re: Alcatel Pushes Ethernet-Over-MPLSScolwell: " There is a big difference between a forwarding table that can hold full routes and one that can hold your own infrastructure routes."
I didn't say there wasn't a difference. I simply stated that the 7450 must support routing to be able to support MPLS signalling, thatGÇÖs it.
Scolwell: "All the 7450 needs is routing for signalling which can be done safely in s/w."
Incorrect. The 7450 needs to be able to route packets for all IP management and control traffic, e.g. OSPF, SNMP, ICMP, etc. Also, routing is always done in software using a routing protocol or an automated OSS (or manually). Routing refers to the calculation of a path(s) to a specific destination. Packet forwarding/switching using software or hardware based tables is a separate issue.
Scolwell: "The saving is in avoiding the need for carrying an internet routing table or many fragmented enterprise VPN routing tables."
Yes, or more accurately, the saving is in avoiding the need to support BGP and RFC2547.
re: Alcatel Pushes Ethernet-Over-MPLSParticle_man said: " So last I heard, an ethernet frame didn't have a label. In this case MPLabelS is going to struggle a bit."
I think perhaps you are assuming that the 1662 or 7450 only support Ethernet switching and therefore some proprietary mechanism must be used to encode the MPLS label and forward the packets.
This is not the case, the 1662 supports standard MPLS switching and Ethernet access over a STM-4 RPR ring. The 7450 supports standard MPLS switching over Ethernet.
The MPLS packet including label(s) is carried inside an Ethernet frame. Packets are switched at the Data Link/MAC layer based on the destination address in the Ethernet header. Packets are switched at the MPLS layer using labels (the Ethernet header is stripped-off before and added-on after MPLS switching takes place).
re: Alcatel Pushes Ethernet-Over-MPLSComrades, I don't thknk there's anything proprietary going on in the 7450 (or, at least no more than usual - remember all this VPLS stuff is still in draft).
The way I understand this is that there is a move (not just within Alcatel) towards using MPLS Control Plane protocols to set up Layer 2 Ethernet services. In order to achieve customer traffic separation (and help with stuff like QoS and OAM), VPLS encapsulation is used (which in turn is based on Martini).
MPLS control plane messages (LDP or BGP for signalling of VPLS tunnels) will need to be routed across the switch. Do they pass across a "fast" or "slow" path? Not sure I can answer that because it depends on the way the product works, and if caching is done, etc. It shouldn't matter that much because these packets are only used to set up VPLS tunnels, which should be very long-lived.
Of course there needs to be a Denial of Service protection mechanism to protect the Control Plane against deliberate flooding, but that applies to both Fast and Slow implementations.
So the idea is that the 7450 is very much VPLS-focussed. But then, that's not a big surprise as Alcatel (Timetra) had a big hand in VPLS development.
re: Alcatel Pushes Ethernet-Over-MPLSTo judge from tmc1's comment it appears moot. That aside, there are actually a few boxes out there that do as suggested. They typically put the CPU on a switch port with ACLs etc that give it the control traffic suitably (one hopes) throttled. Don't take me for an advocate :-)
-------> Good point. Yes Alcatel may have reprogrammed the fast path packet forwarding chips not to forward packets based on IP addresses and therefore force IP packets to use the slow path. I personally doubt this though, what are the benefits of doing this? Surely it makes more sense to let control/mgnt traffic use the fastpath in order to take advantage of its policing, ACL, and queuing capabilities?
re: Alcatel Pushes Ethernet-Over-MPLSComrades, Is there anything proprietary going on in the new Alcatel RPR product? Well, probably. If my understanding of RPR is correct you cannot mix Vendor A and Vendor B RPR boxes in the same ring.
I'm sure there are exceptions to this (not least because there are technology partnerships like Native-Alcatel and Cisco-Riverstone).
Should we expect interoperability at this level? Well, I guess not. If you think about it, you don't usually mix two different vendors' SONET or SDH ADMs in the same ring either.
Between rings is different - as long as the RPR implementation isn't "pre-standard" then you're in good shape to achieve interoperability. But I think you also lose QoS at that point if I'm not mistaken. Some vendors have announced fixes for this, but they are - you guessed it - proprietary :-)
I'm not up to date on RPR the way I used to be, so please feel free to correct me.
Some of these combinations seem to mix the stack up a bit and at first glance seem to be a bit obscure especially at the framing layer - PWE3 variants for instance. With people trying so hard not to use the term SONET (even if using it to only for framing) it is sometimes hard to figure out what layer is doing what these days.
re: Alcatel Pushes Ethernet-Over-MPLS Say_Yes_2_MPLS & all,
Some clarifications/confirmations from the past messages:
Alcatel 7450 ESS is a MPLS-based Ethernet Service Switch which fully supports L2 MPLS-based services like VLL (draft-martini) as well as VPLS (draft-vpls). Both strictly aligned with standards (see MPLS Forum 2003 tests).
Naturally it fully supports IGP like OSPF/ISIS TE for proper MPLS/MPLSTE/MPLSFRR setting-up, etc. Its only difference against Alcatel 7750 SR (same product family) is the support for L3 VPNs (RFC2547 bis), which is translated (as already mentioned) with no-BGP support.
Other differences like specific/optimized interfaces for Metro Space exists on A7450 because its positioning in MetroEthernet arena (i.e. more dense GigaEthernet modules, etc.)
In regard where/how IP routing/signalling is done, as every packet processing in the boxes (A7450 & A7750) it is done @wire-speed on Network Processor presents en each line card.
Consequently performance is is not affected because enabling any feature like filters, QoS processing, etc., on the box.
Other remarkable features of the boxes (77750 & 7450) includes Hierarchical QoS -8000 Queues per interface-module- & MPLS/Service OAM tools -Service, LDP, SDP, MAC pings-.
I didn't say there wasn't a difference. I simply stated that the 7450 must support routing to be able to support MPLS signalling, thatGÇÖs it.
Scolwell: "All the 7450 needs is routing for signalling which can be done safely in s/w."
Incorrect. The 7450 needs to be able to route packets for all IP management and control traffic, e.g. OSPF, SNMP, ICMP, etc. Also, routing is always done in software using a routing protocol or an automated OSS (or manually). Routing refers to the calculation of a path(s) to a specific destination. Packet forwarding/switching using software or hardware based tables is a separate issue.
Scolwell: "The saving is in avoiding the need for carrying an internet routing table or many fragmented enterprise VPN routing tables."
Yes, or more accurately, the saving is in avoiding the need to support BGP and RFC2547.