Segment Routing Testing Discussion

[MuhammadKhidr0164]

Dear All , 

I am doing my thesis on Segment Routing and trying to show the benefits of SR by using GNS3 and building a network and implementing SR and testing some use cases 

I'd need your help and suggestions the network is as shown below 

the IGP is OSPF and then SR is applied 

- my first question is : How to show that ECMP is working 
- my second question is regarding a scenario i did and wish you can help with it 

I am running IPerf3 test between PE-01 and PE-02 in different cases while congesting the link between P-01 and P-03 ( by another test on the same time ) using UDP packets 
- first without SR ( OSPF only ) 
- 2nd with segment routing with a policy to send the traffic trough P-04 then PE-02 while the other routes will be decided automatically by IGP so i have 2 segments only.
- I fixed the traffic from PE-01 to PE-02 to be 512 K and changes the other traffic ( between P-01 and P-03 ) from 512K , 1 Mb , 2 Mb , 4 Mb and so on 

what I found is that with OSPF only with increasing the BW gradually i found losses  and losses increases with BW increase ( I mean the changeable BW ) however I found that was solved when using SR 

is that correct ? Can SR solve such problem ? or I have a problem in my use case ?

all your comments and suggestions for another use cases that shows SR benefits are welcome 

Thanks in advance 

[Giuseppe Larosa]

Hello Muhammad,

about your questions:

1) to show that OSPF with SR is able to support ECMP, I think it depends on how you have configured the SR paths.

2) Possible benefits with traffic over the emulated environment

As we have discussed in a previous thread you have an emulated environment in GNS3 or VIRL.

Data plane aspects in emulation provide relative results.

With Segment Routing like it happens with more traditional MPLS TE you can move away from IGP best path choices and you can move some traffic over other alternate paths and avoid potential congestion.

I think this is the meaning of your tests with traffic generated by Iperf with a concurrent traffic that creates congestion when traffic level increases on the background.

Hope to help

Giuseppe

[MuhammadKhidr0164]

Hi Giuseppe, 

Many Thanks for your continuous support 

What i saw in my test is that however i didn't force the traffic to another direction and let it go to P_04 using IGP the performance got better when i used SR 
Does that show a benefit os SR ? ( Results and justification is important for the thesis :) ) 

another point that in my current topology is there any suggestion for any scenario that shows SR benefits ? 

Thanks again and much appreciate your support