[e2e] a question on edge-to-edge DiffServ flows

[Jaeyoung Kim]

Hi, I have an idea on monitoring edge-to-edge DiffServ flows and would
like to have some comments or related papers on this. 

When managing a DiffServ domain, it is important to have the network-wide
traffic status of each QoS class in DiffServ network. Since each DiffServ
router can monitor QoS (throughput, drop rate, etc.) of each DiffServ
class it handles, I think it is possible to extract edge-to-edge traffic
status by combining the QoS information monitored in each router and 
routing path from one edge router to another.

In the following example network topology with four edge routers (E1-E4)
and two core routers (C1-C2),

 E1---\           /---E3
       \         /
       C1-------C2
       /         \
 E2---/           \---E4

Let's assume there are four different edge-to-edge traffic of a DiffServ
class, E1->E3, E1->E4, E2->E3, E2->E4. Thus, E1 and E2 perform source
edge routers, and E3 and E4 perform sink edge routers.

In this situation, I'd like to have traffic status of one Edge-to-Edge
DiffServ flow. The problem here is that all the traffic of the same
DiffServ class are aggregated and undistinguished in DiffServ domain.
That is, the QoS information monitored at each router is not for each
edge-to-edge flow, but for aggregated class overall. However, with simple
rule, I think we can extract QoS of each E-to-E traffic.

Let's assume that we are monitoring the throughput of a DiffServ class
in each link within a certain time interval. The throughput is indicated
in the following figure.

 E1---\3         4/---E3
       \         /
       C1-------C2
       /   10    \
 E2---/7         6\---E4

The throughput monitored in E1->C1 link is 3, E2->C1 is 7. And the two
traffic is aggregated in C1->C2 link and the amount is 10. Finally, in
C2 router, the traffic is splitted to 4 for C2->E3, and 6 for C2->E4.