Of course a network card will withstand 15g's. As for the antenna, that's a different matter. However, at mach 3, that rocket's going to be well out of transmission range awfully quickly!
Of course, you can try using more directional transmitters - but if you go TOO narrow on the beam, then you lose reception if the rocket tilts even a small amount.
I've used a similar-sized monitor in 1280x1024. For regular computing, it's a waste of money - you could pick up a smaller monitor with a higher resolution, and be better off.
The two really good uses for a large monitor with that low of resolution are (a) computing for the visually impaired, and (b) watching movies.
It's a great idea. Those with more fuel-efficient cars, which pollute less and damage the roads less pay less. Those that want to live the "high on the hog" life style, using large, inefficient engines in huge, heavy vehicles pay more.
The more you abuse the roads and the ecosystem, the more you pay. I really don't see anything wrong with it. In fact, I'd rather see the gasoline tax raised to at least $2 per gallon. THEN we'd see American car companies bring their motors out of the 1960's, technology-wise. We'd also see people re-evaluating whether having a 5,500-pound house on wheels is really a *necessity* once they have two children.
When I was younger, I had plenty of "hands-on" experience. I think I got my first Radio Shack "XXX-in-one" kit at about 7 or 8 years old, and started building circuits then - both those in the books, and those of my own.
In high school, I took 2.5 years of electronics, which was partly theory, mostly hands-on. So far, so good. But it wasn't until I took some incredibly dry, boring, theoretical classes filled with lots of equations using calculus that I really made a big jump in my understanding, and hence, my abilities.
Hands-on is great if you're going to be an electrician. If you're going to be an engineer, though, you really need to understand the underlying, fundamental principles - and you just can't understand them without a lot of dry, boring theory and equations.
Your statement is quite ironic, seeing as how the carpentry that you mentioned is one of the very, very few areas where fractional measurement DOES have some strong merits over metric. : )
However, we've seen in the past when Microsoft *deliberately* stole entire programs wholesale, then only paid the authors when they were found guilty.
Anybody remember DOS 5.0? I remember looking at their new, nifty scandisk and defrag programs, and INSTANTLY recognizing them as a wholesale ripoff of Norton utilities. I remember thinking "How in the world can they get away with this?" Of course, later on, they were found guilty and forced to pay some royalties. However, in the mean time, they had stolen someone's programs in entirety, and used them not only to attract more customers, but to specifically keep those customers from buying the competing product!
Not all of MS's examples are that egregious, but they are certainly numerous. Like someone else said, proprietary systems have a control of the law *minus* financial incentives, opens source models have the law *plus* pride and accountability.
Think about if I was the CIO of a company and I'm going to be running my business on an operating system that has an intellectual property foundation that, by almost everyone's admission, is built on quicksand
What a load of crap. He's essentially saying that closed-source code is somehow more guaranteed to be more legitimate. I'd say that the reverse is true: There's a lot more incentive to do things legally when the entire world gets to see your source code than when virtually nobody does.
I would suspend any new Linux-related activities until this is all sorted out.
How unfortunate that he left out the "... and buy SCO instead."
We certainly have a point of contention regarding their interpretation of that contract.
Earlier he said that companies should abandon linux-related projects until SCO's suit is worked out. Now, he's implying that despite the fact that SCO is lying out of their teeth, and that Novell is calling them on it, noooooobody should abandon SCO-related projects. Sounds a bit hypocritical.
Here's what it really boils down to: SCO is the armpit of the Unix world, and the headquarters are in a city that's virtually the center of the MLM universe, and also known for immensely brain-dead business executives. SCO sucks, and is going downhill. Why? Because of Linux. Amazingly, Linux is also eating into Microsoft's server market share. Now the two team up, and decide to try and get rid of Linux. It's really not hard to see what's going on.
While on the surface, this might seem like a reinvention of IP tunnelling and VPN's, there are a couple of important features bundled in that set it apart:
1. It turns each node into a router. While you can establish a VPN with other tool kits, you still have to enable and configure the routing manually.
2. It's entirely user-land - it's a standalone program that a user can plop on their machine and be on their way.
The best part about it is that you can get through firewalls. The worst part about it is that you can get through firewalls.
Most people are pretty polar in their opinions of firewalls, with most of those people seeing them a fascist mechanism to control what they can see. In some (perhaps most) cases, that can be true. However, firewalls are much more than that: They can (and often are) used to protect YOU, the clueless end-user, from the other bad people on the Internet.
