I have different locations (customers) where I have to suggest the best way to connect devices.
The one location I am talking about has several catalyst 2900 switches all configured for 100 MBit full duplex with the workstations.
All switches are connected through fiber (1000 MBit) with each other. Then there are 3 servers (but 1 main server) which are connected also to a catalyst 2900 at a speed of 100 MBit full duplex.
I am afraid that if more workstations try to access the server simultanious the switch/server connection may fail due to performance problems.
I have this sort of problem in more than 1 location so what is suggested to be the best way to configure the switches when the server/swich connection cannot be upgraded. Should I choose the fasted acccess method? 100 MBit full duplex for all workstation and maybe creating a mayor bottleneck. Or choose a bit slower method and configuring the workstation at 10 MBit for a better "balancing" or "steadier throughput" on the netwerk with less "peaks" in bandwidth.
What should I choose and why should I choose this solution??
If performance is the key, then I would suggest rethinking the connection for the servers to the switch. I know you said its not an option, but turning 100mb/sec ports down to 10mb/sec would really be a backwards step.I suggest a gigabit connection to the server in question, and setting the client ports up to use 100/full.
When we keep everything on 100 MBit, wouldn't the switch connected to the server get overloaded with pakkets from the other switches through the 1000MBit backbone?
I was thinking when 10 switches with each 24 ports on which all 100 MBit workstations can generated a lot of traffic. The backbone can handel the traffic since it is 1000 MBit. The "server-switch" is also connected to the backbone and can get up to 1000 MBits of traffic destinated for the server. Since the port where the server is connected to can only handle up to 100 MBit shouldn't there be packets that get discarded somewhere on the way? And should those packets be retransmitted again and again??????
Is there something I can check to confirm this?
PS. Some of my locations (customers) can bundle some cards for load-balancing to get a 200 MBit connection. But this is not an option for all of the location.
I wouldn't use the network to throttle back your users for the server. As said before, that would be backwards. It would also be pretty darned time consuming also.
Unless you have alot of broadcasting or multicasting going on in your network, I really wouldn't sweat the fact that you aren't using Layer 3 yet. Depending on what type of Layer 3 device you select, and where you put it in the network, you could turn it into a bottleneck instead of your server. That's a more expensive way to end up right where you're at now. If everyone is pretty much always talking with the server, and if there is not much broadcast/multicasting going on, I don't think Layer 3 would buy you much anyway.
The one idea about using Fast Etherchannel is a good possibility. Uplinking the busy server(s) to the switch with a Gigabit Ethernet card would be a good way to go, if you can get an open GigE port for it.
To see if you have a performance problem to begin with, look at the SH INT statictics on your switches. Here's some stats you can review:
FastEthernet0/2 is up, line protocol is up
Hardware is Fast Ethernet, address is 0007.501d.cf82 (bia 0007.501d.cf82)
MTU 1500 bytes, BW 100000 Kbit, DLY 100 usec,
reliability 255/255, txload 1/255, rxload 1/255
Encapsulation ARPA, loopback not set
Keepalive not set
Auto-duplex (Full), Auto Speed (100), 100BaseTX/FX
ARP type: ARPA, ARP Timeout 04:00:00
Last input never, output 00:00:01, output hang never
0 output buffer failures, 0 output buffers swapped out
If you haven't cleared the counters, you may want to do that a couple days before you do this, so you have a known period of time to analyze. Keep in mind that the analysis would be for that period. It's best to do this periodically from time to time to get a feel for the hotspots and issues across your network...
-Look at your input and output queues (12th line) to see if you are dropping alot of packets. Also look at the deferred, throttles and overruns for clues of link saturation.
-Look at the ratio of input and output packets versus the broadcasts and multicasts to see if you have any problems there. This will tell you whether to worry about adding Layer 3 at some point.
-Errors such as CRC/FCS, late collisions, frame errors indicate either duplex mismatch, a media problem (bad cable), or an installation issue such as a location's cable being too long.
Start with that. If you want to tell us what types of switches and uplinks you have, that may be helpful.
We are pleased to announce availability of Beta software for 16.6.3. 16.6.3 will be the second rebuild on the 16.6 release train targeted towards Catalyst 9500/9400/9300/3850/3650 switching platforms. We are looking for early feedback from custome...