What's next after 100mbps NICs
- Thread starter heng1028
- Start date
They have 1gbps and 10gpbs networking standards right now. However, the equipment for these standards is quite expensive. I saw a 1gbps NIC selling for $330 dollars.
is there any newer technologies for NIC?? not just increasing the transfer rate.
what i mean is new technology invented or whatever
what i mean is new technology invented or whatever
330 is not even that bad. Some of 1g adapters that I saw were like 600.
You can try wireless, but I doubt that there would be any cost efficiency. Err:cost efficiency in the near future. The main problem that we had when we installed our wireless here was that we had to pay for the freaking tower too. 20 feet one. Our pockets felt it all the way at the place where I work.
You can try wireless, but I doubt that there would be any cost efficiency. Err:cost efficiency in the near future. The main problem that we had when we installed our wireless here was that we had to pay for the freaking tower too. 20 feet one. Our pockets felt it all the way at the place where I work.
Or you could go with a couple Intel server NIC's and "team" them together.
It works pretty good. I'm using it right now.
Apparently you can team up to 4 ports together, though I'm only using 2.
It works pretty good. I'm using it right now.
Apparently you can team up to 4 ports together, though I'm only using 2.
the only thing is about dual nics is that your ping times s*ck. if you are playing a lot of multiplayer you might reconsider.
<< the only thing is about dual nics is that your ping times s*ck. >>
What makes you says that?
Mine got better. They dropped to almost half of what they were in UT.
I'm trying to set up dual NIC's in my 2000 server. They are both installed and shown to be working properly, but I've noticed throughput has gotten slower rather than faster. Is there a trick to make sure they work together instead of against each other?
In fact, Sandra shows no throughput with both nics hooked up, but one I disconnect one I'm fine.
"The main problem that we had when we installed our wireless here was that we had to pay for the freaking tower too"
20 feet? Shouldnt have cost much at all. What were your alternatives. Running fiber or a T1. Most likely it was considerable less expensive than those solutions. Wireless has its issues, for most, at least the customers I deal with, cost is THE plus. Efficient? Unless you lease the tower, which it sounds like you did not do, it is very efficient. One time cost. Mileage varies, and I am not starting an argument, I have just never heard an debate from one of our potential customers about the "added" cost of wireless. In fact, the business we get is more about saving money than the flexibility of wireless, which is another major selling point.
20 feet? Shouldnt have cost much at all. What were your alternatives. Running fiber or a T1. Most likely it was considerable less expensive than those solutions. Wireless has its issues, for most, at least the customers I deal with, cost is THE plus. Efficient? Unless you lease the tower, which it sounds like you did not do, it is very efficient. One time cost. Mileage varies, and I am not starting an argument, I have just never heard an debate from one of our potential customers about the "added" cost of wireless. In fact, the business we get is more about saving money than the flexibility of wireless, which is another major selling point.
I'd say that gigabit copper is probably the next thing that will trickle down to price levels low enough to entice home users, although the big bottleneck will clearly still be the pipe between you and the net (unless you're the kind of user who likes to do major data moving tasks like maybe pulling uncompressed video streams over the wire off your home server in the basement while you sit in the living room).
The next COOL thing, however, might be a 100 mbit signaling standard that will work over a single pair of copper, kinda like the Home PNA stuff on steroids.
The next COOL thing, however, might be a 100 mbit signaling standard that will work over a single pair of copper, kinda like the Home PNA stuff on steroids.
Chexi, if you're trying to double your throughput, then aggregating two NIC ports (sometimes also called "teaming"
whether by installing two aggregate-able NICs or using a single dual-port NIC is only part of the equation. You also need them to be plugged into a switch that will support port aggregation. Cisco calls this feature "Fast EtherChannel" on their switches that support it, and Nortel calls it "Multilink Trunking." I'm sure other manufacturers support it but I don't know what they call it.
If you think about how TCP/IP works, an IP address is only associated with a single MAC address and vice versa, so there is extra intelligence required both at the switch level and at the driver level on the NIC in order to get around that design "limitation" and let you increase your throughput. Actually, with the proper equipment you can usually team up to 4 NICs together in full duplex giving you 400mbit in and 400mbit out simultaneously. Pretty cool! Although you eventually have to start looking at cost. For some it might be cheaper to jump straight to gigabit ethernet.