After I clear out counters on firewall rules, it's not uncommon to see 10-20 (sometimes more) incoming attacks within 5 seconds.
So, this will be great for letting people browse the web from work. On the other hand, it will expose them to propagation of worms and attacks which would have otherwise been caught by the firewall.
Is this a good program? Overall, I think that it's a good thing that NullSoft created it. We simply need to realize that with all of the benefits it brings, it will also bring a few negatvies with it.
I'll bet that the REAL reason they need more IP addresses is because most people already have nearly all of APNIC's address space black-holed.
On a more serious note, remember that APNIC (the organization in question) covers more than just China, Taiwan, Japan, etc. - it also covers Pacific nations such as New Zealand, Australia, etc..
All I'll have to do is a little DNS spoofing or hijacking, or inject a route into the router of my choice, and guess what? A whoooooole bunch of people just got their hard drives wiped out.
A man with a bad heart is walking along a busy sidewalk, brushes his arm against a woman's jacket, falls over in cardiac arrest.
How much do you think the lawsuit will be for?
Yes, the low amperage charge won't hurt a healthy person... but get a person with a bad ticker (maybe a few defect AV nodes to boot), and it can be *really* bad.
Shoot, you don't even have to have a heart attack. Pretty much *any* accidental touch that resulted in a shock would be grounds for a lawsuit. The actual damage that occurred would only increase the amount.
How hard would it be? I see you haven't been around long. 3DFX made a number of multi-processer video cards not very long ago. When they were going out of business, I picked up a dual-processer, 64 meg card for a ridiculously low price.
They even had some 4-processer (!) cards with 128 megs of memory on them - but they never really produced them in quantity.
Early on, you could put more than one 3dfx card in the machine, in SLI mode, and each card would render every other line. If I recall, their multi-processer cards worked the same way, but without the SLI cable.
... you can't just run the computer on a GPU: Because it isn't made for it. It's the age-old question of a general-purpose processer verses one made for a specific purpose.
GPU's can do astounding amounts of math because they're made specifically to handle certian types of math, and nothing else. CPU's can do much more, just not anywhere nearly as quickly.
GOOGLE can complain. By making it five times faster, they can spend:
-five times less on servers -five times less on power for the servers -five times less on data center real estate -five times less on cooling the data center -five times less on replacing dead hardware -much less on paying people to maintain the machines
The list doesn't stop there, either. The costs involved with running a high-traffic web site are very significant.
We went from a 4xP3/700 Xeon (1MB cache) to a 2xAthlonMP 2000+. The Xeon system, with 4-way memory interleaving, gave a theoretical bandwidth of 400 MB/second, with the Athlons having to fight for the 266 MHz memory bandwidth.
Overall, we went from 2800 MHz of P3 Xeon cycles to 3332 MHz of AthlonMP cycles, and lost most of the cache while we were at it. That alone certainly doesn't account for *twice* the performance - although increased memory bandwidth and lower SMP overhead (2 vs. 4 processers) does help explain some of it. Getting twice the performance on an RDBMS application with only a FOURTH of the cache is pretty impressive! I can't wait to get my hands on some Opterons, with 1MB cache and oodles of memory bandwidth.
Supposedly, the ServerWorks chipsets in the 4xP3 Xeon has a crossbar-type switch in front of the processers, so that each one can get 100 MHz to/from the memory at the same time. I haven't dug deep enough to verify that, but I'm not entirely convinced that it's true.
Just the other day, my boss was telling me that now when he and his wife stay in someone's guest bedroom, he's always nervous about whether there is an X10 camera hidden in there.
I suppose I could build him one of these, and he could have at least some idea of whether he's being watched or not...
ViaWest's principle connection into Utah is an OC3, at 155 megabits. If the DDOS was using 100 t1's worth (the full 155 megabits), then all of their other customers were also being effectively DDOS'd.
The other 10% that they mention are a couple of local t1's that ViaWest has in case of backup. Certainly not enough to supply data to all of their data center and other customers if their primary connection went down.
ViaWest DOES have a pretty good network for their size. I've been in their data center on a fast ethernet segment, and on a single download from Microsoft.com, had a sustained transfer rate of over 10 megaBYTES per second, nearly the entire 100 megabits! Getting that sort of bandwidth from a single download, from a site hosted several states away on a different network, is pretty darn decent.
Yes, a t1 for 160 people is too small. I know people in codos with less than 100 users on a t1, and it *sucks*.
