Solved

Maximum nodes for Ether and FDDI

Posted on 1998-08-03
14
891 Views
Last Modified: 2008-02-26
I just read a book, which says the maximum nodes for a ETHER network is 1024, and for FDDI is 500, for Token Ring is 260.  Is there anyone can tell me why they are 1024, 500, and 260. Are the maximum "hard" limitations? If there are more than one VLANs in a network, are these limitations for any VLAN or for the whole network? Is there any similar limitation for FAST ETHER network?
0
Comment
Question by:lmnetware
  • 6
  • 4
  • 4
14 Comments
 
LVL 2

Accepted Solution

by:
calger earned 100 total points
ID: 1592967
For some information on Ethernet networks:
http://www.lantronix.com/htmfiles/mrktg/catalog/et.htm
http://www.betanet.co.uk/ETutorial.htm

The real restrictions reside in the way the network works.  For example, in Ethernet you have a "collision-domain".  Every time a node wishes to transmit on the wire - it just sends out its packet, risking a collision with another packet.  If this occurs, the node resends the packet a few ms later.  The more nodes you have on a network, or the "chattier" these nodes are, the more likely you are to have collisions.  So, if you had a network of 50 nodes, but they all talk all the time, you would have many collisions and might need to segment your network.  If, however, you had the same 50 nodes with very little traffic - you would be fine - no need to segment.  You could take this model up to x number of nodes and it would generally hold true.  If, however, you do happen to install 1000+ nodes on a SINGLE network - just the SAPs and "hello" packets, etc. of the various protocols would generate enough traffic to cause network bottle-neck problems.  I would say that there is little to no use for ANY network to have 1000+ nodes all on the SAME segment.  Bridge/route that puppy!

Token and FDDI work with a token passing system - the node doesn't transmit till it receives the "token" and is allowed it's chance to "talk".  The problem with large networks in this type of scenario is that the node then has to wait longer and longer for the token to be passed around the ring.  This can cause great latency problems.  Again, segment, segment, segment!!!

Regarding your question of Fast Ethernet - the more bandwidth you have available does allow for more transmission - but the generally idea of how often each node transmits and the likely-hood of collisions doesn't change.  Its still just one circuit and there can only be so many nodes trying to talk at the same time.

In conclusion, you should never allow your network to grow to the point where collisions/latency becomes a crippling problem.  You can always throw in bridge or a router to segment your network to break up the size of the collision domains or ring.  So, to answer your question, No, there is no HARD limit where the network automatically fails when you reach 1025 nodes on an ethernet network - but you will experience collisions and bottlenecks that will force your network to be unusable - and you will generally reach that point long before you reach 1000 nodes.

Good luck!

0
 
LVL 3

Expert Comment

by:paulnic
ID: 1592968
The numbers you read are for networks WITHOUT switches, bridges or routers.  If you have a VLAN, you have a switch, which dramatically changes the assumptions on which each number is based.

The token ring number is a hard limit for NICs in single ring, without any switching, bridging or routing.  With Ethernet or Fast Ethernet, the limit would only be for the number of 10BaseT or 100BaseT stations on a single switch port, interconnected by hubs.

Bottom line:  In a good network design you'll never come anywhere close to these per-segment limits.  We have about 1500 10BaseT NICs on a single switched LAN, but no more than 20-30 per segment.

Peace^^Paul
0
 

Author Comment

by:lmnetware
ID: 1592969

Hi, calger and paulnic,

Thanks for your anwsers to my question. There are little contravention between your anwsers about the if the limit
is "hard" or not. But it does not matter. Now I am interested
in Paul's 1500 NIC network. I also has a LAN with 1200
nodes and with much have traffic. The problem is when the traffic is too heavy, some switch may go down. The switchs often go down are BAY NETWORK 28115. Sometimes the SQL Server runing on the network may also goes down during heavy traffic. We also tried Sybase instead of SQL SERVER, but the situation is about the same. I do not know the traffic in you network, would you give me some suggestion basing on your network.

Thanks

lmnetware  
0
 
LVL 2

Expert Comment

by:calger
ID: 1592970
Do you have any bridges or routers in your 1200 node network?  

