1 members (Scott35),
161
guests, and
30
robots. |
Key:
Admin,
Global Mod,
Mod
|
|
|
Joined: May 2005
Posts: 247
Member
|
The specified maximum distance for a cat5 segment is 100m. This includes 90m of structured wiring, and 10m of patch cables. The 100m segment will support the full bandwidth.
People have successfully exceeded 100m with full duplex links between switches, but it is not recommended as a standard practice.
If you are running a link between buildings, it is recommended that you run fiber to eliminate any possible differential in ground voltages between endpoints.
|
|
|
|
Joined: Jul 2004
Posts: 625
Member
|
Techie, It was my understanding that 300' would be the cutoff. I would expect that even if we are close we will just have less bandwidth. Witch compaired to what is going on now would be a hughe improvment.
ob If the link fails, it won't be a "just have less bandwidth" failure--it will fail by having a high error rate, which results in dropped packets, which results in end-to-end retransmissions. This is a Very Bad Thing(tm). However, 285' shouldn't be a problem. As techie pointed out, the maximum length of a 100 megabit Cat5 segment is 100 meters, or around 330'. Assuming that the everything is installed properly, that should work with no problem. Using Cat6 cable will probably provide you a bit of extra insurance, but isn't really necessary, because the connection is designed to operate flawlessly across 100m of Cat5 cable. To expand a bit on one of techie's comments, 100m is a hard limit for a half-duplex link. This is because of the "slot time" built into the Ethernet retransmission algorithm. If you make the length longer than that, you will have collisions that aren't detected, resulting in lost packets. Full duplex, on the other hand, doesn't use use the collision detection and backoff algorithm, so there isn't a hard limit as there is with half-duplex. So one can stretch beyond the 100m spec, but it is definitely not recommended practice. I note that normally hubs are half-duplex, and switches are full-duplex. techie, I agree that fiber isn't a bad idea between buildings. However, the original Ethernet installation at PARC saw around a 90V difference in ground potential from one end of the coax to the other, so we specified something like 2000V isolation in the coupling transformers when we wrote the standard. This isolation was carried over into the BaseT specs. So there really shouldn't be any problem using copper between two buildings, unless there might be more than 2000V difference in ground potential between the buildings. (In which case you have much, much larger problems to worry about...)
|
|
|
|
Joined: Feb 2002
Posts: 114
Member
|
You CAN get over 100 Mbps at 300m on coax using Moca adapters. This may get pricey as the technology is pretty new. http://www.mocalliance.org/en/index.aspThe issue with all networking technologies is that you will almost certainly be violating the ANSI/CSA/UL 60950-1 listing when you run between buildings. The interface ports of most networking equipment are classified SELV and as such they are not subject to the overvoltage testing (power line cross) that is required in the US and Canada. In other coutries you will only be violating circuit separation clauses of the safety standard. When you run these indoor circuits outside, this make the interface a "cable distribution system" (rather than jsut a SELV circuit that happens to connect to coax) which adds testing requirements that may not have been considered during the NRTL investigation. If I was uncertain of the conditions of the product Listing (and unable to stop myself from proceeding with the installation anyway), I would install the equipment on and over a non-flamable surface and/or locate it inside a fire enclosure. I only say this cuz I know some people are just going to run stuff outside anyway. May as well make some effort to encourage them to do it the least wrong way possible. Ethernet will have the same violation of Listing issue.
|
|
|
|
Joined: May 2005
Posts: 167
Member
|
[To expand a bit on one of techie's comments, 100m is a hard limit for a half-duplex link. This is because of the "slot time" built into the Ethernet retransmission algorithm. If you make the length longer than that, you will have collisions that aren't detected, resulting in lost packets. Ethernet can run up to 500m on a half-duplex link (thick coax). There is a limit for the total network diameter to ensure that collisions are still detected, but there's no way you're going to exceed it with a single twisted pair link. I note that normally hubs are half-duplex, and switches are full-duplex. Hubs are always half-duplex. Switches can be either half or full duplex according to either the port settings or autonegotiation. One of the most annoying problems I've encountered is admins forcing the ports to full-duplex which breaks autonegotation. The end result of this is terrible performance unless you set the network card in the PC to also run in full-duplex. Unmanaged switches will not allow you to change the port duplex settings manually. They are autonegotiation-only.
Last edited by brianl703; 01/16/08 06:17 PM.
|
|
|
|
Joined: Jul 2004
Posts: 625
Member
|
Ethernet can run up to 500m on a half-duplex link (thick coax). There is a limit for the total network diameter to ensure that collisions are still detected, but there's no way you're going to exceed it with a single twisted pair link.
I should have clarified that the 100 meter limit is for 100 megabit and gigabit Ethernet. For 10 megabit Ethernet the limit is 500 meters per backbone segment, but you can go something like 2500m (IIRC) end-to-end, through repeaters. It was designed so that you could have a 1500m fiber segment bridging between two "ethers." Hubs are always half-duplex. Switches can be either half or full duplex according to either the port settings or autonegotiation. One of the most annoying problems I've encountered is admins forcing the ports to full-duplex which breaks autonegotation. The end result of this is terrible performance unless you set the network card in the PC to also run in full-duplex.
Unmanaged switches will not allow you to change the port duplex settings manually. They are autonegotiation-only.
Yes, you are quite correct that switches can support half-duplex. I was thinking in terms of going switch to switch for the OP's situations, but I did omit that most switches can operate either way.
|
|
|
|
Joined: May 2005
Posts: 247
Member
|
Ethernet can run up to 500m on a half-duplex link (thick coax). There is a limit for the total network diameter to ensure that collisions are still detected, but there's no way you're going to exceed it with a single twisted pair link.
I should have clarified that the 100 meter limit is for 100 megabit and gigabit Ethernet. For 10 megabit Ethernet the limit is 500 meters per backbone segment, but you can go something like 2500m (IIRC) end-to-end, through repeaters. It was designed so that you could have a 1500m fiber segment bridging between two "ethers." Actually, the 100 meter limit is for all twisted pair based Ethernet, 10baseT, 100baseT, and 1000baseT. (the following is mostly of historical interest, since no one actually builds shared networks of this size any more.) For 10base5 (Thicknet) the limit is 500 meters per segment. For 10base2 (Thinnet) the limit is 185 meters per segment. The 5-4-3 rule applies. Between any two stations in a shared Ethernet network, you may have up to 5 segments, with 4 repeaters (hubs are considered multi-port repeaters), but only 3 segments may be populated with stations. If you are using 10base5, this means that there can be a maximum of 3000 meters between end stations before a router or bridge is required. 3000 meters is based on 5x 500m 10base5 coax segments, and 10x 50m AUI transceiver cables (ie: [HOST]- 50m-(X)- 500m-(X)- 50m- [REPEATER]- 50m-(X)- 500m-(X)- 50m- [REPEATER]- 50m-(X)- 500m-(X)- 50m- [REPEATER]- 50m-(X)- 500m-(X)- 50m- [REPEATER]- 50m-(X)- 500m-(X)- 50m- [HOST]If you need to extend the network beyond these limits, you need to segment the collision domain with a router or a bridge (a switch is essentially a multi-port bridge). A DEC remote repeater uses up to 1000 meters of fiber optic cabling to connect two network segments. FOIRL (Fiber Optic Inter Repeater Link) also has a 1000 meter limitation. 10base-FL, which replaces FOIRL, has a 2000 meter limit. I believe that you are allowed to replace one segment with a fiber optic remote repeater, based on the old DEC configuration guidelines.
|
|
|
|
Joined: Apr 2003
Posts: 362
OP
Member
|
Hi Guys Thanks for all your help! I installed two cat 5e cables Fri and it works great measured 290 feet to new switch. The IT guy is very happy!
Ob
Choose your customers, don't let them choose you.
|
|
|
|
Joined: May 2005
Posts: 167
Member
|
For 100baseT, there is apparently a hard 100 meter limit that is due to collision timing issues, in addition to signal strength issues.
For 10baseT, the 100 meter limit is due to signal strength issues ONLY. Modern cat5e cable has less attenuation than does the cable for which 10baseT was originally speced (which is not even cat3 cable, but just plain old telco grade twisted pair, like what might have been used to wire a Merlin phone phone system in 1985) and therefore it is apparently possible to run 10baseT up to 150 meters on cat5e cable without issue.
|
|
|
|
Joined: May 2005
Posts: 247
Member
|
For 100baseT as a shared media (ie: using a hub), there is a hard limit due to timing. There is also a limitation on the number of hubs in the network, again due to timing issues.
The basic problem is that the frame must be able to reach the furthest point in the network and return before the originating station has finished transmitting the frame. The timing is based on the minimum frame size of 64 bytes.
The rules effectively limit a shared 100baseT network to a single hub driving 100m segments, or two "class II" hubs driving 100m segments, separated by up to 5m of cable linking the hubs.
I should mention that ethernet has been tested successfully over short lengths of barbed wire.
I should amend my posting above. IEEE 802.3 notes that the non-populated link segments are supposed to be full-duplex (twisted pair or fiber), and the length of the AUI cables used with fiber transceivers is 25m.
|
|
|
Posts: 404
Joined: March 2007
|
|
|
|