Aricent recently conducted a webinar on the topic of MPLS-Transport Profile. We had several hundred people participate and they asked more questions than we were able to answer in the time we had! Here are answers to the ones we didn't get to in the webinar.
What is the difference between MPLS-TE and MPLS-TP?
TE stands for Traffic Engineering. Traffic Engineering is the term used to describe the methodology of pre-determining the routes that user data packets (traffic) will take through an operator’s network. This is done by binding the user data packets to specific MPLS tunnels (LSPs) at the ingress point in the network. The tunnels are laid out with constraints about which nodes they will go through in the network, which nodes they will avoid, what QoS treatment will be applied, what bandwidth guarantees are available, etc. The result is that the paths that traffic takes through the network are controlled well and the operator thereby has much more predictable network behavior.
TP stands for Transport Profile. A Transport Profile refers to a set of attributes for the technology that make it usable in a transport network. One aspect of the Transport Profile is the capability to do Traffic Engineering. Other aspects of the Transport Profile are connectivity verification, path tracing, tools for problem isolation, performance measurement, performance management, fault management, etc. TE is thus one toolset within the set of TP attributes.
How does MPLS-TP compare to Carrier Ethernet networks (which have extensive OAM and redundancy)?
The primary advancement of MPLS-TP over traditional IP/MPLS has been the clear definition and standardization of OAM techniques and toolsets. The OAM standards for MPLS-TP are currently standardized with interoperability also established between early adopters. The OAM standards borrow from what has worked well for Carrier Ethernet and SDH/SONET and apply these specifically tailored for MPLS. Thus, the MPLS-TP OAM capability is comparable and equal to that of Carrier Ethernet.
Which specific performance metric is similar to SONET/SDH with the MPLS-TP approach?
From a performance perspective, the most important metric is the time to move traffic from a failed path to its redundant protection path. This must be accomplished within 50 milliseconds i.e. there must be less than 50 milliseconds of outage for the specified service (an LSP or a pseudo wire) in the network. This is made possible in MPLS-TP by a clear definition of how service redundancy is to be provisioned, how service monitoring must be done, and how failover of traffic to a protection path must be carried out.
What are the key pass/fail criteria of an MPLS-TP network? Do you have a test plan you can share with me?
The key criteria for an MPLS-TP network would be – (a) ability to provide OAM and (b) ability to guarantee sub 50 milliseconds service restoration. We will be happy to work out test planning strategies based on specific equipment or networking needs with you. Some standard interoperability scenarios can be seen at EANTC and ISOCORE web sites.
Given the various elements large incumbents have in their respective networks, what are the challenges of incorporating MPLS-TP into their networks?
The challenges are the same as for bringing in any new concept or technology into an existing and functional network. With a pure Ethernet network, it is more difficult since the forwarding plane must start dealing with labels. With an MPLS network, this is much easier since the forwarding plane is more or less identical. A gradual software upgrade path with gradual introduction of the OAM concepts can be put into place.
Which of the following is used to create MEPs – IEEE8021.ag or Y.1731 BFD?
MEPs can be realized using Y.1731 for MPLS-TP. IEEE 802.1ag can be used to create MEGs but it does not define the performance management and fault management attributes like done by Y.1731 and therefore it does not lend itself for use in MPLS-TP networks in the way that Y.1731 does. BFD can be used to realize the same results as Y.1731 but the definition of an MEP or MEG is not clearly laid out in BFD as it is in the Y.1731.
What is the convergence time for dynamic failover to an alternate path using MPLS-TP?
Transport network requirements are for failover in less than 50 milliseconds and this has been demonstrated and proven to be practical, realistic and feasible with MPLS-TP.
Given the scalability limitations of "connections" e.g., ATM VCs, Ethernet EVCs and MPLS PWs, what mechanism does MPLS-TP define to provide a mesh of PWs?
MPLS-TP does not do anything more than MPLS (or IP/MPLS) from the scalability aspect. It is as good or as bad as MPLS. MPLS is inherently more scalable than provider bridged Ethernet because the 32 bit label space is a highly scalable alternative to the 4094 service VLANs limitation of provider bridged Ethernet. When compared to ATM, there are 2 primary aspects wherein MPLS scores in scalability – (a) ATM cells are very small whereas MPLS packets can be very large, because of which the transport overhead is far lesser with MPLS than with ATM, and (b) MPLS provides a multi level hierarchy like pseudo wires within LSP tunnels or H-LSPs (Hierarchical LSPs) which offers better scalability than a single flat space for addresses or labels.
What does an operator with Carrier Ethernet do? Must they migrate to MPLS-TP? If so, how would they do that?
