[c-nsp] Nexus 5000 / Nexus 2000 SFP+ with LRM

[Michael Balasko]

Sorry to be late to the convo here, but I can personally attest that LRM'S work fine. 

Our 6513 with 6704's are glued to our N5K's with Xenpack LRM, and Merge Optics (Digikey special 280 per unit) SFP+ LRM's. 
We bought them because for our DC LRM is the sweet spot and Cisco doesn't offer LRM. (NPH till July)


 N5K-DC-02# sho interface transceiver
Ethernet1/1
    sfp is present
    name is MergeOptics GmbH
    part number is TRX10GDL0610
    revision is B00
    serial number is EM0838-00247
    nominal bitrate is 10300 MBits/sec
    Link length supported for 50/125um fiber is 220 m(s)
    Link length supported for 62.5/125um fiber is 220 m(s)
    cisco id is --
    cisco extended id number is 4

Ethernet1/2
    sfp is present
    name is MergeOptics GmbH
    part number is TRX10GDL0610
    revision is B00
    serial number is EM0848-00015
    nominal bitrate is 10300 MBits/sec
    Link length supported for 50/125um fiber is 220 m(s)
    Link length supported for 62.5/125um fiber is 220 m(s)
    cisco id is --
    cisco extended id number is 4

Ethernet1/3
    sfp is present
    name is MergeOptics GmbH
    part number is TRX10GDL0610
    revision is B00
    serial number is EM0838-00254
    nominal bitrate is 10300 MBits/sec
    Link length supported for 50/125um fiber is 220 m(s)
    Link length supported for 62.5/125um fiber is 220 m(s)
    cisco id is --
    cisco extended id number is 4

Sho cdp neigh - 

TBA05520665(COH-DC-6513-02-248)Eth1/1        168    T S       WS-C6513      11/4

Other side: (yes, that’s CatOs)

6513-720-02 (enable) sho cdp neigh 11/4
* - indicates vlan mismatch.
# - indicates duplex mismatch.
Port     Device-ID                       Port-ID                   Platform
-------- ------------------------------- ------------------------- ------------
11/4     N5K-DC-02                       Ethernet1/1               N5K-C5010P-BF

Sho port
Port  Name                 Status     Vlan       Duplex Speed       Type
----- -------------------- ---------- ---------- ------ ----------- ------------
11/4  Trunk NX5K-02 1/1    connected  trunk        full       10000 10G EDC1310

How? 

N5K-DC-02# sho run | inc uns
service unsupported-transceiver


Mike

-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Nick Hilliard
Sent: Monday, May 10, 2010 4:53 AM
To: cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] Nexus 5000 / Nexus 2000 SFP+ with LRM

On 10/05/2010 08:34, Marian Ďurkovič wrote:
> LRM SFP+ is just part of the stuff you need. For LRM to work, the 
> switch linecard must have appropriate EDC functionality. If it's not 
> there, it simply won't work.

To give some back-ground on this, LRM is long-reach multimode.  As it's multimode, modal dispersion comes into play pretty quickly, and even over relatively short distances, it causes severe signal distortion - this is one of the primary distance limiting factors of multimode.

On xenpaks, x2 and xfp, the dispersion compensation is performed on the transceiver (by the EDC), and you end up with a fully digital signal being transmitted from the transceiver's electrical interface to the line card.
However as the SFP+ form factor is really tiny, there isn't enough room to house various components such as an EDC or a CDR (clock / data recovery).
For SFP+, these components are housed on the line card, if at all, and in many cases the line card simply won't have EDC.  Perhaps the n5k main board doesn't have EDC processors, which would make it unsuitable for LRM.

> (One more "thanks" to all people who thought that analog interface 
> between SFP+ and linecard is a good idea...)

Fibre and transceiver deployments are all about choosing the appropriate technology.  If you need to run fibre over longer distances, doing this over MMF probably isn't the best idea.  I appreciate that lots of organisation have cartloads of legacy 62.5µ MMF and that they tend to be unhappy about the prospect of changing longer runs to use SMF, but 62.5µ wasn't designed for longer runs at very high speeds.

In some senses, you might as well complain that SFP+ isn't physically large enough to house enough lasers for LX4.  10G standards like LX4 and LRM were only created to try to deal with legacy plant deployments which weren't really designed for anything more than 100M-FX.  Anyone sensible MMF deployment done over the past couple of years will have been OM3, where you can use SR transceivers instead of LRM or LX4.

If you need distances longer than 200m, LR + SMF is a better choice of technology to use.

Nick
_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/