When is consumer 10GbE going to happen?

Discussion in 'Networking, Telephony & Internet' started by Smokin Whale, Sep 21, 2014.

  1. bucket23

    bucket23 Member

    Joined:
    Dec 10, 2008
    Messages:
    71
    I haven't had an issue with cat6A cables and gromming them into the rack.
    You just need to follow the manufactures instructions. If you follow them and there is an issue, then go to the manufacturer.
     
  2. Aetherone

    Aetherone Member

    Joined:
    Jan 15, 2002
    Messages:
    8,576
    Location:
    Adelaide, SA
    It's pretty much all simplicity. Media render box here with 10g, bit bucket over there with 10g, a stack of 1g clients hanging off the 1g ports. One simple cable to connect them each to the network, minimal bottlenecking when multiple clients decide to hit the storage silo.
    I'd love a 4+16 port version of the switch but I suspect they're going to need a die shrink and significant power efficiency improvement to achieve that passively.
     
  3. bcann

    bcann Member

    Joined:
    Feb 26, 2006
    Messages:
    5,666
    Location:
    NSW
  4. davros123

    davros123 Member

    Joined:
    Jun 18, 2008
    Messages:
    2,930
    fwiw, My tplink with 4 x 10G and 24 x 1G is passive.
     
  5. Aetherone

    Aetherone Member

    Joined:
    Jan 15, 2002
    Messages:
    8,576
    Location:
    Adelaide, SA
    How's it handle having the SFP slots stacked with 10GBASE-T transceivers?
     
  6. davros123

    davros123 Member

    Joined:
    Jun 18, 2008
    Messages:
    2,930
    Good point, I run optical. it's on mydesk and it's cool as a cucumber. It'll be in a rack soon with a POE below and a server above and so we'll see how it like being in tight spaces soon.

    Cpu is pretty low but I at last have most things.


    [​IMG]
     
    Last edited: Feb 1, 2019
  7. Quadbox

    Quadbox Member

    Joined:
    Jun 27, 2001
    Messages:
    6,218
    Location:
    Brisbane
    This is a common misconception by the way. Maximum signal speeds for electrical signals moving along a cable and optical signals moving along a fibre are both at the speed of light of the medium in question, which no matter what is lower than the vaccuum (and probably even air) speed of light. (for a 50ohm impedance coaxial cable for example the speed of light is 2x10^8m/s, so 2/3 the vaccuum speed of light, which is still *amazingly* fast)

    There's not an inherently higher latency in *copper* connections, hence why a twinax connection is just as low latency as a multimode fibre connection is (slightly more so, in fact). The latency comes in because making these signals work on twisted pair involves a hell of a lot of trickery to manage noise, crosstalk and dispersion, and so the electronics is just flat out more complicated, once you're up at 10GB. An optical transciever, on the other hand, is comparatively simple electrically (the fact you've got a laser diode in there doesnt really increase latency particularly, and there's a lot less signal processing involved).

    EDIT - (to clarify, twinax on the other hand runs on coaxial cables that are well and truly capable of supporting RF without all the stuffing around. Not that different to antenna pigtails.)
     
    Last edited: Feb 5, 2019
    Luke212 likes this.
  8. davros123

    davros123 Member

    Joined:
    Jun 18, 2008
    Messages:
    2,930
    Plus opticals go 300m and only cost $22 for 10G Intel compat. modules at FS.com and arrive in 2 days for just $8 ;)
     
  9. evilasdeath

    evilasdeath Member

    Joined:
    Jul 24, 2004
    Messages:
    4,855
    I'd love for this to happen
    https://www.extremetech.com/extreme...ber-for-faster-networks-more-accurate-sensors

    subsea cables would be great, but imagine a cut, would suck in crud and would likely need to lay the whole thing end to end.

