Two key differences. (1) you want to do this purely in data plane fast path. (2) you don’t want to keep a streaming session from each node for scalability reason The path probing approach can meet both requirements.
Haoyu From: Tianran Zhou <zhoutian...@huawei.com> Sent: Thursday, February 27, 2020 5:14 PM To: Haoyu Song <haoyu.s...@futurewei.com>; Robert Raszuk <rob...@raszuk.net> Cc: ippm-cha...@ietf.org; spring@ietf.org; i...@ietf.org Subject: RE: [spring] Monitoring metric to detect and locate congestion This is a good point. In this way, the probe is generically used to collect node data. At each node, the probe will be send to the slow path to get the node data. Data could be carried in the payload which has more space. Then, what’s the difference to existing way that Robert mentioned? Use streaming telemetry to export from each node. As far as I know the state of art, many device support streaming telemetry can export in a period of 1-10 seconds. For customized data, the period could even be reduced to 10-100ms. Best, Tianran From: ippm [mailto:ippm-boun...@ietf.org] On Behalf Of Haoyu Song Sent: Friday, February 28, 2020 4:44 AM To: Robert Raszuk <rob...@raszuk.net<mailto:rob...@raszuk.net>> Cc: ippm-cha...@ietf.org<mailto:ippm-cha...@ietf.org>; spring@ietf.org<mailto:spring@ietf.org>; i...@ietf.org<mailto:i...@ietf.org> Subject: Re: [ippm] [spring] Monitoring metric to detect and locate congestion Can the combination ingress+egress queuing information on each transit node be collected by simply visit the node? If so, one or more probing paths that can cover all the nodes are sufficient. Best, Haoyu From: Robert Raszuk <rob...@raszuk.net<mailto:rob...@raszuk.net>> Sent: Thursday, February 27, 2020 12:18 PM To: Haoyu Song <haoyu.s...@futurewei.com<mailto:haoyu.s...@futurewei.com>> Cc: ruediger.g...@telekom.de<mailto:ruediger.g...@telekom.de>; ippm-cha...@ietf.org<mailto:ippm-cha...@ietf.org>; spring@ietf.org<mailto:spring@ietf.org>; i...@ietf.org<mailto:i...@ietf.org> Subject: Re: [spring] Monitoring metric to detect and locate congestion The point is not to assure all egress ports are inspected. The point of any really useful end to end path probing is to find out if combination of ingress+egress queuing on each transit node my packet may traverse are not congested. Looking at Eulerian path algorithm I do not see such guarantees. Best, R. PS. Path probing is A1 to B1. It is not A1 to B1, B2 .. Bn. where An are the ingress ports to the network and Bn are the egress ports. On Thu, Feb 27, 2020 at 9:14 PM Haoyu Song <haoyu.s...@futurewei.com<mailto:haoyu.s...@futurewei.com>> wrote: Hi Robert, The Eulerian path algorithm guarantees to visit all the edges of a graph. In the SR context, we can consider the sub-path between two segments an edge. Haoyu From: Robert Raszuk <rob...@raszuk.net<mailto:rob...@raszuk.net>> Sent: Thursday, February 27, 2020 11:50 AM To: Haoyu Song <haoyu.s...@futurewei.com<mailto:haoyu.s...@futurewei.com>> Cc: ruediger.g...@telekom.de<mailto:ruediger.g...@telekom.de>; ippm-cha...@ietf.org<mailto:ippm-cha...@ietf.org>; spring@ietf.org<mailto:spring@ietf.org>; i...@ietf.org<mailto:i...@ietf.org> Subject: Re: [spring] Monitoring metric to detect and locate congestion Hi Haoyu, > which applies Eulerian Path algorithm to find the minimum set of paths with > network-wide coverage. In practice networks use ECMP. ECMP decision may happen at each hop. Your end to end flows get spread over all ECMP paths. So limiting number of probed paths is inaccurate to the fundamental objective of the exercise. That is infact main challenge with any end to end path probing today ... if you do not cover all possible paths your packets may take between ingress and egress you just do not get full picture of the network. Thx a lot, R. On Thu, Feb 27, 2020 at 8:44 PM Haoyu Song <haoyu.s...@futurewei.com<mailto:haoyu.s...