2

I have a 10G XPF+ optical cable with market updates from a stock exchange. This cable goes into a switch, which then mirrors every packet to a couple of computers over two ports. The problem with using a switch for mirroring is that there is latency overhead, even with a pass-through switch (~200ns).

Are there "optical" solutions (I'm thinking of a beam splitter of some sort) which would allow for close to zero latency 10G mirroring?

Randomblue
  • 1,135
  • 5
  • 15
  • 33

3 Answers3

7

Are there "optical" solutions (I'm thinking of a beam splitter of some sort) which would allow for essentially zero latency 10G multicast?

No.

Chopper3
  • 100,240
  • 9
  • 106
  • 238
  • 7
    Ye' cannae' break the laws of Physics, Jim! – tombull89 Sep 09 '13 at 12:09
  • "beam splitter" effectively splits the photon stream, so only half the photons goes each way. this possibly breaks the signal sent. Amaright ? – mveroone Sep 09 '13 at 12:18
  • @Randomblue - certainly prismatic optical networking adds VERY little, not zero, but very little latency to the equation - the issue here though is that you still need to turn the data into regular ethernet and there's no real way of doing that without some form of switch involved - which inherently brings latency. Bay Networking used to sell a range called ONS that was designed for just this thing but I think that line got dropped or sold-on when they went bust - perhaps you could look around as I'm sure there are better products to use than 'regular' switches. Good luck. – Chopper3 Sep 09 '13 at 12:22
  • @Randomblue I'm not sure I see your point - even light has a finite speed. – Dan Sep 09 '13 at 12:22
  • @Randomblue - yes, but you need *a* switch of some form to do that sorry – Chopper3 Sep 09 '13 at 12:53
  • That's an error message waiting to happen "Insufficient critical photons detected". – Tom O'Connor Sep 09 '13 at 13:02
  • Randomblue: Ethernet over Fibre is not an analog signal. It's a digital one... – Tom O'Connor Sep 09 '13 at 14:19
  • @Randomblue I'm very aware of optical network taps, I work for the biggest buyer of them in Europe, they don't provide multicast-like functionality at all, they're more similar to a span port. – Chopper3 Oct 25 '13 at 22:39
  • @Chopper3: "I work for the biggest buyer [of optical network taps] in Europe" -> Do you work for GCHQ by any change? Or the NSA? – Randomblue Nov 08 '13 at 06:21
  • 4
    @Randomblue If you come into chat, we've started compiling a list of people who've asked Chopper3 direct questions and then disappeared. I'd suggest locking your doors. – Dan Nov 08 '13 at 11:32
5

My old firm manufactures a port replicator that solves the problem you're having. It's a Layer 1 device that can take a market data feed and multiplex to a number of ports at <10ns latencies. This bypasses the Arista switch for one-way market data consumption.

ewwhite
  • 194,921
  • 91
  • 434
  • 799
1

Yes, port mirroring is possible using network taps, with latency an order of magnitude less than a using a switch. There is a video demonstration here.

Randomblue
  • 1,135
  • 5
  • 15
  • 33