[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Pce] Virtual Network Topology (VNT) changes and IP layer routingloops

Hi Kohei, For the folks who haven't read the "micro-loop" drafts. These describe transient loops that can occur at the IP layer or in MPLS with LDP. In the GMPLS world we may also use the term LSP somewhat loosely to apply to SDH, WDM, G.709 connections. Since these are all explicitly routed (and due to their circuit switched nature too) we won't get loops at these layers.
However, when we use GMPLS to set up a new SDH, WDM, G.709, etc... connection (or tear down a connection) between IP routers (modifying the IP layer VNT) this looks like a link up or down "maintenance" action to the IP layer and hence the issue with micro-loops so well described in the drafts. 

Regards

Greg B.

Kohei Shiomoto wrote:

Hi Greg

Thank you for your kind notice.

I think that the avoiding micro-loop is an important work. Micro-loop
could unnecessarily amplify traffic load as much as 128 times, which
should be a serious problem for today's wire-speed packet forwarding at
high-speed link.

So far micro-loop avoidance has been discussed in the context of
maintenance operations including link install/removal, IGP link cost
change, etc. As you pointed out, micro-loop may form in the process of
VNT reconfiguration if IP datagram is forwarded directly on top of VNT
(if we use explicit-routed LSP, we could avoid micro-loop). Micro-loop
avoidance should also be considered in the context of VNT reconfiguration.

Thank you again for your pointer.
--
Kohei


Greg Bernstein wrote:
Hi Don, I'm using the term "Virtual Network Topology" related to the data plane not the control plane. For example, when I interconnect IP routers via a WDM or SDH switching layer the resulting IP layer topology is termed a "virtual network topology" since the connectivity amongst the IP routers is not necessarily the same as the fiber topology. It is these changes to the IP layer topology that can cause the traffic disrupting micro loops. Note that other ways that micro loops occur are: (1) during IGP link weight adjustments and (2) maintenance operations.
The connection to GMPLS/PCE is we are trying to make setting up these virtual topologies quicker and potential for traffic engineering purposes. There have been some good papers in the literature on using GMPLS and advanced algorithms for the VNT design problem. 

Greg B.

Don Fedyk wrote:


Hi Greg

Whoa....
  [Snip]
1) These virtual topologies are for control traffic and they are solid
most of the time.
2) The existing data traffic is not disrupted when the control traffic
has microloops. 3) There may be a delay in signaling of new or rerouted connections when 
the control traffic has microloops. This is analogous to Graceful
recovery of the control plane where the data plane is momentarily unable
to adapt to new changes.
4) Many of the optical technologies we have do not need response to
signaling in less than 100s of milliseconds so the delay is not
critical.
5) We need a clear separation of protection which can be locally driven
and fast (50 msec) and routing and rerouting which are slower
restoration mechanisms. As long as protection mechanisms are independent
of the control plane primary protection is safe IMHO.
At the Routing working group, Alex's draft draft-ietf-rtgwg-microloop-analysis-01.txt provides some analysis and a method to reduce the impact (duration too) of the transient loops. More recently the drafts: 
(1) draft-bryant-shand-lf-applicability-01.txt
(2) draft-bryant-shand-lf-conv-frmwk-02.txt
(3) draft-francois-ordered-fib-01.txt
Address this problem more generally including in (3) a method that guarantee loop free convergence. 


Loop Free Alternates are a good thing.  But with more advanced
mechanisms it is very hard to determine if the cure is worse than the
symptoms.
The problem is that these have generated very little interest at the RTG WG and may not move forward. This area is not within PCE or CCAMPs charter but can have an impact on the adoption of GMPLS in multi-layer/region networking. If you're interested please take a look and comment to the RTG WG. Note I wasn't involved with writing these, but came across them when considering the effects of GMPLS changes on the IP layer VNT. 
There was a lot of interest and a genuine amount of push back.
Personally I would be very concerned about tight coupling of a control
plane convergence to impacts on a GMPLS control data plane.
Regards,
Don


Thanks

Greg B.

--
===================================================
Dr Greg Bernstein, Grotto Networking (510) 573-2237







--
===================================================
Dr Greg Bernstein, Grotto Networking (510) 573-2237