Advertising $600
Total Monthly cost: $4000
To cover expenses you need 200 customers. However, at about 140 customers you are going to start running into busy signals and have to purchase a second T1 from the Telco. Add another 45 customers to break even.
This doesn't even cover staffing costs, etc. You can put about 220 dialup ports to a single T1. This translates to ~1700 customers before you need to bring in a second data T1. Using a caching server saves you about 20% on bandwidth. Expect to gain a maximum of 50 accounts per month for the first 6 months. You may do better, but your numbers should always be conservative. Figure that you are going to run through the rest of the 50K before you break even. Also, your first users are going to be the people who consider unlimited access to be a 7x24 connection. Until you get up to ~96 ports, your modem ratio is going to have to be about 6.75 to one. Above 96 ports, you can run ~8 to 1. Forget the old 10:1 ratio. People are spending more time online now. Once you have to go above 48 ports, you will have to buy more hardware for your access server. Usually another 10K.
Also realize that you are probably not going to pull enough users to become a target for a buy-out. You're not going to get rich running an ISP at this stage in the game. It's for the most part going to be a hobby, and a way for you to get T1 access for yourself. Consider the competition you are going up against. You need a hook to bring customers in. Is it custom programming? Online Gaming? Web Hosting? If you can't find a niche, your going to get creamed.
Sorry for the length, but I saw too much misinformation posted, and wanted to get some real facts out.
The problem is that when a chip is remarked, they have fake serial numbers on them. Unless you have an internal unique identifier, it's impossible to know that it's an original chip. (except for the badly done remarks)
Personally, I see a few places where this could come in handy. #1, it would allow single processor tracking from production to end user. You can check with Intel and see if the Pentium II-450 you just bought started life as a P2-400... A big problem with a number of small computer manufacturers. I've no problem with overclocking, but when a company tries to sell me a re-marked processor....
#2, as a method of tracking stolen processors. Buy your computer, check the ID with Intel, and if it's listed as stolen, go back to the store to get a refund...
I think something along the lines of a BIOS utility that allows you to write your ID out to floppy, would be a better way to go. Then you could have the choice of "installing" the ID or not.
For robotic changers, one of the best solutions I've seen is Arkeia by Knox Software. www.knox-software.com. It's commercial, and expensive, but worth the money if you need large-scale backup.
Big file systems are great when they work, but FSCK is a nightmare. We have 24 9.1 Gig Drives on a three channel DPT controller. We tried it as a single volume, but weird things happened with the DPT firmware.
We originally used this as a Usenet news server. We tried 24 seperate volumes to have the maximum number of spindles, but Linux has a limit of 16 SCSI drives in the 2.0 kernels. We ended up creating 12 2-drive stripe sets. (no redundancy) We then created 6 partitions. 5 that were 2 gigs in length, and one with the remainder. We used a patch to allow the partitions to be handled as 2 gig files. This was very fast, and had no FSCK issues as there were no file systems. If a few articles were mangled because of a crash.....
We ended up outsourcing our usenet service, and had this server to reuse. We created 3 volumes of 7 drives each, along with 3 hot spares. (One hot spare in each external drive chassis) Each volume is ~50 Gigs in size. One thing we have found is that if we HAVE to fsck the whole thing, (150 Gigs) you need about 4 hours. The PCI bus just doesn't have the bandwidth to check huge volumes in a reasonable time. We end up checking "/" and mounting it rw. We then mount the rest of the volumes "ro". We can then restart basic services (mail, web) and continue the fsck on the read-only volumes.
It's a balance you have to strike. If you really need that large of a file system, understand the time to restart. For us, just a basic reboot takes 12 minutes. With FSCK, it's ~4-5 hours of time to babysit. If you don't need that much space, look at setting up several individual file servers. It will help spread the load.
Slashdot Mirror
On top of that, their "2 hour window" is bull. I checked before I left home, and there wasn't anything there about reconfirming it. When I got to work (15 minutes later) it's already closed.
Read the original post. It's his "PAYING" customers that didn't pay up. As I read it, non-open source web hosting that he uses to keep the doors open.
From another ISP, it sounds like they are talking about the Cisco NetFlow software. This software is used to have more fine-grained control of a large network. It allows you to see what areas of the Internet are sending/receiving most of your data, and allows you to change the routing to what looks like sub-optimal, but is actually better. Shaping traffic on a per-flow basis is actually a very processor intensive process, so most companies would rather use controlled congestion...
You might want to consider offering access to the 1 meg network in an open manner, but surcharge bandwidth usage (K transfered) above a point. I assume you could use IP accounting to manage this. You could also set it up so that when a new user brings up a new machine, a program watching the ARP tables could see the dynamic ARP entry, add a firewall rule to block all requests but port 80 and DNS. Then redirect port 80 to a local page where they can "log in." After they log in, create a static ARP entry for them and tie the accounting for their IP. If you do a search, MIT has done something like this on their network.