    But yes Quad, over Fibre the signal is serialized 64b/66b (https://en.wikipedia.org/wiki/64b/66b_encoding) efficient encoding and very low overhead

    vs
    10G-BASE-T which is (from wikipedia)

    The 802.3an standard specifies the wire-level modulation for 10GBASE-T to use Tomlinson-Harashima precoding (THP) and pulse-amplitude modulation with 16 discrete levels (PAM-16), encoded in a two-dimensional checkerboard pattern known as DSQ128 sent on the line at 800 Msymbols/sec.[52][53] Prior to precoding, forward error correction (FEC) coding is performed using a [2048,1723]2 low-density parity-check code on 1723 bits, with the parity check matrix construction based on a generalized Reed–Solomon [32,2,31] code over GF(26).[53] Another 1536 bits are uncoded. Within each 1723+1536 block, there are 1+50+8+1 signaling and error detection bits and 3200 data bits (and occupying 320 ns on the line)

    All this takes a fraction longer to do than 66b/66b

    This might be a few microseconds which is non issue to most people, but is a measurable difference. Additionally needs more power, again not much in a small situation, but i can tell you in a situations with 100 of these things it's not worth it just to be able to use copper.

    That being said, i used early generation 10G fibre interfaces, cisco xenpaks and gbics :) SFPs are a shit load better!!!
     
    Last edited: Feb 5, 2019
  10. davros123

    davros123 Member

    Joined:
    Jun 18, 2008
    Messages:
    2,930
  11. evilasdeath

    evilasdeath Member

    Joined:
    Jul 24, 2004
    Messages:
    4,855
  12. cvidler

    cvidler Member

    Joined:
    Jun 29, 2001
    Messages:
    12,215
    Location:
    Canberra
    Cool thing is while it appears to be complex math, it's easy to implement in hardware so can happen at full line rate (no performance impact, no CPU usage).

    The fancy stuff done to copper 10GE, is all done in hardware too, but still results in more overhead than the fibre encoding.
     
  13. evilasdeath

    evilasdeath Member

    Joined:
    Jul 24, 2004
    Messages:
    4,855
  14. Quadbox

    Quadbox Member

    Joined:
    Jun 27, 2001
    Messages:
    6,218
    Location:
    Brisbane
    If you're the sort of person who has a signal generator and a reasonably high frequency capable oscilloscope btw, it's actually relatively easy to do an experiment to determine the speed signals propagate in the cable. It's one of those little physics experiments that I feel's worth actually doing yourself just to convince yourself it's true. Only need about a 100MHz oscilloscope, two ten metre lengths of (50ohm) coax, one t-piece, one 50ohm terminator and one coax shorting end. T-piece in the middle of the two cables, with the oscilloscope attached to it. one end to the signal generator, the other end you can either short, leave open, or put a terminator on. Capture the outgoing pulse and its reflection in the same time sector. It's quite interesting to see the behaviour

    EDIT - of course if you've got a pair of 20 or 30m long coax leads it's even easier :p
     
    Last edited: Feb 7, 2019
    cvidler likes this.
  15. cvidler

    cvidler Member

    Joined:
    Jun 29, 2001
    Messages:
    12,215
    Location:
    Canberra
    used to do that with the 10base2 network at school. pull the terminator, watch everyone start complaining.
     
  16. evilasdeath

    evilasdeath Member

    Joined:
    Jul 24, 2004
    Messages:
    4,855
    ahh the days of school

    We had a classroom full, about 25 PCs, which were these thin clients at the time, they just did touch typing classes and stuff like that. 10Base2 network so it would fall apart all the time when a connector was loose somewhere. Add to the fact that all 25 PCs were on the same circuit breaker!, so if you turned them on in lots of 5 and let them boot it was fine, but turn on all 25 at the same time and it would trip the breaker.
     
  17. Quadbox

    Quadbox Member

    Joined:
    Jun 27, 2001
    Messages:
    6,218
    Location:
    Brisbane
    This is why all the cool kids were using tokenring :p
     
  18. cvidler

    cvidler Member

    Joined:
    Jun 29, 2001
    Messages:
    12,215
    Location:
    Canberra
    but if you disconnect the cable, the token would fall on the floor and you'd be stuffed.

    we're blessed with todays' individually switched collection domains, and non-zero latency wifi.
     
    Aetherone likes this.

Share This Page

Advertisement: