Composite Link Framework in MPLS (Curtis Villamizer)
- This is working group last call, please comment on the framework
- Discussion: none raised
MRT Algorithms Analysis Progress (Alia Atlas)
- [Pierre Francois] Do you have interactive forwarding?
- [Alia Atlas]: I do not have interactive forwarding.
- [Pierre Francois]: It makes a lot of difference.
- [Alia Atlas]: I'd like to talk about the details offline.
Operational management of Loop Free Alternates (Stephane Litowski)
- Discussions: none
- Asked for WG Adoption.
- Alvaro: will take to list.
Interactions between LFA and RSVP-TE (Stephane Litowski)
- Alvaro Retana(chair) Comments?
- xx (Huawei): To use the LFA and FRR in the same network is complex.
- Stephane: We use the LFA in the engine, why not use it with FRR.
- xxx (Huawei): How does this with work tunnels?
- Stephane Litowski: Using it across multi-vendors is not the best solution, but we need to document it.
- Rob Shakir (BT): This draft does a good job of describing the interaction between the two functions.
Microloop prevention by introducing a local convergence delay (Pierre Francois)
- Alia Atlas: I agree that in terms of simplicity, how does this compare to PLSN.
- Pierre Francois: Similar, I can share the results with you.
- Alia Atlas: I think this with be useful to compare to the PLSN.
- Alvaro Retana: How many people have read the draft?
- [25+ hands]
- Alvaro Retana: Please discuss this on the list.
Advertising MPLS labels in IGPs (Hannes Gredler)
- Alvaro Retana: How many people have read this draft?
- Alia Atlas: It is a useful change.
- Andy Malis: Will discussion happen here or on the mpls list?
- Alia Atlas: Our list is not so busy that we cannot have an "cc"
- Alvaro Retana: It would be good to have more comments from the room.
- Alia Atlas: Otherwise, we have a small group of people discussion (5+)
- Alvaro Retana: We would appreciate if when we take documents to the list, we have discussion.
[Introduction to PANET section]
- Curtis: I was thinking we would have a big open discuss to talk about taking up this work. There is no correlation between the problem statement, the use case, and the requirements.
- Alvaro Retana: We'll have a discussion at the end of the session.
PANET Problem Statement + Requirements (Beichuan Zhang)
- Curtis: The first dilemma I have is do I comment on the drafts or the presentation? If you look at the problem statement draft it has a list of research areas. (Curtis then listed a set of research areas from the draft). This is all that is in the problem requirements. The requirements make an assumption about the result of the research. We need to take a step back and understand this is the research area. In the slides, it is suggested we focus on an area such as DC or edge providers. Originally, we started out with the research of a core area.
- Curtis: (continuing) I think we should treat this as a research area. This is not a question but a comment. How do the chairs feel about this fact?
- Alia: We are an incubator that for work that might make sense to in the IETF someday. We had a discussion with Lars about the research. We are investigating whether this work has enough substantive to do in the IETF, or is substantive work.
- Beichan: I would like to hear your comments offline.
- Curtis: I'm sure this will be an interesting discussion.
- [Eric Osborne]: I know we don't do business cases. The back of the envelope economics say we only solve a nickel a day, we should not go further. Network engineering is all about trade off financially and technically.
- [y-person]: This is a tradeoff. For example, vm-migration is about some tradeoff. The scope in Data Center is about switches versus the servers. The servers are larger.
- Don Fedyk: Where does this come up from your customers? Customer do talk about the power-foot print. Is there something simple we can do that will get us a long way toward what you are doing. My comment is that this is a research area. Perhaps a simple approach will allow you to work on this.
- Alvaro: We should continue on with the presentation.
Use Cases for Power-Aware Networks (Mingui Zhang)
- Curtis: The use cases draft is not correlated to the other one. We have core use cases and DC use cases. We should be able to move traffic off component links. There is no evidence that the reduction in the power will reduce the whole piece. There are two types of power in the router - forwarding blades and fans. At least 30% of the power is in the fans. If you can turn chips off on a card, then you get a cool blade. If you have 3 cool blades and 1 hot blade, you still have to run the fans to cool the hot blade. The air flow through the chassis may still be the same.
- [Correction to the notes made by Curtis: The comment I made regarding air flow is that if you concentrate traffic onto one card that card runs hotter and the fan speed is determined by the amount of airflow needed to cool the hottest card. So air flow has to increase, not stay the same and therefore fan power will increase.]
- Rob Shakir (BT): I did't read the draft. This is an area that telcos care about. The issue of unused links is that there is also capital investment into such unused links, not just power usage. .... If I look at my networks, I will be reducing power by reducing parallel infrastructure. I think that if I was reducing power, I would do this first. If the re-engineering costs of the network are more than the power savings.
- Stewart: (speaking as individual contributor) I want to address some of the things said. You want to understand the problem, and understand what you might gain. If you look at the phone, you run different hardware based on power consumption. This is a long-haul research problem. We need to carefully determine the individual pieces of the problem statement.
- Pat Thaler (IEEE 802 chair, Broadcom): I want to mention the 802.3az (energy efficient ethernet). This work tried to reduce the power by shutting it down quickly and waking up so no traffic would be missed. This is in the ms worth of time. If you wish longer time, this could negotiate this within the 802.1az specification. We have done better with the components that are worked in home. There is an additional work to be done in the fiber links. The shutdown and bringing back up cannot be done quick enough to share power. This effort will take some active signaling.
- Alia: This is an excellent discussion. However, we would like to have a chance to have the last two presenter talk.
A Framework and Requirements for Energy Aware Control Planes (Manuel Paul)
- Alvaro: In interest of getting the last presenter sometime, we will cut conversation here.
Reducing Power Consumption using a Metric Based Approach (Shankar Raman)
- Alvaro: This is Shankar's first IETF. Please provide him with "first-time" courtesy.
- Alia: We have 5 minutes for discussion.
- Don Fedyk: Why did you pick the telecommunication within India? I did a quick scan of the web.
- Shankar: The people who wish this work are an ISPs.
- Don Fedyk: ISP are not the heaviest consumer of the power.
- Shankar: The manufacturing gets power. The ISPs does not get power.
- Don Fedyk: We are squeezing power from an area that does not take enough power. The house appliances (refrigerators) take more power.
- [Eric Osborne]: You have the information to indicate whether this is appropriate information. Please write up this information so we can look at it.
- Curtis: Thank you I have a better insight of what you are facing. The power is different based on whether you are using solar or diesel power. You can use TE so that you can avoid the diesel power links. You can avoid the diesel links by using bump of the information.
- Shankar: Thank you for the input. We did do some of these experiments.
- Dimitri: You are showing a black and white box approach. What is your viewpoint of looking at the white box versus black boxes? Lots of comments that were raised that were looking a specific metrics that were configured. The question is whether you want to discover instead of computing.
- Shankar: We are looking at this from the top level.
- Dimitri: You need to decide if it black box or white.
- Alvaro: We need to narrow down. Thank you for the input.