Need fast answer on max length of cables

[Paul Victorey]

Working to design a 100 Mb network that will consist of a central switch, connected to a second switch (connected to 2 or 4 hubs) and the central switch will also be connected to about 6 to 8 hubs.

My question: What is the max length of UTP cat5 cable I can run (under Fast Ethernet standards) between the central switch and the stuff that it connects to?

If I have the same 100m as between PC and hub, I won't need to run fiber.

Also, what is the difference in price between a switch that will accept fiber connections and one that will use cat5, and the difference in price between a hub that allows fiber uplink and one that won't? I can actually run 2 stranded multimode fiber for the exact same cost as Cat5 UTP.

------------------
Paul M. Victorey
------------------
I am not responsible for any problems that may arise as a result of following my advice. This includes, but is not limited to, computer failure, loss of data, nuclear war, famine, boils, no clean laundry, your daughter running off with a biker gang, or armageddon. Take my advice at your own risk.

[HAL9000]

I do believe that it is still 100M.

------------------
If it ain't broke, you're not pushing hard enough!

[Omletteboy]

Yup, cat 5 max length is 100m

[Paul Victorey]

Even between a switch and a hub?

I had heard that hub to hub would only allow 33 meters... but then this might be between hubs in a stack?

------------------
Paul M. Victorey
------------------
I am not responsible for any problems that may arise as a result of following my advice. This includes, but is not limited to, computer failure, loss of data, nuclear war, famine, boils, no clean laundry, your daughter running off with a biker gang, or armageddon. Take my advice at your own risk.

[mbossman2]

if given a chance to run either cat5 or MM fiber, i would most definitely run fiber.

the reason for that is that cat5 is realistically capped at 100mb (i know there is a standard for gigabit over copper but that is EXTREMELY expensive as it is right now only available in high end core switches and the price delta between the fiber gigabit blade and copper gigabit blade is approx 50%) and fiber has no upper limit to speed.

so with fiber you can run 10/100/1000 mbps and probably higher with no additional wire costs.

as to the price delta with switches with fiber and those that can't might be 10-30%, factor in the cost of rerunning fiber in a year or 2 and it all evens out.

[Great_One]

As far as the fiber vs. cat5, several points:

Speed will be the same on either media: 100mb
Fiber would be a better choice if you thought that someday you may move to gigabit ethernet on the uplinks. It supports higher band widths, longer distances,and is immune to interference but does cost more than cat5.

Fiber requires special tools to terminate and test. Their are 3 different types of connectors available so you need to match them up to the connectors on your switch. Probably wont be a do-it-yourself project.

Fiber ports on switches work the same way as the cat5 rj45 ports, just a different connection. Are both ethernet ports.
And of course you pay more for fiber ports on switches.

[Nutrocker]

There are some hubs and switches whose uplink ports only work 5mtrs or under! The manufacturers thought "these will only be used in cabinets near each other!"

Other than that 100Mtrs it is.



------------------
Regards
Don
<A HREF="http://
http://www.nutrocker.co.uk" TARGET=_blank>
http://www.nutrocker.co.uk</A>

[bob]

Great_One - hope to see more posts. You clearly know your stuff.

[Dave22]

Input on your pricing query:

There are switches that come with the option of installing a gigabit uplink module (SC or ST). You could always get one like that, so if you decide to utilize a fiber gig link in the future, you would just purchase the module instead of dishing out all the money up front. The other option may be to get a switch that supports FEC or some other type of port aggregation to create a bigger pipe without having to go to fiber.

Just a thought.

[vitalstatistix]

few points
a) u need to plan out a network topology, like u want a star schema with singular link or multihoming or dual-homing. Again multi-homing could be based on leaf or mesh.

b) based on your design above you will need either a (CORE + Edge switching system) or an entry level network as others have described it.

I assume its a wide spread network and high-availabilty is a criteria.
have you done your network thru put estimation as in average throug-put and peak throughput. THose numbers will help you decide the switch you need. a) Switching fabric capacity (giga bits) and b) Switching Speed (million packets per second)

My suggestion a layer 3 switch like Xylan Omni S/R9 at the core and Omni-Stack on the edge. They all come with all possible interfaces that u will ever need.
Put you core links on Fibre, i wud suggest go Leaf on core-edge routes. Make use of Vlans.

any layer 4 switching requirment like Radius or LDAP authentication or DHCP services, IP Mobility between Vlans

Use SNMP to monitor the network and MRTG / Cricket to monitor the bandwidth. (MRTG on Linux is better and easier)

now the bottomline, what sort of budget do you have. That will define if u cud get a Ford or a Lamborghini Diablo.
cheers

gtfx

[Paul Victorey]

Actually, due to the fabulous communication in my office, although both my boss and I thought I was designing the network, someone (at a lower level than my boss in fact) already contracted and outsourced this work over a month ago, so it looks like this job's been taken from me. They should be running cable soon.