Yes, 10 kbps is less than a 14.4 modem, as a 14.4 modem will reach up to.... 14.4 kbps. You're confusing bits with bytes. Here's a tip: You rarely, if EVER, see anything other than "bits" when talking about network equipment. 56k modem? 56 kiloBITS. T1? 1.544 megaBITS. T3? 45 megaBITS. Ethernet? 10 megaBITS. FE? 100 megaBITS. GE? 1 gigaBIT. I use a "56k" modem, I regularly get 35+ kbps.
Here's the real problem with a t1 for 160 people: While it *can* be done, in order to keep the few bandwidth hogs from ruining it for everyone, you need to limit each user's bandwidth to the point where it's really not worth even $20 per month. Think about it. Where *one* single user with a P2P app can eeeeasily saturate a t1, what are you going to do when you have only *five* people with P2P apps? With 20, 50, or 100 connections each, the person next door trying to check his online banking (through one single connection) is going to have an awfully hard time trying to compete for bandwidth.
So, what do you do? Let's say that you limit each person's bandwidth to 512k. No good, those five people still max out the line with ease. 256k? Well, they don't max it out, but they come awfully close. It won't take many people with legitimate traffic to fill the rest, and bam, things start to suffer. I won't even *touch* the subject of what happens when one person sets up an open relay/proxy.
Given the choice of paying $15 per month for a connection I'm not sure of, and paying a measly $40 for a *guaranteed minimum* of 768k, I know which one I'll take.
I also know which one MOST people will take. By the time they pay for a second phone line and an ISP, guess what! That's $40, and they're on a measly 56k connection. Give them the chance to spend it on a *real* connection, and you're in business.
I'm not blowing smoke out of my bung hole here, I'm speaking from experience.
First, forget DSL. It simply isn't reliable enough. Being out of connectivity is bad enough, having 100+ people mad at YOU because they're out of connectivity is even worse. Use a t1, multiple t1's, or a t3.
Here's the easy, cheap way to do it. Go over to ebay, buy yourself a Cisco 1720 with a Wic-1T-DSU card in it. Your t1 plugs into the WIC card, and ethernet port on the router plugs into your switch. You'll be able to do bandwidth limitting and port filtering as well. From there, the only question left is the distance involved to the condos, which would dictate the structure of the ethernet design.
There are a few flaws with that design: First, with everyone on the same L2, there's no end to the mischief that someone can cause. Second, virii capable of exploiting the "network neighborhood" will spread like wildfire.
If you want to do things a bit better, put a firewall/router in each building, and wire those back to your central distribution switch. The "router" can be a $40 machine from the thrift store, with a couple of 4-port ethernet cards in them. Each ethernet port can be on it's own subnet, with appropriate firewalling on a *per port* basis. That will help you prevent lots of accidental and intentional problems that can crop up.
Of course, with 160 units, 1 t1 is pretty small. That only guarantees each unit about 10 kbits/second, which is lees than a 14.4 modem. Of course, not everyone is going to be on at the same time, but even if 1/10th of the people are on, that only guarantees them about 100 kbits/second.
When you also look at the fact that some people will use as much bandwidth as possible, then it gets even harrier. Let's say that you can each individual's bandwidth at 256k, with bursts to 512k. That means that it only takes 6 people downloading ISO's, using their favorite P2P app, watching streaming porn, or anything else to really make the connection suck for everyone.
Shop around, and see if you can get a good deal on a larger connection. Not long ago, I was offered a full DS3 (45 mbits/sec!) from Broadwing for $6k per month. While $40 per month might sound high on a per-unit basis, remember that would *guarantee* 768 kilobits per resident! There are very few places you can get that sort of *guaranteed* bandwidth for $40 per month *anywhere*.
Perhaps from a marketting perspective, but certainly not from a technological perspective. We took a $13,000 quad P3-Xeon machine, replaced it with a $3,000 dual AthlonMP, and guess what - the loads dropped in *half*.
They were (and are) very good performers. Their only limitation was a memory bandwidth limit. AMD went to all of the trouble to give each AthlonMP it's own independent bus, but they never took the time to mate that with a dual-channel memory controller, so that each processer could actually *utilize* the entirety of it's bus. Even so, they were (and still are) very capable machines.
Slashdot Mirror
Of course a network card will withstand 15g's. As for the antenna, that's a different matter. However, at mach 3, that rocket's going to be well out of transmission range awfully quickly!
Of course, you can try using more directional transmitters - but if you go TOO narrow on the beam, then you lose reception if the rocket tilts even a small amount.
steve
I've used a similar-sized monitor in 1280x1024. For regular computing, it's a waste of money - you could pick up a smaller monitor with a higher resolution, and be better off.
The two really good uses for a large monitor with that low of resolution are (a) computing for the visually impaired, and (b) watching movies.
steve
It's a great idea. Those with more fuel-efficient cars, which pollute less and damage the roads less pay less. Those that want to live the "high on the hog" life style, using large, inefficient engines in huge, heavy vehicles pay more.
The more you abuse the roads and the ecosystem, the more you pay. I really don't see anything wrong with it. In fact, I'd rather see the gasoline tax raised to at least $2 per gallon. THEN we'd see American car companies bring their motors out of the 1960's, technology-wise. We'd also see people re-evaluating whether having a 5,500-pound house on wheels is really a *necessity* once they have two children.
When I was younger, I had plenty of "hands-on" experience. I think I got my first Radio Shack "XXX-in-one" kit at about 7 or 8 years old, and started building circuits then - both those in the books, and those of my own.
In high school, I took 2.5 years of electronics, which was partly theory, mostly hands-on. So far, so good. But it wasn't until I took some incredibly dry, boring, theoretical classes filled with lots of equations using calculus that I really made a big jump in my understanding, and hence, my abilities.
Hands-on is great if you're going to be an electrician. If you're going to be an engineer, though, you really need to understand the underlying, fundamental principles - and you just can't understand them without a lot of dry, boring theory and equations.
steve
Your statement is quite ironic, seeing as how the carpentry that you mentioned is one of the very, very few areas where fractional measurement DOES have some strong merits over metric. : )
steve
However, we've seen in the past when Microsoft *deliberately* stole entire programs wholesale, then only paid the authors when they were found guilty.
Anybody remember DOS 5.0? I remember looking at their new, nifty scandisk and defrag programs, and INSTANTLY recognizing them as a wholesale ripoff of Norton utilities. I remember thinking "How in the world can they get away with this?" Of course, later on, they were found guilty and forced to pay some royalties. However, in the mean time, they had stolen someone's programs in entirety, and used them not only to attract more customers, but to specifically keep those customers from buying the competing product!
Not all of MS's examples are that egregious, but they are certainly numerous. Like someone else said, proprietary systems have a control of the law *minus* financial incentives, opens source models have the law *plus* pride and accountability.
steve
What a load of crap. He's essentially saying that closed-source code is somehow more guaranteed to be more legitimate. I'd say that the reverse is true: There's a lot more incentive to do things legally when the entire world gets to see your source code than when virtually nobody does.
How unfortunate that he left out the "... and buy SCO instead."
Earlier he said that companies should abandon linux-related projects until SCO's suit is worked out. Now, he's implying that despite the fact that SCO is lying out of their teeth, and that Novell is calling them on it, noooooobody should abandon SCO-related projects. Sounds a bit hypocritical.
Here's what it really boils down to: SCO is the armpit of the Unix world, and the headquarters are in a city that's virtually the center of the MLM universe, and also known for immensely brain-dead business executives. SCO sucks, and is going downhill. Why? Because of Linux. Amazingly, Linux is also eating into Microsoft's server market share. Now the two team up, and decide to try and get rid of Linux. It's really not hard to see what's going on.
steve
While on the surface, this might seem like a reinvention of IP tunnelling and VPN's, there are a couple of important features bundled in that set it apart:
1. It turns each node into a router. While you can establish a VPN with other tool kits, you still have to enable and configure the routing manually.
2. It's entirely user-land - it's a standalone program that a user can plop on their machine and be on their way.
The best part about it is that you can get through firewalls. The worst part about it is that you can get through firewalls.
Most people are pretty polar in their opinions of firewalls, with most of those people seeing them a fascist mechanism to control what they can see. In some (perhaps most) cases, that can be true. However, firewalls are much more than that: They can (and often are) used to protect YOU, the clueless end-user, from the other bad people on the Internet.
After I clear out counters on firewall rules, it's not uncommon to see 10-20 (sometimes more) incoming attacks within 5 seconds.
So, this will be great for letting people browse the web from work. On the other hand, it will expose them to propagation of worms and attacks which would have otherwise been caught by the firewall.
Is this a good program? Overall, I think that it's a good thing that NullSoft created it. We simply need to realize that with all of the benefits it brings, it will also bring a few negatvies with it.
steve
I'll bet that the REAL reason they need more IP addresses is because most people already have nearly all of APNIC's address space black-holed.
On a more serious note, remember that APNIC (the organization in question) covers more than just China, Taiwan, Japan, etc. - it also covers Pacific nations such as New Zealand, Australia, etc..
steve
All I'll have to do is a little DNS spoofing or hijacking, or inject a route into the router of my choice, and guess what? A whoooooole bunch of people just got their hard drives wiped out.
Yep. This will sure make my life easier.
steve
Of course I won't, if I run all of my money through the microwave for a few seconds, first. : )
steve
A man with a bad heart is walking along a busy sidewalk, brushes his arm against a woman's jacket, falls over in cardiac arrest.
How much do you think the lawsuit will be for?
Yes, the low amperage charge won't hurt a healthy person... but get a person with a bad ticker (maybe a few defect AV nodes to boot), and it can be *really* bad.
Shoot, you don't even have to have a heart attack. Pretty much *any* accidental touch that resulted in a shock would be grounds for a lawsuit. The actual damage that occurred would only increase the amount.
steve
1. "Get certified in three weeks, and make $$$ as a programmer!"
2. The market demands features above all else, and wants them *now*.
It's the old adage: Quick, cheap, good: Pick any two.
steve
No, the interleaving he's talking about is for page refreshing, not actually interleaving multiple memory sets.
steve
How hard would it be? I see you haven't been around long. 3DFX made a number of multi-processer video cards not very long ago. When they were going out of business, I picked up a dual-processer, 64 meg card for a ridiculously low price.
They even had some 4-processer (!) cards with 128 megs of memory on them - but they never really produced them in quantity.
Early on, you could put more than one 3dfx card in the machine, in SLI mode, and each card would render every other line. If I recall, their multi-processer cards worked the same way, but without the SLI cable.
steve
GPU's can do astounding amounts of math because they're made specifically to handle certian types of math, and nothing else. CPU's can do much more, just not anywhere nearly as quickly.
steve
Don't buy them. Educate everyone you can, so that they won't buy them.
If they don't make money, they'll go away... just like Divx did.
steve
GOOGLE can complain. By making it five times faster, they can spend:
-five times less on servers
-five times less on power for the servers
-five times less on data center real estate
-five times less on cooling the data center
-five times less on replacing dead hardware
-much less on paying people to maintain the machines
The list doesn't stop there, either. The costs involved with running a high-traffic web site are very significant.
steve
We went from a 4xP3/700 Xeon (1MB cache) to a 2xAthlonMP 2000+. The Xeon system, with 4-way memory interleaving, gave a theoretical bandwidth of 400 MB/second, with the Athlons having to fight for the 266 MHz memory bandwidth.
Overall, we went from 2800 MHz of P3 Xeon cycles to 3332 MHz of AthlonMP cycles, and lost most of the cache while we were at it. That alone certainly doesn't account for *twice* the performance - although increased memory bandwidth and lower SMP overhead (2 vs. 4 processers) does help explain some of it. Getting twice the performance on an RDBMS application with only a FOURTH of the cache is pretty impressive! I can't wait to get my hands on some Opterons, with 1MB cache and oodles of memory bandwidth.
Supposedly, the ServerWorks chipsets in the 4xP3 Xeon has a crossbar-type switch in front of the processers, so that each one can get 100 MHz to/from the memory at the same time. I haven't dug deep enough to verify that, but I'm not entirely convinced that it's true.
steve
Just the other day, my boss was telling me that now when he and his wife stay in someone's guest bedroom, he's always nervous about whether there is an X10 camera hidden in there.
I suppose I could build him one of these, and he could have at least some idea of whether he's being watched or not...
steve
ViaWest's principle connection into Utah is an OC3, at 155 megabits. If the DDOS was using 100 t1's worth (the full 155 megabits), then all of their other customers were also being effectively DDOS'd.
The other 10% that they mention are a couple of local t1's that ViaWest has in case of backup. Certainly not enough to supply data to all of their data center and other customers if their primary connection went down.
ViaWest DOES have a pretty good network for their size. I've been in their data center on a fast ethernet segment, and on a single download from Microsoft.com, had a sustained transfer rate of over 10 megaBYTES per second, nearly the entire 100 megabits! Getting that sort of bandwidth from a single download, from a site hosted several states away on a different network, is pretty darn decent.
steve
Yes, a t1 for 160 people is too small. I know people in codos with less than 100 users on a t1, and it *sucks*.
