What happened to routing?

  • Thread starter Thread starter smop
  • Start date Start date
  • Replies Replies 35
  • Views Views 3,112
They have been using Sify for a while now, not only for Hurricane Electric IP addresses but also for a few other destination IP addresses. However, they still primarily rely on Airtel as their upstream provider at MAA, BOM, and DEL. Delhi and Chennai appear to be less affected by the Airtel routing issue due to their greater utilization of the Cloudflare backbone compared to Cloudflare Mumbai, which, for some reason, continues to rely more on Airtel rather than the Cloudflare backbone. On the Cloudflare backbone, latency is normal or lower.

Code:
HOST: xxxxx-xxxxxxx-xxx.local                                 Loss%   Snt   Last   Avg  Best  Wrst StDev
  1. AS13335  172.70.217.66                                    0.0%    10    6.6   8.4   6.6  12.7   2.3
  2. AS13335  104.28.0.0                                       0.0%    10  177.8  26.4   6.9 177.8  53.2
  3. AS13335  8.34.69.1                                        0.0%    10   48.0  18.3  10.9  48.0  12.3
  4. AS13335  162.158.226.88                                   0.0%    10    8.2  12.7   8.2  31.3   7.1
  5. AS9583   1.7.160.209                                      0.0%    10   90.6  19.9   9.0  90.6  24.9
  6. AS???    ???                                             100.0    10    0.0   0.0   0.0   0.0   0.0
  7. AS???    ???                                             100.0    10    0.0   0.0   0.0   0.0   0.0
  8. AS???    ???                                             100.0    10    0.0   0.0   0.0   0.0   0.0
  9. AS???    ???                                             100.0    10    0.0   0.0   0.0   0.0   0.0
 10. AS???    ???                                             100.0    10    0.0   0.0   0.0   0.0   0.0
 11. AS???    6939.sgw.equinix.com (27.111.228.81)            90.0%    10   84.4  84.4  84.4  84.4   0.0
 12. AS???    hurricane-electric.sgix.sg (103.16.102.81)      90.0%    10   84.6  84.6  84.6  84.6   0.0
 13. AS???    ???                                             100.0    10    0.0   0.0   0.0   0.0   0.0
 14. AS???    ???                                             100.0    10    0.0   0.0   0.0   0.0   0.0
 15. AS???    hurricane.nwax.net (198.32.195.42)              70.0%    10  247.2 240.1 234.4 247.2   6.5
 16. AS???    eqix-sv8.hurricaneelectric.com (198.32.176.20)  90.0%    10  246.9 246.9 246.9 246.9   0.0
 17. AS6939   100ge14-1.core3.fmt1.he.net (72.52.92.66)        0.0%    10  247.2 259.0 242.3 356.7  35.6
 18. AS6939   he.net (216.218.236.2)                           0.0%    10  237.9 245.1 236.6 301.0  19.8
 
Before 2-4 months they were using AS9498 Airtel as upstream.
Still they use Kerala Vision (AS138754~AS9498) as upstream in COK colocation
 
The routing to the Hurricane Electric endpoints using Sify is quite poor. All Hurricane Electric endpoints, regardless of their location, are consistently routed via Singapore to reach the destination on Sify upstream through Warp. This adds an overhead of 60 ms.
 
@Anurag Bhatia
This may be completely Dumb or Idiotic doubt

Why Indian ISP'S Like (ACT , Excitel , Other Regional Tier 2 ) have Indian upstreams like BSNL(AS9829), Airtel(AS9498), Sify(AS9583) etc.

Why they can't peer or buy transit from a Tier-1 ISP. Is there any Indian Govt Policy ? Or is it expensive to buy the transit

Why this Tier 2 Regional ISP doesn’t depend on IX for peering and most of the peering and transit goes the upstreams
 
Last edited:
Hi

Why Indian ISP'S Like (ACT , Excitel , Other Regional Tier 2 ) have Indian upstreams like BSNL(AS9829), Airtel(AS9498), Sify(AS9583) etc.

Basically networks like ACT, Excitel as you mentioned in your example offload 80-85% of their traffic on PNIs (Private peerings) + IX peering (incase of Excitel), local CDN caches of Google/Facebook/Akamai etc. For this 80-85% they may or may not depend on large telcos depending on area. E.g most of content is present in Delhi/Mumbai/Chennai. Those + say another 50-100km range of those often has dark fibre by various players and ISPs can make use them to connect their nodes with leased dark fibre.

For remaining 15-20% traffic they rely on upstream transit. This overall volume is low to justify going outside of India. Plus an ISP cannot build just to Singapore as that would mess up routing with EU (like the case of BSNL). So one would need atleast Singapore + somewhere in Europe, mostly on long term commitments, plus cost of transport from landing station to a colocation (like Equinix, Coresite, Telehouse), plus cost of colo, plus cost of cross connects etc. All this cost adds up and is not worth it for 15-20% traffic unless that's very high in absolute number. Airtel and Jio do that even when for them is barely 6-7% traffic but absolute volume is still quite high (in Terabits). Bandwidth, compute (VMs), storage etc is all turning into commodity market and is hard to operate on your own unless doing it at scale and have your own captive user base for it (Airtel/Jio) or do transit business to have downstreams for it (again Airtel, Sify, Tata Comm etc).

Why this Tier 2 Regional ISP doesn’t depend on IX for peering and most of the peering and transit goes the upstreams
Unsure what makes you say that. You mentioned ACT and Excitel in your example. Excitel does peers at Extreme IX (and Extreme is sister company for them anyway). ACT doesn't peers at IX but that's mostly because they have PNIs and local caches of almost all players they can get. Once you all those i.e direct PNI to Google/FB/Akamai/Netflix besides their local caching nodes, the traffic by % to IX is very low. So it may not justify the effort, cost of maintaining it etc.
 
Back