3 ms·
It would have been fun to be a fly on the wall at Cisco when they first realized people were jamming devices with heaters in SPF sockets.
by oasisbob 5y ago
It would have been fun to be a fly on the wall at Cisco when they first realized people were jamming devices with heaters in SPF sockets.
- eqvinox 5y agoCisco didn't "realize" this was happening, they specified, had manufactured, and sold the very SFPs the GP post is describing. SFP slots in routers are designed with this thermal load in mind. Which is also why plugging a large amount of DWDM optics into a datacenter switch is a bad idea. Datacenter switches are _not_ designed with this in mind. You run into risks of both overheating the switch as well as overloading the PSUs. A small number of high-power optics ain't gonna break the switch though. And: 10Gbase-T SFPs have horrendous power consumption, even worse than DWDM SFPs. At these speeds, the signal over copper is mostly mush, and the PHY contains a none-too-trivial analog & digital signal processor. Which, again, is where limitations for 10Gbase-T SFP usage come from. If at all possible, avoid this shit — there's absolutely no reason to have 10Gbase-T inside a rack, for example. Just use DAC cables or SR optics.
- oasisbob 5y agoI'm being a bit glib, but from the point of view of someone who used SFP optics once in a while, it seemed like the whole "standard" was a bit of a mess. If it was known that vendors would try to push their thermals past what the device cage would support, why not address that better? The general impression I was given is that thermal loads were one of the primary reasons given for vendor-locking pluggable optics, and that the original SFP spec didn't even address it.