@futurewei.com>> wrote: Hi Ruediger, I like the general idea that using pre-determined paths in SR to collect performance metrics. I think this approach provides some unique benefits compared with the other approaches. It is also coincident with some of related research work I’m doing. Here are some thoughts I have. 1. I think IOAM could be used as the standard approach for such probing packets. It can collect the performance metrics mentioned in the draft and does more. 2. An interesting problem raised by the draft but not fully addressed is the method to plan the optimal paths. There is a work called INT-PATH (https://ieeexplore.ieee.org/document/8737529<https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fieeexplore.ieee.org%2Fdocument%2F8737529&data=02%7C01%7Chaoyu.song%40futurewei.com%7C03811f6f80554f7dad6408d7bbeb74f0%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637184492271070940&sdata=CMFuDR26gKPUjOYu3vcVlWNoTP%2FA%2FGiCIEnxERVBZmk%3D&reserved=0>) which applies Eulerian Path algorithm to find the minimum set of paths with network-wide coverage. However, the problem here seems different: you need path coverage redundancy. My question is: do we really need the redundancy to achieve the measurement goal? If so, what’s the best planning algorithm should be? In a real and large scale network, we have constraint on where the probing device(s) can be placed, and we usually want to monitoring the entire network, so an efficient algorithm is necessary. Best regards, Haoyu From: ippm <ippm-boun...@ietf.org<mailto:ippm-boun...@ietf.org>> On Behalf Of ruediger.g...@telekom.de<mailto:ruediger.g...@telekom.de> Sent: Tuesday, February 25, 2020 11:55 PM To: ippm-cha...@ietf.org<mailto:ippm-cha...@ietf.org> Cc: spring@ietf.org<mailto:spring@ietf.org>; i...@ietf.org<mailto:i...@ietf.org> Subject: [ippm] Monitoring metric to detect and locate congestion Dear IPPM (and SPRING) participants, I’m solliciting interest in a new network monitoring metric which allows to detect and locate congested interfaces. Important properties are * Same scalability as ICMP ping in the sense one measurement relation required per monitored connection * Adds detection and location of congested interfaces as compared to ICMP ping (otherwise measured metrics are compatible with ICMP ping) * Requires Segment Routing (which means, measurement on forwarding layer, no other interaction with passed routers – in opposite to ICMP ping) * Active measurement (may be deployed using a single sender&receiver or separate sender and receiver, Segment Routing allows for both options) I’d be happy to present the draft in Vancouver... If there’s community interest. Please read and comment. You’ll find slides at https://datatracker.ietf.org/meeting/105/materials/slides-105-ippm-14-draft-geib-ippm-connectivity-monitoring-00<https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fmeeting%2F105%2Fmaterials%2Fslides-105-ippm-14-draft-geib-ippm-connectivity-monitoring-00&data=02%7C01%7Chaoyu.song%40futurewei.com%7C03811f6f80554f7dad6408d7bbeb74f0%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637184492271070940&sdata=tL%2FtbMJ0Ei2e%2FUFZyBB2GFPEhj8qzCuWXo5upM9GNsE%3D&reserved=0> Draft url: https://datatracker.ietf.org/doc/draft-geib-ippm-connectivity-monitoring/<https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fdraft-geib-ippm-connectivity-monitoring%2F&data=02%7C01%7Chaoyu.song%40futurewei.com%7C03811f6f80554f7dad6408d7bbeb74f0%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637184492271080924&sdata=qteR3ORsHz6mnbguKfMMQwcls9kaSi8NhAh9hFaH7b8%3D&reserved=0> Regards, Ruediger _______________________________________________ spring mailing list spring@ietf.org<mailto:spring@ietf.org> https://www.ietf.org/mailman/listinfo/spring<https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fspring&data=02%7C01%7Chaoyu.song%40futurewei.com%7C03811f6f80554f7dad6408d7bbeb74f0%7C0fee8ff2a3b240189c753a1d5591fedc%7C1%7C0%7C637184492271080924&sdata=cUScWh29Li%2BFBJHIYBm6fcylGx6zrv%2F4RtEM6b5YMvs%3D&reserved=0>
_______________________________________________ spring mailing list spring@ietf.org https://www.ietf.org/mailman/listinfo/spring