Either way your connectivity is going to be tied to a single exchange.
The only way to achieve real redundancy in your connectivity is to use two separate physical paths to two separate ISPs (i.e. SDSL to one and LMDS to another).
If you are hosting critical services on-site and you want inbound redundancy you should be talking BGP to your upstreams. Good luck convincing ADSL providers to agree to that.
Also, there might not be much point in having connections to two providers if they both terminate in a single router that could fail at any point. This means you should have a separate router for each connection, run EBGP to your upstreams, IBGP between the routers, and VRRP to make them appear as a single IP address on the internal interfaces.
Your scenario is wrong in so many ways. There are many real costs of running a large ISP network you are not taking into account: 1) transit costs - unless they are tier1, ISPs have to buy transit (internet access) from other providers... give Level (3) a call and ask how much 2gbps of transit costs 2) backhaul costs - ISPs have to pay for connectivity from the exchange to their network (again, very expensive) 3) engineering staff costs - several $150k/year CCIEs to manage your network, manage BGP with upstreams, providers and peers 4) support staff costs - hundreds of support staff to take "someone stole my internet" calls from customers 5) colocation costs - it is not feasible to build a facility everywhere an ISP has a POP so they have to lease colo space 6) hardware costs - ISP customers expect services like mail, DNS forwarding, web hosting space
Your example doesn't even take into account the true costs of network equipment. The ISP needs aggregation devices (switches) and local endpoints (routers). They need core routers and border routers (each device can be hundreds of thousands of dollars). All these routers need to be placed in pairs for failover.
Maybe the government should sieze control over the main backbone and make the upkeep/upgrade no longer a responsibility of the major providers. ISP's would all compete for the last mile hookups/billing, allowing other companies in who don't already own part of the highway itself.
This is the fun situation we currently have in Australia. Telstra owns most of the backhaul fibre across the country and from DSL exchanges. They refuse to peer with other ISPs and generally charge outrageous rates for any wholesale services provided to ISPs.
They are currently trying to get $5 billion out of the federal government to run Fibre to the Node, so that they can deliver an *amazing* 6mbps to most households. The fact that other ISPs are currently running 24mbps using existing copper infrastructure just fine is lost on them.
I don't know where to start on this one.
Firstly, you need to assess what your exact bandwidth requirements are. This is not difficult - just MRTG graphs and some reporting from whatever firewall / routing equipment you're using right now.
Secondly, you need to figure out what sort of cabling is in your area. As others above have noted, the best solution is an ethernet VLAN over fibre between you and your provider. You could start out at 10mbps and scale up to 100mbps easily if you need to. If there is no fibre running in your area then you'll be stuck with a T1 (but if you are currently fractional, why not upgrade to full?).
Thirdly, you need to make sure that you have the routing equipment to make this work. I can't think of any device with a "nice pleasant UI that doesn't cause me great grief" that could actually forward packets at T3 speeds, let alone 100mbps. Serious equipment is configured from the command line, if you can't figure out the command line then...
Lastly, you'll sound a lot smarter if you don't throw around terms like "redundant OC-48 runs". Is it possible that you will ever need Nx2.4gbps of connectivity? Do you know how much this would cost? Do you know how much the routing gear to handle this would cost? There are a lot of ISPs that don't use anywhere near that much transit bandwidth.
It looks like the cellphone network isn't the only one having trouble: Reuters' web server has ground to a halt and is throwing ASP.Net Server Too Busy errors.
http://today.reuters.com/
I think part of the trick is that it's a very flat support organization - you don't need to escalate to a level 2 or level 3 person on the phone.
This sounds good in theory but in practice it is very expensive to have smart technicians explaining to the AOL crowd that their modem needs to be plugged into the phone line. In any case no smart technician is going to spend 8 hours/day on soul crushing basic technical support.
Dell, on the other hand, makes people jump through hoops when they call in with a problem (like a dead hard drive). This even happens on corporate accounts - the field techs at work have been known to spend 4 hours on the phone going through dell's script.
Dell's phone support should be held up as an example of the worst possible offering. But the actual technicians that come on-site to do repairs always seem to be professional, friendly and extremely capable.
And in other news, the RIAA has sued ARIA in the Australian Federal Court. The USA based RIAA is suing its Australian counterpart, claiming that ARIA violated its copyrighted idea of filing fucking stupid lawsuits. SCO is reportedly also considering taking similar action.
Google has spent years developing its PageRank technology that they call the "heart of our software". Given that Google has patented PageRank, it is hard to imagine how Amazon will be able to produce anything that will be of a competitive nature to Google.
Froogle hasn't been officially released (notice the little "Beta" in the corner of the logo), but I think that it will be a far more comprehensive online shopping service than anything that Amazon could use.
It seems likely to me that Google could use its place in the "Search Engine" market to expand into the "Search Engine for Specific Products/Services" market. It seems highly unlikely that Amazon will be able to use its place as an Online Retailer to springboard themselves into the search engine market.
This to me is a sure sign of a guy stuck in academia.
indicate that the knucklehead around here is not Philip Greenspun.
Greenspun may be a smartass, and possibly an egomaniac, but his technical credentials are first rate. Let's have a look at his achievements over the last decade:
Teaching Computer Science at MIT. How many of you would be asked to teach Computer Science at the most prestigious tech school on Earth?
Founded ArsDigita, and wrote large tracts of the ArsDigita Community System in TCL. That company and product became the foundation for many Fortune 500 companies' websites.
Through his evangelism, course developments at MIT, and the ArsDigita University, helped to bring web courses like MIT's 6.171 to mainstream university teaching.
Created and developed Photo.net, which averages 6 million hits/day. See more statistics here.
The above are why I question the intelligence of people who post comments like:
Why do we care about his opinion? Did he build any large scale scalable web application? Was he involved in maintaining and supporting any commerical projects? Did he make any critical insight of the strength and weakness of these development tools? Sounds like he draw his conclusion mostly from some student project. In that case I would say Microsoft wins hands down. Microsoft is very good at making and packaging complete development tools and is relative easy for novice to learn and use. Does it make VB the best programming language over other alternatives? I won't make that conclusion.
So to answer these supposedly rhetorical questions:
We do care about his opinion.
He has built more large scale scalable web applications the many of us will in a life time.
He was CEO of a company that developed and supported commercial projects for Fortune 500 clients.
He has spent the last decade making critical insights into everything.
Let's make our own conclusions then.
Slashdot Mirror
Apparently this site will be "127.0.0.1 FOR IT PROS". This, one presumes, is what passes for clever humour in the Microsoft world.
Either way your connectivity is going to be tied to a single exchange.
The only way to achieve real redundancy in your connectivity is to use two separate physical paths to two separate ISPs (i.e. SDSL to one and LMDS to another).
If you are hosting critical services on-site and you want inbound redundancy you should be talking BGP to your upstreams. Good luck convincing ADSL providers to agree to that.
Also, there might not be much point in having connections to two providers if they both terminate in a single router that could fail at any point. This means you should have a separate router for each connection, run EBGP to your upstreams, IBGP between the routers, and VRRP to make them appear as a single IP address on the internal interfaces.
Your scenario is wrong in so many ways. There are many real costs of running a large ISP network you are not taking into account:
1) transit costs - unless they are tier1, ISPs have to buy transit (internet access) from other providers... give Level (3) a call and ask how much 2gbps of transit costs
2) backhaul costs - ISPs have to pay for connectivity from the exchange to their network (again, very expensive)
3) engineering staff costs - several $150k/year CCIEs to manage your network, manage BGP with upstreams, providers and peers
4) support staff costs - hundreds of support staff to take "someone stole my internet" calls from customers
5) colocation costs - it is not feasible to build a facility everywhere an ISP has a POP so they have to lease colo space
6) hardware costs - ISP customers expect services like mail, DNS forwarding, web hosting space
Your example doesn't even take into account the true costs of network equipment. The ISP needs aggregation devices (switches) and local endpoints (routers). They need core routers and border routers (each device can be hundreds of thousands of dollars). All these routers need to be placed in pairs for failover.
Maybe the government should sieze control over the main backbone and make the upkeep/upgrade no longer a responsibility of the major providers. ISP's would all compete for the last mile hookups/billing, allowing other companies in who don't already own part of the highway itself.
This is the fun situation we currently have in Australia. Telstra owns most of the backhaul fibre across the country and from DSL exchanges. They refuse to peer with other ISPs and generally charge outrageous rates for any wholesale services provided to ISPs.
