[j-nsp] 3rd party DWDM SFPs

[Richard A Steenbergen]

On Wed, Jun 11, 2008 at 11:55:43AM -0400, Chuck Anderson wrote:
> Does anyone have any experience, good or bad, with using 3rd party 
> DWDM SFPs in M120 or M10i Gigabit Ethernet PICs?  Specifically, I'm 
> looking at using these SFPs:

The only real gotcha I've hit in Juniper's support for 3rd party optics is 
that they don't expose the actual optic EEPROM contents (with things like 
vendor name and part number) via the CLI. This make it very difficult to 
do remote inventory management (aka figure out what color optic is in what 
slot), short of actually pulling out the optic and reading the label 
(which tends to negatively impact the function of the port :P).

If the part number doesn't match something they expect to see as a 
Juniper-branded optic, the only thing you can see is "UNKNOWN", like so:

  PIC 0                   BUILTIN      BUILTIN           1x 10GE(LAN/WAN)
    Xcvr 0                NON-JNPR     KCL005R           UNKNOWN

If you actually want to see the optic EEPROM details you have to look at 
it from the PFE shell, like so:

ADPC1(re0.routername vty)# show xfp 1 identifier
                                    ^-- NOT 0-based port number! Argh! :P
  Identifier:             XFP
  Connector type:         LC
  Encoding:               Unknown
  Vendor name:            FINISAR CORP.   
  Vendor OUI:             009065
  Vendor PN:              FTRX-3811-328-TR
  Vendor Rev:             00
  Vendor SN:              KCL005R         
  Juniper PN:             NON-JNPR
  Juniper Rev:            
  Date code:              071114  
  Ext identifier:         0x90
  Wavelength:             1554.95 nm
  Wavelength tolerance:   0.100 nm
  Max case temperature:   70 C
  Transceiver code:
    SONET long haul:      G.959.1 P1L1-2D2 
  Min bit rate:           9.9 Gb/s
  Max bit rate:           11.1 Gb/s
  Length SMF:             80 km
  Length EBW:             0 m
  Length 50:              0 m
  Length 625:             0 m
  Length copper:          0 m
  Device technology:      0x76
  Max power dissipation:  3500 mW
  Max power PD mode:      1500 mW
  Max current +5V:        350 mA
  Max current +3.3V:      400 mA
  Max current +1.8V:      800 mA
  Max current -5.2V:      0 mA
  CC base:                0x8E
  CC ext:                 0x36
  Diag monitor type:      0x08
  Aux2 type:             +5V voltage
  Transmitter technology: 1550 nm EML
  Wavelength control:     no
  Transmitter:            cooled
  Detector type:          APD detector
  Rx power measure type:  average power
  Supported options:
    CDR 9.95Gb/s:         yes
    CDR 10.3Gb/s:         yes
    CDR 10.5Gb/s:         yes
    CDR 10.7Gb/s:         yes
    CDR 11.1Gb/s:         yes
    Line loopback:        no
    SFI  loopback:        yes
    FEC control:          no
    Soft Tx disable:      yes
    VPS:                  no
    VPS LV mode:          no
    VPS bypass mode:      no
    CMU:                  no
    Soft power down:      yes
    Wavelength tunable:   no
    Transmitter tunable:  no
    BER support:          no

I was actually all set to to complain about a problem I was hitting in 
earlier code (8.4 definitely had this issue) where they made a silly 
assumption about the max length of the part number string and truncated it 
at 11 characters (RIGHT before the part that would identify the color of 
the optic). However, I just checked this on a 9.0 box and it seems to be 
reading the complete part number string and even identifying the 
wavelength information. So basically just be prepared to have to drop to 
the PFE shell to see the actual optic details, or help nag them to expose 
these details via the CLI.

Other than that one, I've got many hundreds of 3rd party optics of all 
varieties in many Juniper's without any issues.

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)