Cisco Switching Lab on GNS3

Finished conducting an internal training on Cisco CCNP Switching Exam 642-813.  Rather than getting physical Switches and Routers for this internal training, I decided to run the lab virtually on GNS3 in my ESX lab.

It took me one week (5 nights and two full days over weekend) to get the lab up and running on Windows 7 as a guest OS in ESX 4.1.  Reasons for spending such a long time is because of the IOS images problem with 3745, 3725 images.  You can't save 3745 and 3725 configuration.  Have also tried to use 1821 images for emulation of PC (because of the 64MB footprint), but hit into CPU utilisation problem.  I just can't bring down the high CPU utilisation.

After many nights and testing, I have finally found the sweet spot of the IOS images for my lab.  Running 3660 IOS image c3660-jk9o3s-mz.124-15.T13.bin.unpacked and 7200 IOS image c7200-adventerprisek9-mz.124-15.T13.bin.unpacked.

3660 is running NM-16ESW (16 port Switch Module) and 7200 running PA-FETX and PA-2FETX.  All the PCs and Servers are emulated by 3660.

As for the ESX 4.1 setup, each Windows 7 is allocated with 8GB memory and 4 vCPU each.

The purpose of this lab is to demonstrate Cisco Switching Design Concept on Core, Distribution and Access Layer.  Access and Distribution Layer is using 3660 with NM-16ESW and the Core is using 7200 with lots of PA-2FETX. 

A brief description of the lab as below,

• Core Switch to Core Switch running Layer 3 IP with EIGRP
• Core Switch to Distribution Switch running Layer 3  IP with EIGRP. (There are some limitation on running NM-16ESW as a Layer 3 Interface.  The 3660 default 2FE Interface cannot do Port-Channel.  Hence, using NM-16ESW to perform Port-Channel to Core Switch but assigned a VLAN IP to the Port-Channel.)
• Distribution Switch to Access Switch running Layer 2 with STP and VTP Domain.
• Enable HRSP for Distribution Switches.

Below is the digram capture from GNS3.

Design Concepts

Core Layer
All interfaces in Core Switch is running on Layer 3 IP.  This is to stop the complication of running Layer 2 STP below Layer 3 IP.  As Spanning Tree will takes longer to converge when a link failed as compare to Layer 3 (EIGRP or OSPF).  Hence, running EIGRP/OSPF is enough to build and able redundancy on this layer. 

If we add STP into this layer, failure of links will casue STP to converge first while putting Layer 3 routing protocols in an unstable stage. STP may takes 50 seconds (Classic STP) or 15 seconds (RSTP) to settle down.  Hence, the network may be down for 15 to 50 seconds (if the STP root is down, 50 seconds worst case). 

Using OSPF or EIGRP, there will be no network downtime should any router or link failed.  OSPF or EIGRP may takes 3 to 5 seconds to stabilise the network routing topology, it may drop some packets during this stage due to wrong routes but TCP packets will still be alive as long as we keep within the 20 seconds of TCP session timeout.

Routing Speed of the devices in this layer is citrical.


Distribution Layer
Switches in this Layers runs the most protocols. As this layer links between the Core Layer and Access Layer, it runs both Routing Protocols (EIGRP/OSPF) facing the Core, and runs STP as a Root or Secondanry Root facing the Access Layer. It will also be the VTP Domain Server for creating and managing of VLANs.  As these switches have IP addresses for VLAN interfaces, it will also be the best place to implement HRSP/VRRP for hosts. Security between VLAN can also be implemented in this layer.

Layer 3 Switching Speed of the devices in this layer is citrical.

Access Layer
Switches in this Layers runs VLANs and STP.  Port security and QoS marking is done here.  Port density and supporting of different devices (such as PoE for Acess Point and IP Phone) is citrical in this Layer.



The rest of the "normal" design concept can be found in Cisco Website.