They are currently trying to get $5 billion out of the federal government to run Fibre to the Node, so that they can deliver an *amazing* 6mbps to most households. The fact that other ISPs are currently running 24mbps using existing copper infrastructure just fine is lost on them.
Actually, Google doesn't pay for a lot of its traffic because of their extensive peering arrangements with ISPs around the world.
I don't know where to start on this one. Firstly, you need to assess what your exact bandwidth requirements are. This is not difficult - just MRTG graphs and some reporting from whatever firewall / routing equipment you're using right now. Secondly, you need to figure out what sort of cabling is in your area. As others above have noted, the best solution is an ethernet VLAN over fibre between you and your provider. You could start out at 10mbps and scale up to 100mbps easily if you need to. If there is no fibre running in your area then you'll be stuck with a T1 (but if you are currently fractional, why not upgrade to full?). Thirdly, you need to make sure that you have the routing equipment to make this work. I can't think of any device with a "nice pleasant UI that doesn't cause me great grief" that could actually forward packets at T3 speeds, let alone 100mbps. Serious equipment is configured from the command line, if you can't figure out the command line then... Lastly, you'll sound a lot smarter if you don't throw around terms like "redundant OC-48 runs". Is it possible that you will ever need Nx2.4gbps of connectivity? Do you know how much this would cost? Do you know how much the routing gear to handle this would cost? There are a lot of ISPs that don't use anywhere near that much transit bandwidth.
It looks like the cellphone network isn't the only one having trouble: Reuters' web server has ground to a halt and is throwing ASP.Net Server Too Busy errors. http://today.reuters.com/
I think part of the trick is that it's a very flat support organization - you don't need to escalate to a level 2 or level 3 person on the phone.
This sounds good in theory but in practice it is very expensive to have smart technicians explaining to the AOL crowd that their modem needs to be plugged into the phone line. In any case no smart technician is going to spend 8 hours/day on soul crushing basic technical support.
Dell, on the other hand, makes people jump through hoops when they call in with a problem (like a dead hard drive). This even happens on corporate accounts - the field techs at work have been known to spend 4 hours on the phone going through dell's script.
Dell's phone support should be held up as an example of the worst possible offering. But the actual technicians that come on-site to do repairs always seem to be professional, friendly and extremely capable.
$60 a barrel oil? What $60 a barrel oil? Must be nice not to have to pay your electricity bills...
Yeah, his parents sure are going to be pissed when they get that first electricity bill....
"including totally non-gratuitous photos of naked people using technology"
As if photos of naked people could ever be gratuitous to Slashdot readers!
And in other news, the RIAA has sued ARIA in the Australian Federal Court. The USA based RIAA is suing its Australian counterpart, claiming that ARIA violated its copyrighted idea of filing fucking stupid lawsuits. SCO is reportedly also considering taking similar action.
Google has spent years developing its PageRank technology that they call the "heart of our software". Given that Google has patented PageRank, it is hard to imagine how Amazon will be able to produce anything that will be of a competitive nature to Google.
Froogle hasn't been officially released (notice the little "Beta" in the corner of the logo), but I think that it will be a far more comprehensive online shopping service than anything that Amazon could use.
It seems likely to me that Google could use its place in the "Search Engine" market to expand into the "Search Engine for Specific Products/Services" market. It seems highly unlikely that Amazon will be able to use its place as an Online Retailer to springboard themselves into the search engine market.
-
Teaching Computer Science at MIT. How many of you would be asked to teach Computer Science at the most prestigious tech school on Earth?
-
Founded ArsDigita, and wrote large tracts of the ArsDigita Community System in TCL. That company and product became the foundation for many Fortune 500 companies' websites.
-
Through his evangelism, course developments at MIT, and the ArsDigita University, helped to bring web courses like MIT's 6.171 to mainstream university teaching.
-
Wrote two of the best technical books I have ever read, and put them up for free in HTML on his website:
Internet Application Workbook and Philip and Alex's Guide to Web Publishing
-
Created and developed Photo.net, which averages 6 million hits/day. See more statistics here.
The above are why I question the intelligence of people who post comments like: So to answer these supposedly rhetorical questions: We do care about his opinion. He has built more large scale scalable web applications the many of us will in a life time. He was CEO of a company that developed and supported commercial projects for Fortune 500 clients. He has spent the last decade making critical insights into everything. Let's make our own conclusions then.