We currently are running a network with over 500 frame points, several FDDI rings connecting major campus locations, and our 4 major centers are connected by T3 lines.  We have somewhere in the neighborhood of 18,000 nodes and innumerable different services on the LANs/WAN.  We experience some network slowdown problems on our slow WAN links (64K) due to the flood of SAP traffic for IPX services.  We are currently handling this situation by filtering some SAPs on the routers as well as trying to implement Netware/IP in most situations.  This is alleviating the problem.  However, in your case, your links - if they are all local - should be able to handle SAP traffic from 1200 nodes.  If you aren't bridging or routing your network, even using switches isn't enough to segment off the traffic from 1200 nodes.
0
 

Author Comment

by:lmnetware
ID: 1592971
Hi, calger,

It is greatful to share your experience. Yes, I have a local network with 1200 nodes without any dedicated router in the network. I have five servers, 4 NETWARE 4.11 and 1 WINDOWS NT. I used the four NETWARE servers as routers. Actually, the traffic through the routers is light, Their main task is to block the SAPs. I used 5 BAYNETWOKS 28115, and devide the whole network into five VLANs. I know the traffic on the network is heavy, and my application need the connectivity of the whole network. Even in the peak traffic case, every user can find the network is very fast by runing any program. But the problems is sometimes the switch goes down, or the SQL server goes down. At the begining, we used SYBASE on NETWARE instead of WINDOWS NT and SQL SERVER, we found the SYBASE server is not stable.
0
 
LVL 3

Expert Comment

by:paulnic
ID: 1592972
HI, lmnetware...

I am happy to see somone else who understands that 4 or 5 servers is enough for 1200 nodes.  And you're the first actual experience I've heard with Sybase on Netware.  We have a small Oracle on Netware application that seems stable, but our big Oracle apps are on Sun boxes

I'm concerned about the problems you report with the 28115 switches.  Partly because I respect Bay engineering, and partly because I have believed that a 28115 should be very hard to disrupt with extra traffic, since it was supposed to do fine with all ports running at full wire speed.  You might need to describe  how you have things connected, and how many NICs per server, to give an idea of traffic flows.  Have you spoken with a Bay technical person about your problems?  Do you have traffic statistics that point to traffic as the likely cause?  Our switches, for what it's worth, are Cabletron, but I had been hoping for Bay.

IAs far as SQL Server going down, depending on the messages, I wonder if someone in the MS-SQL forum on Experts Exchange could help you.

Calger...I'd heard that a switch acts like a fast MAC-layer bridge, and because it basically delimits a collision domain, you don't need bridges if you switch.  You may decide to use routers (or smarter switches) between biggish switched VLANS, but that's because many switches can't filter broadcast traffic...and sometimes you don't want them to!

Peace^^Paul
0
 

Author Comment

by:lmnetware
ID: 1592973

Hi, Paulnic

It is a pleasure to hear from you again. Sybase in netware is very common in China, but all the system I heard and saw here are all small. The number of clients can not comparable to my case. So most of their system are stable.

I also know somthing about Cabltron. But the price in china is much higher than BAYNETWORKS. As my experience, 28115 is not as good as it has been said. I almost consult all BAYNETWORKs experts in china about my case, but no one can give me a nice anwser. I gave them the topology of my network, and told them the problem.

I used two NICs for every NETWARE servers, and one NIC for Windows NT server. 5 VLANs are create through the 5 28115. If we refer to one VLAN as the center VLAN, all 4 NETWARE servers connect to the center VLAN, and also to the left 4 VLANs respectively. Every Netware server just connects two VLANs. It is difficult to describe without a figure, but I hope you know what I say.

Lmnetware  
0
IT, Stop Being Called Into Every Meeting

Highfive is so simple that setting up every meeting room takes just minutes and every employee will be able to start or join a call from any room with ease. Never be called into a meeting just to get it started again. This is how video conferencing should work!

 
LVL 2

Expert Comment

by:calger
ID: 1592974
Paul - if you can configure your switches for filtering then you're absolutely right - they basically are low-level bridges.  They can't connect different media types like a true bridge, but that's not the issue here.  The catch to this manner of thinking, though, is that if a switch CAN be set for filtering doesn't mean it HAS been set for filtering.  To properly determine if traffic is the problem, however, you probably are going to have to hook up a sniffer or you a similar software-based system.

