Re: [Pce] Concurrent Optimization and More

["Daniel King" <[e-mail]>]

Hi David, 

 

Aha ,"real-time". Well as we know real-time is relative. We talked
previously about the PCE GCO application. Performing a GCO, via an NMS on a
centralized PCE, on a complex network might require 6hrs of computation
time. This could be considered real-time and perhaps is an acceptable amount
of time to wait for a near-optimal solution. Another scenario might be
performing path computation for a local backup path after an existing backup
path fails. A local backup path solution could be required within a few
seconds and computed via a distributed PCE that has local network knowledge.
Both scenarios would require a real-time view (topology, services and
resources) of the network at the time of the path computation request.  The
mechanisms are in place to support these architectures and I am aware of
vendors implementing some of these ideas.

 

We should not mention specific vendors on the WG mailing list. I am
compiling a list of PCE vendors and implementers at:
www.pathcomputationelement.com. Any vendors not currently listed are welcome
to contact me and add their name and information.

 

Br, Dan. 

 

 

From: [e-mail] [mailto:[e-mail]] On Behalf Of
[e-mail]
Sent: 20 June 2008 08:26
To: [e-mail]
Subject: Re: [Pce] Concurrent Optimization and More

 

Dan,

 

Many thanks for your fast reply. Are you familiar with any "real-time"
implementation of few of the PCE ideas? Anyone that implements the
distributed approach?

 

Warm regards,

 

David Amzallag,

BT

 

  _____  

From: Daniel King [mailto:[e-mail]] 
Sent: 19 June 2008 12:50
To: [e-mail]
Subject: RE: [Pce] Concurrent Optimization and More

Hi David, 

 

A few quick responses:

 

>>1. How does the protocol for computing a single path works in a
hierarchical network (more than one level of PCE's)?

 

So far this requirement has been out of scope for the current work. There is
growing momentum in the application of PCE to multi-layer networks, ASON,
and "domain-paths." So this topic will require further thought and
discussion.

 

>>2. What the draft for the concurrent optimization referred to: only inside
of a sub-domain or also an inter-domain optimization which requires
communication between the PCE's?

 

As above. The pce-global-concurrent-optimization draft focuses on single
domain optimisation, specifically the PCC-PCE communication needs and
protocol extensions to support the concurrent optimisation. Although, the
authors do mention that a Global Concurrent Optimization (GCO) is applicable
to a Virtual Network Topology (VNT) and multi-layer traffic engineering for
new deployments. 

 

>>3. Can we assume that the node's demands are splittable (i.e., can be
simultaneously satisfied by more than one path)?

 

I may need some further clarity on this question. Do you mean that there may
be more than one possible solution to a path computation, or that the
solution to the computation may be a set of "parallel" paths with some
assumption of load sharing (e.g. inverse multiplexing)?

 

The answer to the first question is yes and is already done, although some
work might be required to allow the PCE to supply a choice of paths to the
PCC.

 

The answer to the second question is also yes, but (very simple) protocol
extensions will be needed to handle this case.

 

Other relevant drafts you may find interesting for this topic include:

 

A Path Computation Element (PCE)-Based Architecture 

http://www.ietf.org/rfc/rfc4655.txt

 

A Framework for Inter-Domain Multiprotocol Label Switching Traffic
Engineering

http://www.ietf.org/rfc/rfc4726.txt

 

A Per-Domain Path Computation Method for Establishing Inter-Domain Traffic
Engineering (TE) Label Switched Paths (LSPs)

http://www.ietf.org/rfc/rfc5152.txt

 

Evaluation of existing GMPLS Protocols against Multi Layer and Multi Region
Networks (MLN/MRN)

http://www.ietf.org/internet-drafts/draft-ietf-ccamp-gmpls-mln-eval-05.txt

 

A Backward Recursive PCE-based Computation (BRPC) Procedure To Compute
Shortest Constrained Inter-domain Traffic Engineering Label Switched Paths

http://www.ietf.org/internet-drafts/draft-ietf-pce-brpc-09.txt

 

Br, Dan

 

 

From: [e-mail] [mailto:[e-mail]] On Behalf Of
[e-mail]
Sent: 19 June 2008 07:25
To: [e-mail]
Subject: [Pce] Concurrent Optimization and More

 

Hi,

 

I have three basic questions (but I might the correct place in RFCs or
drafts for that...);

 

1. How does the protocol for computing a single path works in a hierarchical
network (more than one level of PCE's)?
2. What the draft for the concurrent optimization referred to: only inside
of a sub-domain or also an inter-domain optimization which requires
communication between the PCE's?

3. Can we assume that the node's demands are splittable (i.e., can be
simultaneously satisfied by more than one path)?

 

Many thanks,

 

David Amzallag

BT

 

_______________________________________________
Pce mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/pce