Slow speed on international and different city servers

  • Thread starter Thread starter vIjAy_kHaSa
  • Start date Start date
  • Replies Replies 22
  • Views Views 3,985

vIjAy_kHaSa

Regulars
Messages
323
Location
Panchkula
ISP
Airtel Xstream Fiber
Guys I am facing slow speed issue on servers located outside my city on speedtest.net. I get above 100 mbps if I test on servers of Chandigarh Tricity and around 30 - 40 mbps on servers like Noida, Bangalore, Mumbai of airtel. In case of international servers like canada, singapore, USAspeed further drops to below 20 mbps most of time.

Do you guys also suffer from this low speeds or I am an exception
 
Unsure of why you would have issue from Chandigarh to Noida but regarding International traffic note that EIG Cable (London-Mumbai) is down since last two days. They might be having capacity crunch on that segment via other cables (just a guess). I do see 2-3% packet loss during peak times from last two days on India-Europe segment on my connection (which is a downstream of Airtel).

My friend is tweeting regular updates on EIG cable on the thread here:
Source
 
And i thought that why my connection was having random packet loss issues from last 2 days. So it’s due to this cable :(
 
I am not sure if this is related, but I am on the TATA backbone. (103.2.132.0/24). Speed to my ISP and most other ISPs within West Bengal is fine. But any traffic leaving West Bengal, particularly downloads, are very slow. Traffic to Jio Kolkata from my IP is slow since it gets routed via Mumbai.

I am on a 50 Mbps plan, and download ranges from 6 - 25 Mbps randomly to Airtel, Excitel, even to alliance Hyderabad. Latency to Alliance Kolkata was ~6 ms last year, this year, it was 12 ms, and now it's ~18 ms. I raised a ticket last 22nd, and they said it's location specific and that they are working on it. God knows what these ISPs are up to!

Last year, the download speed from all google services was limited to 5 Mbps (both on TATA and airtel backbones - alliance uses both); I was told it was a peering issue. They didn't care to shift it to raw bandwidth. It took them 20- 30 days to fix it.
 
Last edited:
I am not sure if this is related, but I am on the TATA backbone. (103.2.132.0/24).
Your ISP is Alliance and thus I would suggest you as well as others here to avoid using terms like "I am on Tata backbone" or "I am on Airtel backbone" unless you are a direct customer of Tata Comm or Airtel etc. Alliance is an independent network, has it's own autonomous system number and has it's own interconnection policy. For IP transit they have a mix of Tata Comm, Tata Tele, Airtel and Vodafone besides extensive peering they have in exchanges in Mumbai. Also as a matter of fact your IP is covered in three BGP announcements - https://bgp.he.net/ip/103.2.132.0 (/24, /23 and /22). The least specific /22 is visible behind Airtel as well. Thus remember that many of these networks are extensively peered and typically use IP transit for much less of the overall capacity.

Last year, the download speed from all google services was limited to 5 Mbps (both on TATA and airtel backbones - alliance uses both
Again that's not correct understanding. Alliance has peering with Google besides GGC caching nodes. Thus everything cached is local within West Bengal and non-cached traffic is served via their peering with Google.

They didn't care to shift it to raw bandwidth.
What do you even mean by that?


Anyways likely the issues you are getting might not even be related to peering. Probably just congestion.
 
avoid using terms

Alliance usually provides 3 IP blocks at my location. On one, the traceroutes go through airtel; on the other two, traceroutes go through TATA. I doubt Alliance even cares to block websites at their own level. Website blocks vary across the 3 IP ranges; I guess that is done at TATA/ Airtel end. These are why I mentioned 'backbone'. From your posts, I see that it's wrong to do so. Should we call these transit providers?

Also, why did Alliance announce my IP thrice? I saw some materials on a global youtube outage in 2008 and how they fixed it using more specific announcements. If /24 's are already announced, why do they need to announce it again in /22, or am I missing something here? They are doing the same for IPv6 (though they do not provide IPv6 yet to end users):


What do you even mean by that?
Google services (except drive) were limited to 5Mbps for each subscriber, and the call centre told me it was due to some peering issue. I don't know if that was true. If a caching node/ direct peering was broken, why didn't they redirect the traffic to one IXP/ transit instead? Wouldn't that solve the issue? Using a VPN fixed it, as expected. Some of my friends in another town also faced this issue, but some did not, even if they were from the same town. My neighbour was not facing this issue(!); we are from the same LCO.

Can latency jump from 6 to 12 to 18 ms due to congestion? DDoS attacks to a machine increase latency. Is congestion enough to increase latency? Also, my speed test to Alliance Kolkata was fine, and maybe other ISPs peer with Alliance at Kolkata, so they were fine too. If there is no congestion up to Kolkata, why is there congestion to Delhi/ Mumbai and that too, specifically for my town? (When the issue first started, the speed to alliance Kolkata was bad, but later it got fixed.) I am not saying that there is no congestion; I'm just curious.

PS: I am a networking noob. TIA.
 
Back