On Wed, Jul 04, 2001 at 05:27:17AM -0300, Rubens Kuhl Jr. wrote:
> There is the documented weirdness of not allowing each VLAN to have
> interface a unique MAC, the MAC of the gateway is the global to all router
> interfaces on SUP2. MSFC2 architecture is really good, much better than
> MSFC1 and not just a faster processor; PFC2 has addressed the MLS
> demand-driven issues of PFC1 by going topology-based (CEF) and have more
> thinking on how to keep the MSFC safe, doing things such as the ARP
> throttling.
Don't forget that the SUP2 restricts 16 HSRP groups in hardware.
What other functionality is limited by this so-called chip?
> We are investigating some ACL x Fragment handling issues but cannot say
> right now that there are problems, just that there may be some.
GSR-esque?
> When comparing just Ethernet routing, GSR full range of possibilities of
> what can happen to a packet may be more or as complex than Cat6K/SUP2. What
> is really strange is the way OSR do WAN routing, that's probably you are
> talking about... but as they seem to add complexity serially on the life of
> a packet and not by expanding the possible ways where packets go or are
> changed, this would help in keeping it a dependable product.
Hrmn... so you are talking about GSR Fiber to BMA
Fiber -> Transceiver -> Framer -> POS ASIC -> FIFO -> BMA
then GSR "buffer" management
BMA -> Buffer -> IP + Buffer Header -> L3FE
onto the switch fabric
L3FE -> FIA -> Scheduler -> SLI ASIC -> Tranceiver -> XBR -> LC SLI ASIC ->
Port -> Transceiver -> Dest FIA -> Dest LC SLI ASIC
and through the output queue
BMA <-> FIA -> (possible leap to R5k in multicast) -> POS ASIC -> Framer ->
Fiber
Yeah I guess it can't get any more complicated than that!
> >Does this bug affect SupIOS (Native)?
>
> Not tested, but this would probably affect the SupIOS version that was
> released at the same time as 6.2(2). If it also affects 6.1(x), probably
> all SUP2-capable versions of Hybrid and SupIOS are affected.
I haven't tested Native code, ever. Anyone have any comments?
> As of today, OSM modules still requires Hybrid. SupIOS is the direction
> Cisco is going with no ETA...
Expect it to happen relatively soon. I don't think Cisco can progress
with CatOS + MSFC IOS from a simply provisioning / management functionality
standpoint.
> But as complexity goes, MPLS on OSR with Sup2 forwarding only Ethernet/IP
> is a big hack on how to map LSPs to VLANs, tags to IPs, MPLS tables to
> FIBs... sounds like making round air filters fit into square holes.
Does any other vendor do AToM or EoMPLS? There are some really good air
filters. There are some really bad Layer 3 Switches.
> Versions on that bug id were of non-released versions; we've come to this
> running 12.1(7a)E1 and 6.2(2), but aside that everything is equal including
> router configs.
Specific LC's tested? Anything else to know?
-dre
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:43 EDT