Yes, 10 kbps is less than a 14.4 modem, as a 14.4 modem will reach up to.... 14.4 kbps. You're confusing bits with bytes. Here's a tip: You rarely, if EVER, see anything other than "bits" when talking about network equipment. 56k modem? 56 kiloBITS. T1? 1.544 megaBITS. T3? 45 megaBITS. Ethernet? 10 megaBITS. FE? 100 megaBITS. GE? 1 gigaBIT. I use a "56k" modem, I regularly get 35+ kbps.
Here's the real problem with a t1 for 160 people: While it *can* be done, in order to keep the few bandwidth hogs from ruining it for everyone, you need to limit each user's bandwidth to the point where it's really not worth even $20 per month. Think about it. Where *one* single user with a P2P app can eeeeasily saturate a t1, what are you going to do when you have only *five* people with P2P apps? With 20, 50, or 100 connections each, the person next door trying to check his online banking (through one single connection) is going to have an awfully hard time trying to compete for bandwidth.
So, what do you do? Let's say that you limit each person's bandwidth to 512k. No good, those five people still max out the line with ease. 256k? Well, they don't max it out, but they come awfully close. It won't take many people with legitimate traffic to fill the rest, and bam, things start to suffer. I won't even *touch* the subject of what happens when one person sets up an open relay/proxy.
Given the choice of paying $15 per month for a connection I'm not sure of, and paying a measly $40 for a *guaranteed minimum* of 768k, I know which one I'll take.
I also know which one MOST people will take. By the time they pay for a second phone line and an ISP, guess what! That's $40, and they're on a measly 56k connection. Give them the chance to spend it on a *real* connection, and you're in business.
I'm not blowing smoke out of my bung hole here, I'm speaking from experience.
steve
Do not worry about viruses
Deciding you're not going to worry about them is one thing. Dealing with 160 irate customers that *think* you should is entirely another.
How much do you want to spend on public IPv4?
I don't spend *anything* on public IPv4. I justify the usage to my upstream provider, and get all that I need, for free.
steve
First, forget DSL. It simply isn't reliable enough. Being out of connectivity is bad enough, having 100+ people mad at YOU because they're out of connectivity is even worse. Use a t1, multiple t1's, or a t3.
Here's the easy, cheap way to do it. Go over to ebay, buy yourself a Cisco 1720 with a Wic-1T-DSU card in it. Your t1 plugs into the WIC card, and ethernet port on the router plugs into your switch. You'll be able to do bandwidth limitting and port filtering as well.
From there, the only question left is the distance involved to the condos, which would dictate the structure of the ethernet design.
There are a few flaws with that design: First, with everyone on the same L2, there's no end to the mischief that someone can cause. Second, virii capable of exploiting the "network neighborhood" will spread like wildfire.
If you want to do things a bit better, put a firewall/router in each building, and wire those back to your central distribution switch. The "router" can be a $40 machine from the thrift store, with a couple of 4-port ethernet cards in them. Each ethernet port can be on it's own subnet, with appropriate firewalling on a *per port* basis. That will help you prevent lots of accidental and intentional problems that can crop up.
Of course, with 160 units, 1 t1 is pretty small. That only guarantees each unit about 10 kbits/second, which is lees than a 14.4 modem. Of course, not everyone is going to be on at the same time, but even if 1/10th of the people are on, that only guarantees them about 100 kbits/second.
When you also look at the fact that some people will use as much bandwidth as possible, then it gets even harrier. Let's say that you can each individual's bandwidth at 256k, with bursts to 512k. That means that it only takes 6 people downloading ISO's, using their favorite P2P app, watching streaming porn, or anything else to really make the connection suck for everyone.
Shop around, and see if you can get a good deal on a larger connection. Not long ago, I was offered a full DS3 (45 mbits/sec!) from Broadwing for $6k per month. While $40 per month might sound high on a per-unit basis, remember that would *guarantee* 768 kilobits per resident! There are very few places you can get that sort of *guaranteed* bandwidth for $40 per month *anywhere*.
steve
The AthlonMP line was never effective?
Perhaps from a marketting perspective, but certainly not from a technological perspective. We took a $13,000 quad P3-Xeon machine, replaced it with a $3,000 dual AthlonMP, and guess what - the loads dropped in *half*.
They were (and are) very good performers. Their only limitation was a memory bandwidth limit. AMD went to all of the trouble to give each AthlonMP it's own independent bus, but they never took the time to mate that with a dual-channel memory controller, so that each processer could actually *utilize* the entirety of it's bus. Even so, they were (and still are) very capable machines.
steve