Good Luck
Dan
AT&T has made it very clear that they intend to become a monopoly again. They plan to start offering cable telephone access (have tried it in CA) and are going to fight tooth and nail to keep any outsiders out... If you want cable telephone, you will HAVE to use AT&T Long Distance. If you want Net access, you HAVE to use AT&T's @HOME.
AT&T is acting like a phone company, but doesn't want to be regulated like a phone company. The Telecomunications Deregulation Act of 1996 stated that ALL phone companies have to open their networks to competition.
- My thoughts
Dan
The exemption that you are describing was put in place in the late 80's or early 90's. At the time, it was felt that modems and other data communications services needed to be protected. Since then, the telcos have been working to remove this restriction. Since Reed Hunt left the FCC, along with 5 other members of the FCC board, the FCC has been more than happy to do what ever the telco's want. There is NO leadership at the FCC at this time. They are only reacting to what is placed before them.
I realise that before the per-minute charges go into effect, congress will probably pass a law, but with all of the money the telcos can spread, do you think it will be able to get through congress?
We have run into the results of this already with GTE. GTE has declared us to be an IXC (Inter-eXchange-Carrier) and has prohibited us from purchasing any services that are local-tarrifs, including partial T1 circuits. I have 25 partial T1's already, but I can't purchase any more period. I can't move them, I can't change them, I can't touch them at all except to pull them out.
.03/minute = $2,610.00/day * 30 days per month = $78,300 per month in usage. No ISP in existance can afford to absorb those charges and still provide flat rate. Assuming a 10:1 user ratio, the ISP should have 1000 customers. The average customer would have a $79 usage portion of their bill along with the standard $20 base rate.
This ruling says that you can't be charged long distance on your PHONE BILL. However, the telco can go to court now and claim that since the FCC declared ISPs to be IXC's, they should be charged the same as Sprint, MCI, or other long distance carriers for each minute of access that the ISP uses on the TELCO network. Based on an ISP with 100 phone lines:
100 lines * 60 minutes= 6,000 line-minutes per hour
6,000 line-minutes * 5 hours of peak time (6pm-11pm) = 30,000 line minutes for peak
50% usage for the rest of the day, 3,000*19=57,000 line minutes non-peak
57,000+30,000=87,000 line minutes *
This ruling doesn't do the above, it just opens to door for a future COURT case to open it. This is a form of incrementalism. These rulings are only the opening moves on a chess board. Once the pieces are in place, one move and checkmate.
Not good.
Dan
My first bit of advice: Don't do it.
I'm assuming that you will ignore that bit of advice, so here's what I'd do if I was starting again.
#1, have $50,000 in capital available. Not necessarily cash in the bank, but liquid that you can draw on.
#2, choose your provider. Negotiate the best deal you can get with them. If you know Telco stuff, purchase your own T1 to the provider. If they have frame relay available, make use of it. Full T1 for resale should run between $1500 and $3000/month total (Port from ISP and T1 cost) depending on whether you have to cross LATA's or not.
#3, shop for your equipment.
For your core router, choose Cisco period. No flames please. I've used Livingston, Engage, Cisco, Bay, Emerging Technologies, and Sangoma. From what I've found:
Livingston is good for general use, but the IRX series doesn't support full duplex on the 10BaseT port or have support for bridge interfaces over frame relay. Almost a requirement if you want to support ADSL.
Engage is cheap, and includes a CSU with the router, but it's limited on what you can configure.
Not many Bay users out there to draw on for support.
The guy that builds Emerging Technologies boards for PC's has poor support for Linux, and doesn't want to talk to anything not running Cisco HDLC. (It will work with PPP over HDLC, but not stable.)
Sangoma has good support for Linux, but they are difficult to configure if you've never done it before.
Cisco seems to be the most flexible for different configurations.
For access equipment, either 3Com/US Robotics or Livingston Portmaster 3. Both are about $10,000. You bring in a digital T1 to them and you're done. If available, ask your Telco for ISDN PRI lines rather than DCS. PRI's are 23 lines, where DCS gives you 24 but PRI's allow your equipment to talk to the phone switch with out-of-band signaling rather than having to use the same channel for data and phone call. Just easier to troubleshoot.
I personally like the US Robotics equipment. V.90/X2 I think is a more reliable technology. If you're looking to connect a lot of low-end modems, you might want to go with a 56Kflex based modems in the Livingston box. Personal preference.
You might also want to look at Ascend equipment. They seem to have the best VoIP implementation right now if that's important to you.
For servers, you want at least 3 to start with.
#1, Primary Name Server/Backup Authentication server
#2, Secondary Name Server/Primary authentication server
#3, User E-Mail/Web Server
#4, Buy your billing software. If you don't have an automated billing system, you're GOING to fall behind, and start loosing money. Check out usenet "info.inet.access" for information on different billing packages. Don't scrimp on this step. You want online signup, automatic account creation, etc.
#5, Advertise. You've got the basics, so now you need to get customers to pay for the equipment.
Cost Breakdown:
Router and CSU/DSU $2500
Access Server $10,000
10MB Switched Ethernet hub $500
Servers w/Linux 3@$1500 each $4500
Install on T1 and Phone Lines $7500-$10,000
Billing software and server $4000
Total Startup cost: $29,000-$31,500
Monthly costs:
T1 to ISP $2500
Telco T1 $900
Advertising $600
Total Monthly cost: $4000
To cover expenses you need 200 customers. However, at about 140 customers you are going to start running into busy signals and have to purchase a second T1 from the Telco. Add another 45 customers to break even.
This doesn't even cover staffing costs, etc. You can put about 220 dialup ports to a single T1. This translates to ~1700 customers before you need to bring in a second data T1. Using a caching server saves you about 20% on bandwidth. Expect to gain a maximum of 50 accounts per month for the first 6 months. You may do better, but your numbers should always be conservative. Figure that you are going to run through the rest of the 50K before you break even. Also, your first users are going to be the people who consider unlimited access to be a 7x24 connection. Until you get up to ~96 ports, your modem ratio is going to have to be about 6.75 to one. Above 96 ports, you can run ~8 to 1. Forget the old 10:1 ratio. People are spending more time online now. Once you have to go above 48 ports, you will have to buy more hardware for your access server. Usually another 10K.
Also realize that you are probably not going to pull enough users to become a target for a buy-out. You're not going to get rich running an ISP at this stage in the game. It's for the most part going to be a hobby, and a way for you to get T1 access for yourself. Consider the competition you are going up against. You need a hook to bring customers in. Is it custom programming? Online Gaming? Web Hosting? If you can't find a niche, your going to get creamed.
Sorry for the length, but I saw too much misinformation posted, and wanted to get some real facts out.
Dan
The problem is that when a chip is remarked, they have fake serial numbers on them. Unless you have an internal unique identifier, it's impossible to know that it's an original chip. (except for the badly done remarks)
Personally, I see a few places where this could come in handy.
#1, it would allow single processor tracking from production to end user. You can check with Intel and see if the Pentium II-450 you just bought started life as a P2-400... A big problem with a number of small computer manufacturers. I've no problem with overclocking, but when a company tries to sell me a re-marked processor....
#2, as a method of tracking stolen processors. Buy your computer, check the ID with Intel, and if it's listed as stolen, go back to the store to get a refund...
I think something along the lines of a BIOS utility that allows you to write your ID out to floppy, would be a better way to go. Then you could have the choice of "installing" the ID or not.
My thoughts.
For robotic changers, one of the best solutions I've seen is Arkeia by Knox Software. www.knox-software.com. It's commercial, and expensive, but worth the money if you need large-scale backup.
Big file systems are great when they work, but FSCK is a nightmare. We have 24 9.1 Gig Drives on a three channel DPT controller. We tried it as a single volume, but weird things happened with the DPT firmware.
We originally used this as a Usenet news server. We tried 24 seperate volumes to have the maximum number of spindles, but Linux has a limit of 16 SCSI drives in the 2.0 kernels. We ended up creating 12 2-drive stripe sets. (no redundancy) We then created 6 partitions. 5 that were 2 gigs in length, and one with the remainder. We used a patch to allow the partitions to be handled as 2 gig files. This was very fast, and had no FSCK issues as there were no file systems. If a few articles were mangled because of a crash.....
We ended up outsourcing our usenet service, and had this server to reuse. We created 3 volumes of 7 drives each, along with 3 hot spares. (One hot spare in each external drive chassis) Each volume is ~50 Gigs in size. One thing we have found is that if we HAVE to fsck the whole thing, (150 Gigs) you need about 4 hours. The PCI bus just doesn't have the bandwidth to check huge volumes in a reasonable time. We end up checking "/" and mounting it rw. We then mount the rest of the volumes "ro". We can then restart basic services (mail, web) and continue the fsck on the read-only volumes.
It's a balance you have to strike. If you really need that large of a file system, understand the time to restart. For us, just a basic reboot takes 12 minutes. With FSCK, it's ~4-5 hours of time to babysit. If you don't need that much space, look at setting up several individual file servers. It will help spread the load.