Connections +
Feature

Data Detectives for Ethernet Networks

The need for reliable network performance is greater than ever and increasing by a factor of 10 every five years.


July 1, 2007  


Print this page

This article covers several areas in testing Ethernet Networks. Foremost is understanding common network problems and how to discover them, second it deals with measuring the quality of experience (QoE) of someone actually using a network so that you can duplicate and quantify issues, but also break them down so the likely cause of poor resource performance is located and addressed.

It will be of use to individuals or teams managing Ethernet networks regardless of size and scope that are responsible for resolving issues and responding to calls.

As you read it, the chances are you have got a computer nearby and that PC or Mac is connected to either a home or a business network.

The protocol of that network is likely based on IEEE 802.3 or Ethernet, a shared local networking technology that was designed in the early 1970s by the same engineers who were working on the development of the Internet.

Ironically, Ethernet and Internet Protocol (IP) is becoming the dominant networking technology in the world regardless of distance and it continues to get faster by adding new standards to keep pace with today’s bandwidth demands.

Since Ethernet is being used outside its initial local area network roots, its speed just keeps increasing as data is aggregated together.

Think about how fast and available it has become. Any new laptop network card supports up to 1 Gigabit Ethernet. Optical 10 Gigabit Ethernet has been deployed for the last three years from access networks through service providers core networks and 100 GigE is being worked on by an IEEE committee already with a standard projected in 2010 and products in 2011.

In addition, 100 Gbps bit error rate testers are already shipping to help test designs.

Fast broadband connections

Historically, because of increased bandwidth needs at the networks edge, metro and core networks require the higher speeds first and have made a jump to the 10 times increase in speed when the cost is three times more than the slower rate.

While this is all interesting, the point is that network issues only get bigger over time as we become more dependant on networks.

Can we become more dependent than we already are? E-mail is the standard for business communication and working with people across the country or across the world is not uncommon.

Because of this, many applications are now Web based which by definition relies on network connectivity to function.

So what is driving the speed increase? Fast broadband connections are now affordable and still getting cheaper. We can see the effect it is having on Ethernet. I mean that literally. The killer application that is now possible over Ethernet and driving network growth is video.

Enabled by new applications such as Video chat (Skype, MSN Messenger ) and Movie download sites (Google Video, YouTube, ITunes) , bigger pipes to consumers and business as well as servers connections to transfer content will continue to grow.

The volume, in turn, drives network element prices down and the cycle repeats. Where we were ripping CDs to our hard disks last year, this year it is about storing videos (or trying to) to PCs with huge storage capacity since it has become affordable to have a one TB hard drive at home.

But growth is never without problems. New applications such as IP telephony and video require different network performance than required by e-mail or Web browsing. Network delays and data integrity become more important.

VoIP communication is acceptable only when a network loses less than 0.5% of the data transferred and video is 20 times as sensitive.

Another way to say this is not only does the pipe have to be big enough, but it needs to be free of leaks and have the right consistent pressure.

To summarize so far, the need for reliable network performance is greater than ever and increasing by a factor of 10 every five years. You can’t fix what you can’t see or test.

So, can you test your network, especially in a way that simulates it being used?

Are you getting what you are paying for out of your network elements or service provider?

Did you ever dream of a day when employees could go to a store and casually buy devices that can connect to a network?

OK, it is not a dream, more like a nightmare, and maintaining network inventory and security is harder because of cheap, easy to configure devices like wireless access points. Can you see everything connected to your network?

Can you test to see if you have a network that can VoIP? Does it have low enough latency (delay) to be a viable alternative to your standard telephone system?

Testing provides answers

Here are some problems and tools available to solve Quality of Experience (QoE) issues. The majority of daily network issues involve QoE problems not network outages. Here is a list of some challenges for many IT folks:

People growing up with the telephone as their main communications device are not familiar with the ISO layer 1-7 protocol stack, so most will not troubleshoot problems for you.

The common complaint is the network is ‘not working.’ A more useful complaint (one layer higher) is that the network is slow. Almost always the complaint is unclear. The cause could be a defective network element like a faulty NIC card broadcasting Ethernet frames into the network, or a more subtle problem that creeps up on you because your growing network overloading elements

It is mandatory to know what devices are on the network. Even just doing a compare or delta to a previous time can be very useful later on.

Some sample problems that can cause issues:

Duplicate IP addresses on stations cause inefficient networks. Both stations are affected since resources become confused about which node is the ‘real’ address. Symptoms range from applications failing to poor throughput since there is unnecessary traffic being broadcast

Bad Cabling: Whether copper or optical it is the number one failure. If you cannot connect, check the cable. A simple Wiremap can be useful to measure continuity or discover a crossover cable is being used.

Bad NIC Cards: Networking cards fail and can affect not just a PCs, but the entire network. Everyone has a story like ‘a PC’s NIC card’ flooded the network with broadcast frames. Usually contained in larger networks using different subnet and VLAN assignments, for most of us, it debilitates a network. Finding the ‘Top talkers’ as they are called is invaluable to locating the MAC address and shutting it down.

Port Duplexing Issues: Many devices differ in their speeds and duplex characteristics. Duplex or half-duplex? 10 or 100Mbps? How about Auto-Detect? Some devices require us to match the speed and duplex settings on the switch with the device’s interface. Changes or mistakes happen resulting in switch port and device settings that no longer match. More and more mismatches mean collision and retransmission rates increase and eventually someone notices.

Overloaded routers and switches: Examples are routers with insufficient memory or address table limitations. Perhaps these have not been upgraded as your network has grown. Frame loss, delay and useable bandwidth can all be affected. Unauthorized devices like wireless access points are mostly a security threat, but also a source of unpredictable network performance.

Handheld test equipment has the advantage of being easier to use with specific user interface, portable, rugged and have dedicated network hardware that in operates at full line rate. They analyze 100% of network traffic, not just a sample, up to 1 Gigabit full duplex speeds

Sometimes performance is not the issue. It is more basic than that. Checking for connectivity is what is required. Testers plugged into a port can automatically tell if the network is there, its speed and if the switch duplex setting is correct

Some scopes can automatically discover network elements
and provide details on what servers are available, be it visible elements like Web Servers, file and e-mail servers or more invisible resources like DNS, DHCP, or Routers.

The newest tools can also provide details on what Virtual LAN (VLANs) exist. Auto discovery tools can be passive or politely query the network discovering what’s out there and can store a station list for troubleshooting or comparison later.

Handheld tools exist to graphically show you network utilization (stress) over time and then drill down to show the address of the ‘top talkers’ on the network. You may want to connect test equipment to a switch configured as a span port, which mirrors traffic of the other ports.

We have covered some tools to solve common network problems, but how do you troubleshoot an issue when a user simply cannot get something done? You measure their Quality of Experience or QoE. QoE testing does not just mean determining if a resource is turned on. I know my kids would love it if the only test in school they ever get would be an ‘attendance check’.

QoE testing is testing the way a resource is used, so you can isolate the real issue. Most times it is not the network at all, but the server or an element in between. It is a bonus if you can measure repeatably (like when your boss or customer is looking) and also not have to be an expert on all the resources in your network.

Armed with the proper tools to scope out a network, you can solve common problems and ensure users experience a network that enables productivity.

Peter Schweiger is a business development engineer at Agilent Technologies he can be contacted at peter_schweiger@agilent.com.