11-03-2011 05:42 AM
What is the Rx Unstressed Sensitivity that's shown on the spec sheet for an SFP?
For example the Brocade 4Gbit/Sec ELWL (30 KM) sfp shows (http://www.brocade.com/downloads/documents/data_sheets/product_data_sheets/SFP_4GB_ELWL30Km_DS_01.pdf):
Unstressed sensitivity: 29 μW, -18.0 dBm
Does this mean if you look at the receive power on the sfp and it's above 29 μW you should be fine?
The reason I'm asking is IBM (we have IBM rebranded 48000) saying they dont' like that our long distance sfp show less then 200 μW on our ISLs, but the distanace is 25KM so I'd expect the receive power not to be really high. I'm not aware of any issues (we've not seen any CRC or Encoding errors). Looks like short distanace SFPs have a different unstressed sensitivity, but I can't seem to tell what this really means or what a best practice / ideal receive power reading should be.
11-03-2011 06:20 AM
I've had a similair issue recently, although our vendor is HP.
HP uses -5dBm to -11 dBm as a range which is ok for RX power.
Most of our long haul ISL hover around -7 dBm (or 200uW)
11-03-2011 06:53 AM
Here is the reading for one ISL from each switch (that are connected to each other):
low high low high
Temperature: 26 Centigrade -3840 19200 -2560 17920
Current: 29.076 mAmps 12.620 100.238 20.000 63.246
Voltage: 3215.6 mVolts 50.0 5000.0 150.0 4500.0
RX Power: -9.0 dBm (126.2 uW) 0.0 uW 0.0 uW 0.0 uW 0.0 uW
TX Power: 2.3 dBm (1701.5 uW) 4.0 uW 1584.9 uW 6.3 uW 1000.0 uW
low high low high
Temperature: 30 Centigrade -3840 19200 -2560 17920
Current: 32.034 mAmps 12.620 100.238 20.000 63.246
Voltage: 3235.1 mVolts 50.0 5000.0 150.0 4500.0
RX Power: -9.3 dBm (117.6 uW) 0.0 uW 0.0 uW 0.0 uW 0.0 uW
TX Power: 2.7 dBm (1875.9 uW) 4.0 uW 1584.9 uW 6.3 uW 1000.0 uW
So this is one example where they want it higher but they don't seem to understand that the issue is mainly with the distanace. I'm assuming if the Rx power needed to be higher the 30km SFPs would be made to output more power to over come the distanace limitation.
I'm thinking of testing out some 50km SFPs, but it seems silly unless I can be told how my current receive power is at issue.
11-03-2011 02:08 PM
my understanding is that the SFP should work if the SFP gets an RX signal of -18dbm.
You are currently at a level of -9dbm.
The spf values fit to your distance of 25km.
You have confirmed that the links works fine without any errors. My questions is why did IBM tells you that you have an issue with your link?
11-04-2011 04:32 AM
This all started when we have major performance issues for any traffic that crossed the long distance ISLs. Originally IBM could not find anything and they saw the receive power and wanted us to "fix" that. When we had the issue a month later, we noticed Class 3 Discards on the ISLs and a few other ports. IBM then said we probably have a slow drain device and to fix that. Some other people at my company wanted to still "fix" the low power "issue" IBM saw, even though the new Class 3 Discards were found. I had a hard time going with trying to fix something that could not be fixed (because it was related to our distance and sending power of the SFPs). I also wanted this 200 uW issue/best practice in some official document. I mean, should all customers have to wait until they have issues, talk to support, to find out they should have more then 200 uW for the Rx power? Things were just not adding up for me. We have our ISLs in groups of two (trunking) so taking one out to replace cases the trunk to change back to a standard ISL, thus causing a disruption in traffic as the new routing is propagated to all the edge switches, etc. We did replace a few SFPs, the receive power went up a little bit but not to the 200 uW that IBM wants.
We've sense enabled the bottleneckmon on our ports and need to reproduce the issue to see what it says. It's suspected the slow drain device might be our HP SVSP storage virtualization and the issue has happened when doing an Exchange restores during the day. Our HP SVSP is going away within a month, but I'd like to figure out what's going on. I also wanted a better understanding to know what the SFP's were capable of to know if we really were getting near the marginal capabilities of the optic.
11-04-2011 10:48 AM
Did you try to increase the buffer on your long distance links?
This will may reduce the discard issue and inrease the amount of data you can push over the link.
11-04-2011 11:00 AM
They are already higher then default for the distanace. I'm not sure increasing it will fix the issue if we really do have equipment that's a slow drain device as is suspected. The issue has only come up when doing Exchange restores during the day. If restore are not being done we have no issue of any kind.
11-04-2011 11:03 AM
it will not fix your issue. It will only reduce the impact a bit.
Your restore will overload your storage system. This will cause back preasure and at the end you will see discards. The problem with discards is that other servers or flows are affected as well.
07-25-2014 12:38 AM
I've also found some infos for Brocade SFPs. Unfortunately I can't copy my excel table here. For example I have following values for a 8 GBit/s SWL:
Average Power: -8 dBm
OMA: 302µW / -5,2 dBm
Optical return loss: -12 dBm
unstressed sensitivity: 76 µW / -11,2 dBm
SRS OMA: 148 µW / -8,3 dBm
Can someone explain these values to me? I couldn't find much help when googling for these parameters. I think for receive power the unstressed sensitivity should be the value which is important and which should not be lower then specified. But what about OMA and optical return loss?