Now, I only need to MAINTAIN the network, I have no say in its construction.

------------------
Paul M. Victorey
------------------
I am not responsible for any problems that may arise as a result of following my advice. This includes, but is not limited to, computer failure, loss of data, nuclear war, famine, boils, no clean laundry, your daughter running off with a biker gang, or armageddon. Take my advice at your own risk.

[vitalstatistix]

Paul
ooooops! anywyas, now you get all the time to point fingers at that person / contractor, should anything go wrong! And bargain a good deal in your next appraisal :P

cheeers

[Paul Victorey]

Lol, yup, I get to blame the contractor for everything that goes wrong from this point on. http://www.pcmech.com/ubb/biggrin.gif

Although I was upset at having part of my job in essence taken from me, it really may be for the best, it allows me to work on other projects. So long as the other guy isn't TOO inept I can make it work http://www.pcmech.com/ubb/wink.gif

------------------
Paul M. Victorey
------------------
I am not responsible for any problems that may arise as a result of following my advice. This includes, but is not limited to, computer failure, loss of data, nuclear war, famine, boils, no clean laundry, your daughter running off with a biker gang, or armageddon. Take my advice at your own risk.

[bob]

Paul Victorey - know how you feel. Just designed a whole network - 1500 workstations- GIG links to servers - GIG to switches. 10/100 to workstations. Level 3 hardware VPN.

Project is on hold - to costly.

I could have designed a very cheep network - but they wanted the best with full redundacy.

Go figure....

[mbossman2]

you are not alone. i sell networking gear and get these requests for designs, do them, give them a price and they act like they should be able to implement this for 10% of what i quoted.

you get what you pay for in this world. people seem to have forgotten that.

[glc]

This happens even at the lowest level. I had a prospective customer - a lawyer - who wanted to network 4 Win 9x boxes in his office. He expected me to do the job - hardware, cabling, AND labor, including setting up app sharing - for $500. Yeah right. Try at least a grand to do it right. For $500 he *might* get coax strung across the floor with some old used ISA NE2000 nics.

[Paul Victorey]

Ooh, it gets better. They're connecting all 75 or so clients together, using ony a stack of hubs. As in ALL clients are in one collision domain, as they plan to originally set them up.

I think that my first modification will be to break the stacks of hubs into smaller stacks and add a switch so as to have a snowball's chance in hell to keep collisions at managable levels.

At least they really can't go wrong with the cabling job. Although the length is long on some runs, they have a ~390 foot run they're doing. Although really it should work just fine (in fact I would doubt that particular run will ever be used, I can't see those rooms getting computers ever, they are actually rented out to a preschool, so they shouldn't even use our network anyway.)

------------------
Paul M. Victorey
------------------
I am not responsible for any problems that may arise as a result of following my advice. This includes, but is not limited to, computer failure, loss of data, nuclear war, famine, boils, no clean laundry, your daughter running off with a biker gang, or armageddon. Take my advice at your own risk.

[vitalstatistix]

good save the network! ethernet with a stack of &gt;3 hubs(??) on one collision domain and that too with such extended runs ...
i say u will have either collisions or dead/dropped packets due to timeouts!

gtfx
PS: betch ya, u were LOL and ROLF when u read about this design first!

[Paul Victorey]

Oh, you bet I was http://www.pcmech.com/ubb/biggrin.gif

I was just speechless when I asked "what type of switches are you using?" and I got "oh, we're not using switches".

Now, if he wants to run his network like that, fine, but _I_ am not competing with 74 other computers for control of the ethernet bus http://www.pcmech.com/ubb/biggrin.gif


------------------
Paul M. Victorey
------------------
I am not responsible for any problems that may arise as a result of following my advice. This includes, but is not limited to, computer failure, loss of data, nuclear war, famine, boils, no clean laundry, your daughter running off with a biker gang, or armageddon. Take my advice at your own risk.

[bob]

It will still work --- for now.

1 Hub runs all the other hubs and it connects the servers.

Main problem with collision domain is that if 1 problem - every computer sufferes.

Would be better off with a small cheap switch using large hubs at about the same cost