Case study: How a Cisco router upgrade affects network performance

Sooner or later, we all have to deal with upgrading some software or hardware in our organization. And while step-by-step lists and vendor documentation have their advantages, a case study can sometimes be your best guide. David Davis shares this case study of his experiences upgrading his organization's core router, and he shows you how this upgrade vastly improved performance.

Upgrading equipment is a fact of life—sooner or later, we all have to deal with upgrading some software or hardware in our organization. There are countless resources available that explain the many finer points of upgrading routers; in fact, I wrote an article earlier this year about upgrading routers and switches.

However, as many IT pros will attest, experience really is the best teacher. And while step-by-step lists and vendor documentation have their advantages, a case study can sometimes be your best guide because it walks you through a real-life scenario.

This week, I'm sharing my own case study, taken directly from my experiences helping upgrade my organization's router, and I invite you to share your own upgrade stories in this article's discussion. Let's look at the difference a router upgrade can make on network performance.

Over the past few months—and perhaps even longer than that—network performance between my organization's central data center and our 70 remote locations had been suffering to the point that users began complaining. This case study begins about five years ago, when my organization set up a frame-relay network serving all locations in a hub-and-spoke topology.

Each location had a 56-K circuit, and the central host had a full T1. Initially, this worked fine for the handful of text-based dumb terminals, text-based printers, two Citrix thin client devices, and one Hewlett-Packard laser printer that we had running at the time.

Over time, however, the number of thin client devices grew. Eventually, many sites had as many as 10 thin client devices. Now, while it's true that Citrix thin client devices are "thin," they're not thin enough to support any more than three to four semi-active clients on a 56-K circuit. So, we decided to upgrade our remote 56-K circuits to 512-K MPLS circuits and upgrade our central router to a fractional DS3 9-Mb circuit.

All this time, we have had the same Cisco 3661 router handling the old network as well as the new network during the conversion phase. This router had 256 MB of RAM and R527x CPU, running at 225 MHz. However, before we converted even half of the 56-K circuits to the new 512-K MPLS circuits, we noticed that the CPU utilization on the central site router had increased significantly during the rollout of the new circuits.

We determined this using SolarWinds' Orion Network Performance Monitor, an application for midsize to large enterprises. In the past, I've also used the SolarWinds Engineer's Edition Toolset to graph performance for smaller sites that function on a smaller scale. It graphs router performance statistics (such as CPU and RAM) using SNMP.

In addition to these tools, we also periodically use the show processes cpu history command on the central site router. Listing A shows what this command's output looked like at the time the user complaints began.

As you can see, the old router consistently maxed out its CPU processing throughout the day. From this information, we knew that we had to upgrade our router so it could handle this additional traffic. To replace our Cisco 3661 router, we chose a Cisco 3845 router.

This enabled us to use our existing DS3 module and virtually copy over our configuration file. However, we did have to replace some modules and interfaces to account for unsupported older modules. (For example, the WIC-1DSU-T1-V2 replaced the WIC-1DSU-T1.)

Here's a list of the benefits the Cisco 3845 router offered us:

  • We can expand RAM 1GB, as opposed to the Cisco 3661's max of 256 MB.
  • The 3845 boasts four built-in slots for WIC cards, which meant we didn't have to use up a network module slot with cards to support these.
  • The 3845 includes two built-in Gigabit Ethernet ports, but the 3661 has none.
  • The 3845 sports optional features, including new and interesting modules such as a 36-port Ethernet switch module (inside one of the slots of the router), a network analysis module, and an IDS module.
  • The 3845 is certainly faster. While I haven't been able to find the exact CPU type and speed anywhere, Cisco advertises that the 3800 Series routers "deliver multiple concurrent services at wire-speed T3/E3 rates."

During the upgrade, we were able to copy our existing configuration, bring it into a text editor, search and replace the interface numbers to match the new interface numbers, and paste (or TFTP) the configuration up to the new router. It was a very simple conversion.

After we upgraded the router, users noticed a marked increase in performance. Listing B shows the output from the show processes cpu history command on the new router.

You can see for yourself the drastic difference this new router made in CPU utilization. We hope that this new router will last us for another five years, as the old router did. We continue to use both manual commands and more advanced tools to monitor the performance of this new router. And next time, we hope to predict when this new router begins to reach its maximum throughput—before users start complaining about slowness.

Miss a column?

Check out the Cisco Routers and Switches Archive, and catch up on David Davis' most recent columns.

Want to learn more about router and switch management? Automatically sign up for our free Cisco Routers and Switches newsletter, delivered each Friday!

David Davis has worked in the IT industry for 12 years and holds several certifications, including CCIE, MCSE+I, CISSP, CCNA, CCDA, and CCNP. He currently manages a group of systems/network administrators for a privately owned retail company and performs networking/systems consulting on a part-time basis.