Testing Ethernet in MMU2 H

[asbco85]

Hi all.... For now, we are installing the new MMU2 H in the network, and during testing the link performance in handling Ethernet traffic, some links have less throughput than what is planned to have with the note that all the link condition (RSL, interference...etc) are normal. Any ideas on how to troubleshoot this and what is the probable cause. Also what is the the minimum accepted throughput (for example if the link capacity is 18 Mbps) Thanks in advance.

[lucasnow]

Hi asbco85, what is the frequency that is? Please pass details so we can help you!

[asbco85]

Hi asbco85, what is the frequency that is? Please pass details so we can help you!

The problemtic links working in 38 and 18Ghz

[qwertry]

Hi,

Check QOS settings, policing, port settings and etc. Also you can use Ethernet performance monitoring overview in ML Craft 2.19 and 2.20 for troubleshooting.

BR

[Wizeguy]

Hi all.... For now, we are installing the new MMU2 H in the network, and during testing the link performance in handling Ethernet traffic, some links have less throughput than what is planned to have with the note that all the link condition (RSL, interference...etc) are normal. Any ideas on how to troubleshoot this and what is the probable cause. Also what is the the minimum accepted throughput (for example if the link capacity is 18 Mbps) Thanks in advance.

Hi, some of the most common "faults/mistakes" when measuring the throughput on a L2 Ethernet link over Microwave:

1. RFC2544 (or other throughput measurement methods) is used and results are given as L1 or L3 and not for L2.
This means that. L1: The throughput results will seem to be a bit higher than expected for small frame sizes.
L3: The Throughput result will seem lower than expected for all frame sizes, worst for small frames.
Reason is that Microwave links usually do not transport the L1 preamble nor the inter frame gap.
L3 means that only payload are measured, but the link actually transport L2. i.e. usually the entire Ethernet frame excluding preamble and inter frame gap.

2. The E******* MINI-LINK Craft gives a throughput for the RL-IME at a certain frame size. This is not the throughput which is directly measurable.
It is the link speed which also need to transport the radio link bonding overhead etc, if such configuration is used.

3. Test instrument is not setup correctly or results are not analyzed concerning VLAN tagging and or use of layer of test (see 1.)
For instance sending in untagged frames into the equipment when using L2 switching means that a VLAN tag will be added to the frame transported by the link and thus reducing the measured throughput since the instrument is not counting the VLAN tag.

4. Measuring TCP/IP traffic when the system under test have bottle necks and delay, will lead to some other issues.
Here you will need to adapt the buffer sizes at the congestion points accordingly or change TCP window sizes.
A bit to long explanation needed for me to write here. Try to have a look at the "true speed" measurements that JDSU is driving, for more information.

If no one of the above items is helpful, please give me more to go on.
Like, what is the problem. What is configured, what was the results, what is expected and what and how was it measured?

[asbco85]

Hi Wizeguy,

thanks for your detailed explanation, noting that unfortunately we are not using any Ethernet analyzer during the testing, we only configured two ports in each side of the link on of them is connected to the physical LAN port in the NPU which will be connected to an ordinary PC, then to start a normal LAN test between the two PCs at the link ends.

[Cost_Reducer]

to benchmark an Ethernet link using Software based tools such as iperf gives you for sure no real throughput figures as such tools are working on higher layers, hence measuring or calculating the throughput on those layers. Also, things you need to consider when using SW based tools are generally the performance of the PC's used to benchmark the link. You need to adjust the TCP window size and and and...
To get results you can rely on I rather would test the link with an Ethernet Performance Analyzer giving the fact that you can select the way how to benchmark. Ideally you go for the RFC2544 test testing Latency and max throughput for different frame sizes.


BR

C_R

[asbco85]

Thanks a lot guys, the better thing to do is to use an Ethernet analyzer for testing these links but what can I say to the damn company's policy which don't like to pay additional cost for buying it.:)