Network musings - Optical power monitoring

For a change here are some musing about the rather exciting things called SFP+ transceivers. Ok, they're not that exciting but they power the internet and almost certainly appear on most networks with more than 1 switch and any kind of bandwidth usage.  They're overlooked by many however they hold one of the many keys to network reliability, availability and performance.

One of the newer features of SFPs is DDM/DOM (written about everywhere) however this gives a really good insight into how close to failure your network is yet the majority of people probably do not monitor things like TX and RX powers of optics as a link light clearly proves everything is working well.  Obviously they don't need polling every 15 seconds and probably not every 15 minutes but graphing an hourly RX power over a long period of time will tell you whether your fibre is dirty, has been kinked, trapped or become slightly loose. While the OIDs for doing this via SNMP are widely published for many manufacturers it took me a while to find HP/Aruba ones. Here are the Aruba/HP Procurve ones:

.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.14.<intno> # TX power
.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.<intno> # RX power

While this tells you little without knowing the optics in question harvesting the data will tell you when you will have a problem if things get worse.  Also if you have several connections on the same fibre route harvesting the values via SNMP scanner can tell you instantly if one of them is problematic.  Using your favourite SNMP scanner such information can be collected quickly.  I use braa as it's free and works fairly well from the command line.  Below is the output from a quick scan of 36 switches:

./braa public@172.16.0.0-172.16.0.255:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51

172.16.0.129:0ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4365
172.16.0.130:0ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4463
172.16.0.131:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-3758
172.16.0.132:3ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4732
172.16.0.133:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5094
172.16.0.134:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-3775
172.16.0.135:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5075
172.16.0.136:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4680
172.16.0.137:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4013
172.16.0.138:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5579
172.16.0.139:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4837
172.16.0.140:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4461
172.16.0.141:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5340
172.16.0.142:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5034
172.16.0.143:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5034
172.16.0.144:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4653
172.16.0.145:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5128
172.16.0.146:2ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4989
172.16.0.147:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5710
172.16.0.148:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4873

172.16.0.67:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-6111
172.16.0.68:0ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4177
172.16.0.69:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4803
172.16.0.70:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-6307
172.16.0.71:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-6270
172.16.0.72:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-6104
172.16.0.73:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4784
172.16.0.74:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4231
172.16.0.75:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-6401
172.16.0.76:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5201
172.16.0.77:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-6819
172.16.0.78:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4463
172.16.0.79:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-5494
172.16.0.80:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4584
172.16.0.81:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-4497
172.16.0.82:1ms:.1.3.6.1.4.1.11.2.14.11.5.1.82.1.1.1.1.15.51:-8262

The last 16 have an additional patch in the middle so should be slightly worse than the first 20 however the last value stands out as a full 4dBm worse than the best on exactly the same fibre route. This needs some investigation as it is so much worse than the next worst value (although with a minimum RX of -14000 it is not a service problem... yet).

Luckily the old days of cletops fibre cleaners are gone and the cheaper, faster one-click cleaners have made cleaning the entire fibre path and optics possible without too much hassle.  I've no doubt with a bit of cleaning this value will rise into line with the others.

While getting all excited about optics I should probably mention the economic benefits of bi-directional optics in the LAN. Supposing full price is paid for moderately high end 3rd party 10GE 10km SFP+ conventional pairs of optics will cost c£150 whereas Bi-di pairs come in at c£300 however the fibre demand is halved.  Does your bulk fibre installation on long runs cost less than £150 per core particularly when things like rack space are factored in?  Going low end on optics swings it further into bi-di territory with list prices being <£60 per pair for conventional and <£100 for bi-di.