This largely depends upon the challenges and necessities faced by the operator. If the subscriber base is small and the number of services that have to be offered in the medium term visible future is in the sub 4000 range, then the operator can continue with carrier Ethernet and there really is no strong business case to migrate to MPLS-TP. However, if the subscriber base is growing quickly and service scalability is an issue, then, MPLS-TP provides a viable and efficient solution for the operator. The extent of difficulty in the migration will need to be assessed based upon the design of the network elements. However, with vendors providing support for MPLS-TP and out-of-the-box solutions becoming available from silicon vendors, OEMs and software suppliers, the operator’s challenges in the migration can be significantly alleviated.
What ensures the interoperability between MPLS/TP and MPLS core: Can you elaborate what you mean by mapping client LSP to server LSP? Is that LSP tunneling? Is MPLS/TP compatible with IPv6? What are the benefits of MPLS/TP over other technologies?
Fundamentally, MPLS and MPLS-TP are identical in the datapath, dealing as they do with manipulating labels and moving packets around based upon the labels. This ensures interconnections and interoperability with respect to the datapath or forwarding plane. So, a provisioned network with MPLS and one with MPLS-TP would work together fairly well. The greater interoperability challenges arise with OAM techniques and the control/signaling plane. MPLS did not standardize the OAM toolset so the question of interoperability with MPLS-TP does not arise. The signaling protocols used by MPLS and MPLS-TP are the same (LDP and RSVPTE) and this greatly improves the control/signaling plane interoperability.
Yes, mapping a client LSP to a server LSP is LSP tunneling.
The question of IPv6 compatibility can be talked about in several perspectives. With respect to the traffic being transported, MPLS-TP can carry any type of packet – Ethernet frames, ATM cells, IPv4 packets, IPv6 packets, etc. So, in this respect, MPLS-TP is compatible with IPv6. The next aspect is the signaling protocols that are used. MPLS-TP uses the same signaling protocols like MPLS – LDP and RSVPTE. Since these have been working with IPv6 for years now, in this respect also, MPSL-TP is compatible with IPv6.
What is the maturity level of Aricent's stack? How many companies/platforms are using the MPLS-TP stack?
Aricent has been in the business of providing networking software for a couple of decades now. Software developed by Aricent has been deployed widely in networks across the world, after being established for interoperability, maturity, robustness, and stability. The MPLS-TP technology itself is new and the software for MPLS-TP is also new. However, the software has been developed as extensions to Aricent’s proven, mature, and industry-leading Ethernet switching software. The MPLS-TP software from Aricent has been successfully integrated into some early customer products and is currently being field tested in operator networks.
Which hardware vendor is running your software?
There are multiple hardware platforms running the Aricent software in applications like mobile backhaul, packet microwave, and optical transport nodes.
What would be the best advantages of migrating to MPLS from traditional links such as frame relay?
The biggest advantages are bandwidth scalability and service scalability. At the same time, it is pertinent to keep in mind that Frame Relay is used at the access where it is still in use while MPLS and MPLS-TP are used in the aggregation and core networks. So, it is not really an ‘apples to apples’ comparison of the technology with respect to where it being put to use.
We are dependent on ISP for the traditional MPLS (IP). If we want to migrate to MPLS-TP what benefit we can have in terms of performance and monetary wise?
You cannot migrate to MPLS-TP without the ISP providing support for the same. If/when your ISP offers support, the benefits from your perspective as a subscriber is more transparency into the service level assurances and actual network utilization than you are getting from the ISP. Historically, the availability of metrics for performance has always been immediately followed by better service performance levels, and so you can expect the same. This directly translates into higher availability of your connections, and therefore lower downtimes which will translate into monetary benefits. Direct cost benefits may also be possible when the ISP has better visibility into network performance which would help the ISP to more efficiently use network capacity.
Doesn't MPLS-TP forwarding use the same techniques as Frame Relay virtual circuits?
Yes, in a fundamental way, this is correct. Frame Relay also uses virtual circuit labels or identifiers (called DLCIs) for forwarding traffic (frames) through switches. However, the way of managing label spaces, the protocols available for OAM and signaling, the extent of hardware support available, are all very different for MPLS, all of which contribute to a much higher bandwidth and speed capabilities for MPLS or MPLS-TP than for Frame Relay.
Are there any service provider/operator network deployments of Aricent's MPLS-TP stack?
Yes, the software is going through field trials in some operator networks.
Are these stacks already pre-ported to any standard hardware/chipset(s)?
The hardware support from merchant silicon vendors is being introduced currently and Aricent is one of the earliest software providers working with hardware/chipset platforms to provide an integrated solution.
What protection mechanism is used to provide sub 50ms failover for ELAN services in MPLS-TP?
The APS (Automatic Protection Switching) defined in IETF RFC 6378 is the standard for achieving this requirement.