In our production environment at my company, high availability is a must. Therefore, we do NIC teaming more for _redundancy_ purposes (if one NIC in a "team" fails, the driver knows to kick over to the other NIC) than for increasing throughput.
whether by installing two aggregate-able NICs or using a single dual-port NIC is only part of the equation. You also need them to be plugged into a switch that will support port aggregation. Cisco calls this feature "Fast EtherChannel" on their switches that support it, and Nortel calls it "Multilink Trunking." I'm sure other manufacturers support it but I don't know what they call it. If you think about how TCP/IP works, an IP address is only associated with a single MAC address and vice versa, so there is extra intelligence required both at the switch level and at the driver level on the NIC in order to get around that design "limitation" and let you increase your throughput. Actually, with the proper equipment you can usually team up to 4 NICs together in full duplex giving you 400mbit in and 400mbit out simultaneously. Pretty cool! Although you eventually have to start looking at cost. For some it might be cheaper to jump straight to gigabit ethernet.
In our production environment at my company, high availability is a must. Therefore, we do NIC teaming more for _redundancy_ purposes (if one NIC in a "team" fails, the driver knows to kick over to the other NIC) than for increasing throughput.
<< ." I'm sure other manufacturers support it but I don't know what they call it. >>
Link Aggregation - it's an 802.3x (x being variable, don't know exactly which letter) standard.
Yeah, Teaming is a god send to switched networks. Think about it. You have a full switched network with clients all attached at 100 megabit/full. Then all 3000 lan attached client hit a single server that is 100 megabit/full attached as well. Sure enough, that server NIC is overloaded. Sun has a great feature called "sun trunking" that is the same if not better than the wintel world. We use four gigabit adapters on our sun 10000s. Two pairs of adapters each are paired with each other to provide 2G tx and 2G rx. These "trunk pairs" are then in turn configured for different addresses. Pair A and Pair B have different IP addresses, then you use poor-man's load balancing (a.k.a. round robin DNS) to distribute load.
Next technologies? Packet over sonet has been a god send in the WAN arena, I know there are works on 10 gig ethernet. But really.....how fast do you need it to be? Networks are faster than the machines (yes, even your precious NT/2000 wintel boxes) being put on them. Then LAN is no longer a bottleneck in modern networks. It is the servers/storage/bus technologies that are truly limiting today's LANs.
spidey
ps - be on the prowl for terabit ethernet. Give it five years. It'll come.
Next technologies? Packet over sonet has been a god send in the WAN arena, I know there are works on 10 gig ethernet. But really.....how fast do you need it to be? Networks are faster than the machines (yes, even your precious NT/2000 wintel boxes) being put on them. Then LAN is no longer a bottleneck in modern networks. It is the servers/storage/bus technologies that are truly limiting today's LANs.
spidey
ps - be on the prowl for terabit ethernet. Give it five years. It'll come.
Wow, thanks Tondef. I do have a switch on my home LAN. It's the DSS-8+ D-Link. I'll check if it has this feature, but I doubt it. I don't have identical NIC's either, I was just experimenting with an extra one I had. My first NIC was the OEM 3com that is so raved about, can't remember the actual designation right now. The extra one I had is a Netgear 310TX, which may be the weak link in the chain.
The error I received, which should be no surprise to you because you nailed it, was no IP address found. I don't think what I did actually capitalizes on having 2 NIC's, but by adding IPX protocol on my server and my workstations, at least I now get benchmarks (which not surprisingly are exactly the same as a single NIC). To be honest, I don't really need to use a dual set-up as I only have 2 workstations on my server, and at most during a LAN party will have 8. But I figured as long as I have an extra NIC, I might as well play around a bit
The error I received, which should be no surprise to you because you nailed it, was no IP address found. I don't think what I did actually capitalizes on having 2 NIC's, but by adding IPX protocol on my server and my workstations, at least I now get benchmarks (which not surprisingly are exactly the same as a single NIC). To be honest, I don't really need to use a dual set-up as I only have 2 workstations on my server, and at most during a LAN party will have 8. But I figured as long as I have an extra NIC, I might as well play around a bit
I hope I didn't mislead anybody.
The Intel Server NIC's will only increase your thoroughput going out, since it balances the data amongst each NIC.
For receiving, or incoming, the data still can only use one NIC unless you have a Cisco (or compatable) switch, like TonDef said.
The Intel Server NIC's will only increase your thoroughput going out, since it balances the data amongst each NIC.
For receiving, or incoming, the data still can only use one NIC unless you have a Cisco (or compatable) switch, like TonDef said.
Chexi, if you've only got two workstations running off the server and you absolutely HAD to increase performance, AND you've got an extra 10/100 NIC sitting around, you could always ditch the switch and cross-connect each workstation to a dedicated NIC port on the server. That way each workstation has a dedicated pipe to the server. For ease of use at LAN parties, it's probably best to stick with the switch, though
Yeah, I think I'll be sticking with the Switch due to the LAN parties. Which, let's face it, is the reason I have a server in the first place. I would be happy with increasing only the output because it would make downloading of Unreal Tournament maps and skins to the clients that much faster, not that it isn't fast as it is once I figured out that you are supposed to change multiplayer settings to LAN instead of internet.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 19K
-
-
-
Facebook
Twitter