When Good Networks Go Bad, Part II
| 12/19/2016 11:48:46 AM
Last month my blog addressed, “ Why Good Networks Go Bad”, from a Layer 1, hardware perspective. I discussed why it is easy to jump to a higher layer of troubleshooting---especially when managed end devices such as Ethernet switches are installed- but Layer 1 is the right place to start. Many times an Ethernet network problem can easily be resolved. However, once you have tested cables, checked LED status, and other hardware-related steps, it is time to move up to the next steps for troubleshooting.
While physical device failures can occur in a network, the highest percentage of network issues occur due to misconfiguration of the network devices. Speed, Mode and Duplex are the keys to the network running smoothly or not.
Ethernet Network Devices
Many LANs/Wans have Ethernet switches and media converters. Switches and media converters consistently have copper and fiber interfaces. Since copper cannot cover a great distance and fiber switches are very expensive, media converters are often a cost-effective solution. Switches and media converters are available in single port or multi-port, and with or without management. Let’s look at a typical network segment.
- You have the correct type of fiber connection based on Single Mode fiber or Multi Mode fiber, to extend the network over a distance. Fiber is not a one size fits all. Fiber types and wavelengths are important to match throughout a network segment. Single mode fiber products should be connected over single mode fiber runs, as multi mode fiber will allow too much modal dispersion, and support entirely different wavelengths and distance coverage.
- You have the correct kind of copper cabling to support cross-over or straight-through cabling (if necessary, as some legacy devices may require it). Many devices today offer Auto cross, a feature that will accommodate any type of copper cabling, and some older devices may still have an MDI/MDIX push button that must be set according to the copper cable type. Either feature is easier than trying to make sure you have a specific copper cable.
- You have the right devices to support what protocol must be supported (in this case, Ethernet). Just because a device has RJ-45 ports and fiber ports does not mean it can support the protocol necessary for your application.
For the sake of simplicity, I have chosen an Ethernet media converter for my example; it is a 10/100Mbps copper to 100Mbps single mode fiber, unmanaged media converter.
Example: the media converters have been installed as a pair, to connect to a core switch at the Central Office (CO) and then over 5 miles to a customer premises (CPE), which has a 24 port switch. Since the core switch and the customer switch offer multiple copper ports, the media converters become the inexpensive solution to connecting the copper ports at each end and then over single mode fiber (covering the distance of 5 miles), which has been leased from the city. Although the LED displays link on the copper and the fiber interfaces, the network is not functioning as expected. Packets are being dropped and the business owner at the customer site is unhappy. Since the core switch at the CO is providing connections to other locations successfully, the media converters become the suspected source of the problem.
Why Configuration is important
Configuration must be taken into consideration, for speed, mode and duplex. Some network folks think they can configure the media converters and then force the end points (CO, CPE) to conform. In other words, set the media converter to Auto Negotiation (the mode), so it will advertise the speed (10 or 100Mbps), and by default, Full Duplex (FDX). [Customers often don’t understand how to configure a core switch, or have availability to the CPE device, which can impact the configuration settings.] However, the media converters cannot be the driving force in the segment, the end devices must ultimately be responsible for the configuration.
Some network folks think they need to only configure the end devices ( in this case, the switches at the CO, CPE), and then the media converters will conform to the switch settings. While a media converter can be considered the “piece of wire” connecting the end points, it still has a part in the network configuration. The media converters will try to conform, but mixed signals will impact their performance.
In order to perform a complete assessment of the network segment, the following steps must be taken:
- Any diagnostic features on the media converter should be set to OFF, to not influence the performance of the network segment.
- record the configuration of the switch at the CO
- record the configuration of the media converters
- record the configuration of the switch at the CPE
All devices in a network segment must be configured in the same mode. The mode is either auto negotiation (AN) or Force. If some of the devices are set to AN and others to Force, dropped packets, linking issues, or other errors are inevitable.
All devices in the network segment must be capable of a common speed. If one device is capable of 10/100Mbps, and another is capable of 100Mbps, the devices will agree to the common denominator, in this case, 100Mbps. An Ethernet device is capable of the speed(s) it is designed to support. You cannot expect a 10/100Mbps device to support gigabit speed. This rule is applicable to copper or fiber.
All devices in the network must support the same duplex. While this was a concern in years past, when 10Mbps devices were common, and Half (HDX) and Full Duplex (FDX) were choices, today it is of little concern, as networks can and do perform under FDX. It would be a highly unusual segment that would require HDX, and the end devices would have to be set accordingly, as media converters are open to HDX and FDX; they conform to what the end devices agree to.
To summarize, all devices in an Ethernet network segment must conform to mode, speed and duplex. And to be clear, the switch at the CO and the connected media converter cannot be set to one type of configuration, while the other media converter and the switch at the CPE set to another configuration. This still falls under misconfiguration.
What happens if you do not have access to the CPE to determine that device’s configuration? Then you will have to set up the 3 devices you do have access to, for the same speed, mode and duplex by using the default setting of AN. If errors are still occurring on the network, then change the 3 devices’ settings to a Force mode, Force speed and duplex. The CPE device has to be one of the two. Performance without errors should be accomplished.
What happens if one of the devices supports strictly AN and does not offer a Force mode? Then the other 3 devices must be set to AN, which automatically negotiate the common denominator for speed and duplex. If the application requires a Force mode, all the equipment must be evaluated and possibly some replaced to support that mode.
What happens if you are certain you have all the equipment set to the same mode and the CPE device still won’t pass data as expected? A data sheet will be necessary, so that you can confirm the speed and other features. Sometimes business owners will upgrade Ethernet network equipment to gigabit without thinking of the ramifications to the network equipment recently purchased.
Also noteworthy, and not discussed in this blog, is how management (SNMP) can impact configuration of Ethernet network equipment in a completely different way. That’s a topic for another time.
Although Ethernet equipment such as media converters are pretty simple devices, one cannot overlook how they must interconnect to the end devices for successful linking and passing data. Doing a little homework and calling the vendors first to ask the right questions can minimize installation headaches.
0 Comments Click here to read/write comments