LMNetware - Are most of the users on the same switched segment as the primary server that they are trying to access?
0
 

Author Comment

by:lmnetware
ID: 1592975

Hi, calger,


  Most users in the same segment will access to the same server (the nearest server, not only the primary server).  
0
 
LVL 3

Expert Comment

by:paulnic
ID: 1592976
lmnetware,

Does your setup look something like:   (SW is switch, Sn are servers)

\                                     /
 SW - S1 - SW - S2 - SW
/                   |                  \
                   S3
                    |
                  SW
                 /     \
               PC    PC

I assume your server NICs are 100Mbit, yes?  Which brand?  If they are set for full-duplex, you may want to try half-duplex.  Although this may seem less efficient, it can sometimes be more reliable.

Another option (not too hard to try) would be running TWO NICs from the server into each switch that's NOT in the center...load balancing.  This requires NLSP and attention to Novell tech docs 2926169 and 2909238 from http://support.novell.com/search/kb_index.htm

CALGER.... not to argue, only to clarify, what would a switch need to filter out in order to act as a bridge?  Do you mean packet fragments like collision debris?  Broadcast packets...which I thought cross bridges routinely?  Something else?

As far as connecting different media types, some of our switches have ports for both fiber and UTP copper.  These almost seem to be the only media that people care much about these days unless they do cable TV, and fiber is pushing out coax pretty far there too.

[ I'm not counting wireless yet, though Diamond Multimedia is supposedly about to LAN your two (or more) home PCs for $100 each, at 1Mb/sec and 2.4GHz, with "ether" as the only medium.   ;-)  ]

Peace^^Paul


0
 

Author Comment

by:lmnetware
ID: 1592977

Hi, Paulnic,

Yes I use 3COM 3c59x 10-100M NICs, I have tried full-duplex and half-duplex both, and also changed the port mode of the 28115 according to the mode the NICs works. As to the second way you mentioned, I have no so much ports in 28115, because I must keep some ports for connecting swiths of lower layer.

My topology is like the one you drew, but not the exactly same. Actually, some VLANs include ports in different 28115s. The reason in we must balancing the traffic.

Thanks anyway.

lmnetware

0
 
LVL 3

Expert Comment

by:paulnic
ID: 1592978
Hi, again, lmnetware...

If you have an opportunity to try an Intel Pro 100B NIC in the server, you may want to.  I have used 3com for a long time, and have many running now, but I have heard that the intel cards may do better in servers under some conditions.  I have also seen 3com-related manufacturers like Dell shift toward Intel for server NICs.

How many layers of switch ports do have between workstation and the farthest server?  

I assume everyone has checked the design for "loops" or multiple parallel paths between devices which may cause difficulties?

Peace^^Paul
0
 
LVL 2

Expert Comment

by:calger
ID: 1592979
Paul -

Bridges by default DO filter segment broadcast traffic.  Also, if you are using a switch that connects two different types of media, then it has a bridging-module built in or added to it.  Hence, it really has been converted to a bridge.  Again, not to belabor the point!  :)

lmnetware -

Are you breaking the rule of 5-4-3?  Only 5 segments can be connected via a maximum of 4 repeaters/switches and only 3 of these segments can be populated (all this is true unless you insert a router)?
0
 

Author Comment

by:lmnetware
ID: 1592980
Calger-

I know the rule of 5-4-3. In any VLAN in my network, the maximum number of switchs is 3. The VLANs are connected just via Netware multiple protocol routers. I only use switches not repeaters or bridges in my network.

Paulnic-

I have tried Intel Pro 10/100 NIC, but the situation has not been improved. Actually I also tried the NIC going with COMPAQ server which I am using. No improvement.
0

Featured Post

How your wiki can always stay up-to-date

Quip doubles as a “living” wiki and a project management tool that evolves with your organization. As you finish projects in Quip, the work remains, easily accessible to all team members, new and old.
- Increase transparency
- Onboard new hires faster
- Access from mobile/offline

Join & Write a Comment

Suggested Solutions

Restoring deleted objects in Active Directory has been a standard feature in Active Directory for many years, yet some admins may not know what is available.
A Short Story about the Best File Recovery Software – Acronis True Image 2017
It is a freely distributed piece of software for such tasks as photo retouching, image composition and image authoring. It works on many operating systems, in many languages.
Illustrator's Shape Builder tool will let you combine shapes visually and interactively. This video shows the Mac version, but the tool works the same way in Windows. To follow along with this video, you can draw your own shapes or download the file…

708 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question

Need Help in Real-Time?

Connect with top rated Experts

13 Experts available now in Live!

Get 1:1 Help Now