Networking

3 Simple Techniques To Boost Your Ethernet Performance

Our network expert shares simple ways to spot potential problems before they become end-user nightmares.

Joel is an internationally recognized expert in the areas of security, networking and messaging. Follow him on X (formerly Twitter) @joelsnyder.

Ethernet is a great network technology for linking users together in buildings. Easy to design, install and manage, it works flawlessly — until it doesn’t. Then it can be a great mystery and a business-interrupting nightmare.

Here are three tips for managing Ethernet deployments that will help avoid problems and reduce network downtime for your users.

1. Make Ethernet Switches Manageable — and Then Manage Them

Every Ethernet switch in your building should be a managed switch. Be careful though: One feature you need for monitoring is Simple Network Management Protocol, which now is often stripped out of the lowest level of web-only managed switches.

When SNMP is everywhere, SNMP management tools can track statistics on every port. That’s an important first step because, without an automated view of your ports, you can’t be proactive about problems.
Each management tool is different, meaning terms and capabilities vary. You need to be able to monitor in two areas: port errors and port speed/status (including Power over Ethernet).

Collect this information on each port every few minutes. That’s every port in the network, whether dozens or thousands. For early detection, this could be every 15 minutes or every hour. Watch your reports, then adjust monitoring intervals appropriately.

2. Gigabit Ethernet Should Be Gigabit (or Faster)

Monitor the speed (10/100/1000) and duplex (half or full) of each port. Nearly all should be 1000/full. Some equipment, such as printers and Voice over IP phones, may connect at 100 megabits per second or go into power-saving mode: Configure your management system to ignore these exceptions.

Any new device connecting half-duplex is a performance problem waiting to happen. (Possibly someone dragged in a 20-year-old hub from home. You know it happens.)

Track down these occurrences aggressively, again using a daily or weekly report. The same is true for devices connecting at 10Mbps or 100Mbps. They may seem to work fine but they are signs of old devices on your network — and they’re going to cause trouble.

3. Network Errors Don’t Just Happen — Find Out the Root Cause

In modern Ethernet networks, ports don’t get errors. For example, the port I’m plugged into now has sent and received more than 20 terabytes in 18 months without a single error.

There really is no acceptable level of errors in Ethernet, and any errors indicate either a hardware problem in the switch or device, or a misconfiguration (such as locking in speed and duplex, something that is strongly discouraged in gigabit networks).

Have your management system send a daily (or even weekly) email listing ports that have any errors on them. If there are any, they need to be investigated and resolved.

Here Are Quick Tips on How to Design Ethernet LANs for Maximum Uptime

Ethernet local area networks are incredibly reliable, but every system can break.

With a few design tweaks, you can increase uptime and reduce the likelihood of failure. The key to improving network design is to identify a few of the most failure-prone components and reduce their impact on your network.

The No. 1 least reliable network component? Cables and patches — always. (The second one is power, so add power protection to your list as well.) It doesn’t matter how “good” the quality of the cable you buy, sooner or later, some are going to fail without anyone touching them. The cable path between switch and user should only affect a single user, so it’s normally not worth spending a lot of money trying to do a super-reliable link.

Avoid problems in wiring closets by minimizing environmental factors, such as dust and hot or cold air blowing directly on cables.

If your wires are actually fiber, be very careful with cable management and routing. A sharp bend in a fiber dramatically increases the fragility of the cable and makes it much easier to break accidentally.

For the path between switches, a failure can be a big deal — taking out dozens or hundreds of users. Extra work here is definitely worth the effort. Start by adding redundant links to each stack of end-user switches. As you go down this path, make sure you read up on Multiple Spanning Tree Protocol (802.1s), which you will use to avoid network loops and optimize link usage.

At the other end of the wire, also plan for redundancy. Paired or clustered switches in the core layer will insure that any core device or patch cable can fail without affecting the network. Remember that switches fail occasionally, but patch cables fail a lot. Keep this in mind as you build redundancy at the core.

Stacking two switches with a single wire between them doesn’t go far enough: Cabling redundancy is even more important than switch redundancy.

Finally, make sure you test thoroughly after deployment to be sure that the redundancy actually works.

Yong Hian Lim/Thinkstock