BSNL Broadband routing problem seems to have been resolved

Yes , i was about to post this. routing is back to normal now.
 
However it is still not stable as pings can jump from the normal 70-80ms upto 290-300ms. Resetting the router seems to fix this issue but I have to do it again and again. A fix without guarantee is no fix at all. BSNL is still screwing up.
 
I am getting 130ms to Google right now. Last night it was around 80ms. But it is quite stable for me during the afternoon time compared to few days ago.
 
Yeah, between peak traffic and minimum traffic times latency differs by 50 ms. Well, I guess we have to live with it. I had to reset my router 11 times to get back to normal just now. Online gaming suffers from packet loss of upto 10%.
 
At least the internets is unusable for me this time. Usually there is a packet loss of around 10-15% and pings are in 400-500 ms. It was quite horrendous between noon and 5pm.
 


I have discovered another problem now. BSNL is remotely disconnecting me whenever I have low pings. They are doing this every 2-10 minutes.
I can't play a single game without being disconnected mid session. BSNL is always there to fuck up.
 
Indeed
Microsoft Windows [Version 6.2.9200]
(c) 2012 Microsoft Corporation. All rights reserved.

C:UsersJignesh>ping www.google.co.in

Pinging www.google.co.in [173.194.36.24] with 32 bytes of data:
Reply from 173.194.36.24: bytes=32 time=79ms TTL=55
Reply from 173.194.36.24: bytes=32 time=79ms TTL=55
Reply from 173.194.36.24: bytes=32 time=78ms TTL=55
Reply from 173.194.36.24: bytes=32 time=78ms TTL=55

Ping statistics for 173.194.36.24:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 78ms, Maximum = 79ms, Average = 78ms
 
Hi

Just noticed this discussion.
Yes - Forward path seems fixed for now and there is direct India > Singapore route (including for Softlayer network which I used for testing earlier).
As few members have reported better latency while few reported bad latency - the reason is return path for the prefix to which your IP belongs. E.g I am myself getting 200ms latency with Singapore based servers (on most of networks) and reason is that my IP is coming from 117.206.176.0/20 and this prefix is being announced only at NIXI (for domestic incoming routes) and IPLC uplinks (all outside India and indeed outside Asia).
Currently return path i.e from Singapore to BSNL (Haryana) for this prefix is as:
Telstra Singapore > Tata AS6453 Singapore > Tata AS6453 Chennai > Tata AS6453 Mumbai > Tata AS6453 France > Tata AS6453 London > >>>IPLC>>> BSNL Delhi.


Packets are not going to BSNL Delhi router directly from Mumbai but rather taking a route via Europe because BSNL is not advertising this prefix (and many others) on domestic upstream IP transit links from Tata, Reliance and Airtel.
We do not see this problem on domestic routes because more or less we get routes via NIXI



Some other points about problems experienced by members:
Latency with Google varies quite a lot and my strong guess here is that because BSNL does not has direct peering with Google and goes via Tata-VSNL AS4755 link. On top of that BSNL does not has direct IP port from Tata in North (as far as I can see) and hence BSNL in North always accesses Google via it's Mumbai (and sometimes Chennai PoP). If it was Google's Delhi PoP for North Indian traffic then likely we would have seen lot better & stable setup. On top of this BSNL keeps on having issues at NIXI Delhi and it pulls of almost all routes quite a few times a week from Delhi.
E.g as per NIXI's looking glass
218.100.48.15 4 9829 1035 18556 0 0 0 08:20:11 Active


BSNL's BGP session at NIXI Delhi is down from last 8hrs. This further increases latency with many networks e.g say Tulip.net web server in Delhi (which gives 30ms latency in ideal time):

anurag:~ anurag$ ping -c 5 tulip.net
PING tulip.net (116.214.26.243): 56 data bytes
64 bytes from 116.214.26.243: icmp_seq=0 ttl=56 time=87.899 ms
64 bytes from 116.214.26.243: icmp_seq=1 ttl=56 time=90.490 ms
64 bytes from 116.214.26.243: icmp_seq=2 ttl=56 time=90.367 ms
64 bytes from 116.214.26.243: icmp_seq=3 ttl=56 time=89.966 ms
64 bytes from 116.214.26.243: icmp_seq=4 ttl=56 time=89.867 ms

--- tulip.net ping statistics ---
5 packets transmitted, 5 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 87.899/89.718/90.490/0.939 ms
anurag:~ anurag$



87ms rather then 30. And if we look at route, forward route is via Mumbai as per trace and return is surely via Mumbai since there's just no route possible outside NIXI Delhi/Noida for North Indian routes of BSNL. This adds on latency + tends to chock their Mumbai network.


anurag:~ anurag$ traceroute -a tulip.net
traceroute to tulip.net (116.214.26.243), 64 hops max, 52 byte packets
1 [AS65534] router.home (10.10.0.1) 1.830 ms 0.899 ms 0.980 ms
2 [AS9829] 117.206.176.1 (117.206.176.1) 18.084 ms 17.451 ms 17.667 ms
3 [AS9829] 218.248.169.126 (218.248.169.126) 23.967 ms 25.932 ms 26.994 ms
4 [AS9829] 218.248.246.130 (218.248.246.130) 57.167 ms 58.130 ms 60.774 ms
5 [AS4755] 115.113.128.17.static-mumbai.vsnl.net.in (115.113.128.17) 78.242 ms 52.029 ms 52.416 ms
6 [AS0] 172.31.19.22 (172.31.19.22) 79.360 ms 78.804 ms 79.825 ms
7 [AS0] 172.25.83.37 (172.25.83.37) 78.375 ms 78.600 ms 79.069 ms
8 [AS4755] 115.113.248.250.static-delhi.vsnl.net.in (115.113.248.250) 84.018 ms 85.004 ms 84.301 ms
9 * * *
10 [AS37986] 116.214.26-243.del.tulipconnect.com (116.214.26.243) 89.729 ms 86.852 ms 86.789 ms
anurag:~ anurag$






So they need to fix lot of issues including IPLC which is primary pain for high latency with East Asia. For domestic routing they need to fix their unstable setup at NIXI Noida apart from buying IP transit ports in Delhi from Tata/Reliance/Airtel since Mumbai/Chennai are quite far and any issues at NIXI tend to chock their domestic NLD when traffic is re-routed.



(Btw regarding original issue I have blogged with updates with more detailed explanation here - http://link.anuragbhatia.com/bsnl-issues )



Thanks.