05-27-2008 07:09 AM - edited 03-05-2019 11:15 PM
Hi,
I am facing some issue below :
LAN : Ping from any host below is giving max latency upto 200ms,
where it is expected to be <1ms
as per our LAN standards
any pc > sw01 > core_sw
sw01 > core_sw
core_sw > sw01
WAN :
ping latency is equal to 1 ms latency with remote neighbor router
Our Network :<=WAN=>router > core_sw > sw01 > pc
2nos 3845 routers with point to point leased line connectivity
2nos. 3750 stacked L3 core switches
20nos. 3560 L2 switches connected with redundandant links to both core switches.
Network infrastructure layout implemented by the best team as per international standards
Observation :
STP has selected one of the 20 switches and not core switch
>Request help to reduce network latency
Do making Core switch to be the root bridge for all vlans sort out the issue!
regards
Kaushik Vadali
05-27-2008 07:22 AM
1. Make your Core switch STP root for all vlans
2. Manually prune vlans off unneeded trunk links, decreases STP instances
3. do "show proc cpu sorted", what process is mostly used?
Questions?
1. Are you using jumbo frames anywhere?
2. Are you doing Policy Based Routing?
05-27-2008 08:51 AM
Hi,
I worry about effect of forcing core switch to root switch in a live network.
1. how much time it may take for STP convergence
2. we use only 2 vlans (data and voice) and never thought of pruning
3. show proc cpu sorted output is attached
4. no jumbo frames are used and detected on any interface
5. no PBR in LAN
regards
Kaushik
05-27-2008 09:13 AM
1. What is serving your default gateway for the clients? switch or 3800?
2. 20 switches connecting into stacked core?
Pings between switch and router are what?
Pings between client and switch are what?
Pings between client and router are what?
05-28-2008 01:42 AM
1. Default gateway for the clients is core switch.
2. All 20 switches are connected to stacked core switches
pings stats
core switch to router, client to switch, client to router
90% <3 ms, 5% 4-10ms,5% 10ms to 100+ms
The slowness problem is present for long time and our audit has told us to rectify
regards
Kaushik Vadali
05-28-2008 04:46 AM
Your connection between your switches and core should be <1ms
I would try to change the STP root for all vlans.
How many hosts on this data vlan?
Also, check your MTU to be 1500.
05-28-2008 05:32 AM
Dear Ralph,
Your connection between your switches and core should be <1ms
---Its not! for 10% pings
I would try to change the STP root for all vlans.
--- How much time it may take for STP convergance. As down time of morethan 1 minute is not allowed without scheduling
How many hosts on this data vlan?
--- 1500 hosts in 8 hrs, 24x7 operations
Also, check your MTU to be 1500.
--- MTU is default 1500 on all devices
regards
Kaushik Vadali
05-28-2008 10:09 AM
1500 clients on 1 vlan?
Theres your problem!
Breakdown that broadcast domain to use /24's and not anything bigger.
Since you connect the access switches to the stacked core, you really dont have any STP blocked ports and cannot have a loop.
Changing the STP root is best practice in this case but should not have an impact on this problem. I think you may lose connection for 1-3 secs only.
But I think you have a bigger problem and that may be your vlan size.
05-29-2008 03:38 AM
Dear Ralph,
I agree that we are using more hosts per vlan but cannot use /24 mask as per organization policy. I can use max /22 mask and limit to 1024 hosts
All hosts are DHCP enabled and work on single process, client server based CRM application
I plan for scheduling of core switch to be the root bridge and update you on the outcome soon
One Question - Do intermitant high latency (15 to 150 ms)is the cause for network slowness? Some points to note on Client systems are
- They are not working on internet
- They run and access only single CRM application located on the remote side of WAN
- Accessing common file server located on LAN
regards
Kaushik Vadali
05-30-2008 11:04 AM
So this application sits on another subnet on the remote side of the wan? if so you may have to look into tweaking it to serve multiple subnets and not just this one subnet your using as your data vlan. or is it spanned onto the same vlan?
For ex. Keep in mind that when a pc sends out an arp request for this app server, the switch floods this arp request to the whole vlan. Every host (1500 of them) will hear it and have to process it.
Try this. Create an new test vlan and SVI. Put a test pc in this vlan and point the default gateway to this svi. Do some tests.
Can you send your 3750 and 3560 config without actual IPs?
06-01-2008 01:40 AM
Dear Ralph,
I will try and update soon
regards
Kaushik Vadali
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide