Why FTTH speeds are not growing in India

  • Thread starter Thread starter Kyle Crane
  • Start date Start date
  • Replies Replies 130
  • Views Views 2,397
The 3.3TB cap on 1Gbps (even say 500Mbps plan) is surely bad for end users. I think ISPs are offering 1Gbps just for sake of putting it. They largely do not want end users to be on these plans for now due to GPON and other limitations.


What ISPs can do is provide vouchers for users who need more than 3.3TB to download study material
It likely won't work out. Thing is people hitting 3.3TB cap are using over 10x of what average users are using. ISPs for now want these users to just churn.


To those stating that 3.3TB is more than enough.I agree it is most of the time like 59/60 months.
It's a good point. Instead of 3.3TB a month, 39.6TB year cap will likely solve issues for most of people hitting cap. Unsure how telcos can implement it for non-yearly paying users.


This is pretty much how excitel works in India or all ISPs in the West work.
Would suggest to keep Excitel out of this discussion because Excitel's plans are way better on paper compared to other folks but in reality it's largely manipulated network. There are other threads on IBF about. I also tried their 400Mbps plan in 2023. Speedtest will always gave 400Mbps (literally ALWAYS) but wasn't able to push/pull more than 50-60Mbps even from their large peers like AWS. Excitel for now is great option for someone looking at small LCO backed player which is better in support than telcos, someone who just wants to stream stuff etc. But if goal is to replace 100-200Mbps non-Excitel expensive connection with Excitel for speed than its a bad idea. Thus won't consider their pricing for now as ideal.



India doesn't need to cap data provided ISPs do their homework to identify which connections are what type i.e. home or business.
Again one argue on numbers but as suggested ISPs need it for now because of this goal to offer cheapest service and grab market. If letting high users churn away gives option to serve around 10 users for the same capacity at same ARPU, ISPs will do it. What you said would be valid if ARPU was high enough that in grand scheme of things it doesn't matter. For now it does.


That said bandwidth costs way more here than it does overseas even for ISPs.
This is partially true. When looking at cost keep Jio/Airtel/Vodafone/Tata Comm/BSNL in one bucket and rest of ISPs in other bucket. Cost of bandwidth is not high in India for Jio/Airtel etc. It's similar set of numbers like everywhere else. They do get slightly expensive transit and cost of carrying it to India but that's for barely 5-7% of overall traffic. Rest all Indian cached/peered traffic it's the cost of transport.

For non-Jio/Airtel ISPs specially small ones - they have a) More traffic going to transit than peering (way above 7%, say 20-30%) and even for remaining peered traffic it's often going on the NLD of Airtel/Tata Comm/Vodafone and that is certainly way more expensive than point to point circuits outside of India.


The only real complaint with Indian Internet is the 3.3 TB data cap which is a relatively easy fix for the regulators just revise it upwards to 10TB.
Again, why should regulator get involved in this? It's a cost issue. In hypothetical scenario of regulator getting involved, ISPs will kill higher tier or make it more expensive and same for base tier plans.

3.3TB a month is 10.18Mbps average traffic. Assuming 20% upload, this is around 8Mbps down / 2Mps up for entire month. In reality people hitting 3.3TB cap won't actually running constant 8/2Mbps traffic but likely bursting at 4050Mbps during peak periods for extended periods. That becomes a bottleneck when the aim is to sell cheaper plan with bundled CPE etc.


If not everyone is gonna use full 3.3TB then what's the use of FUP?
The use is to use it as a hard cap (bring down speed to dead slow 1Mbps) if someone is intentionally or unintentionally using over 10x bandwidth of average user. Ideally you want an FUP number which majority of people actually do not hit.


BSNL indeed has higher FUP for gigabit plans wherever available. Check Telangana plans for example, 1Gbps till 12TB and even my 250 Mbps plan has 6TB limit. Idk why others haven't updated their FUP yet.
That's good on their part. BSNL can use this as one of few things to get market share back. That is not the case for other large players for now. As stated in my initial post on this thread imagine everyone except Airtel/Jio is (financially) able to upgrade FUP to say 10TB or just remove FUP, these players would be forced to that as well. For now largely that's not the case. They just don't want those high end users and happy to let them churn away to either leasedline product or other ISPs.



first you bring xpon cost yourself and then talk about distraction. Being a networking pro doesn’t mean you know every major equipment cost to run a isp. Also we are not here to win argument that you still don’t agree 😜
Firstly I do not consider myself as networking pro. Unsure why you bring that up here. About the point - see I stated overall cost is high and somehow thread went crazy with CPE cost. I stated multiple times that while I disagree to the CPE cost and same was confirmed by @thatayush even with 600INR cost, ISPs do not make much on the lowest tier. The whole idea here was to break this assumption that ISPs are making lot of money and can do anything. Reality is that they are making decent money in mobile and for now grabbing market share in fixedline. No one is here to win an argument.
 
@sloj i run a cron job for packet drops for all major public dnses to check for packet loss to move my primary wan weight. Only cloudflare packets get dropped randomly at peak hours make it make sense?
 
@f0r3he4d You see losss with Cloudflare from which network?

I have over 200k IPs in monitoring from my home and servers including 1.1.1.1. Testing Cloudflare 1.1.1.1 every 30 seconds. No visible losses.

cf-airtel-rtk.png
 
Likely it's a BSNL problem and not a Cloudflare problem.

Please share forward trace to 1.1.1.1. It won't give much idea as return path can be pretty wild in case of BSNL and it's hard to know without asking CF for it.

Also, if you have access to a Linux or a Mac, please share output of: dig chaos id.server txt @1.1.1.1 +short
This will give the CF node which you are hitting.
 
@Anurag Bhatia
chiro@PI4:~ $ traceroute 1.1.1.1
traceroute to 1.1.1.1 (1.1.1.1), 30 hops max, 60 byte packets
1 r0ut3r.lan (10.0.0.1) 0.790 ms 0.981 ms 0.964 ms
2 117.247.69.65 (117.247.69.65) 3.265 ms 3.252 ms 3.234 ms
3 10.219.12.173 (10.219.12.173) 3.215 ms 3.376 ms 3.359 ms
4 10.219.12.174 (10.219.12.174) 4.748 ms 5.068 ms 5.437 ms
5 * * *
6 * * *
7 * * *
8 * * *
9 * * *
10 * * *
11 * * *
12 * * *
13 * * *
14 117.216.207.122 (117.216.207.122) 29.576 ms 29.574 ms 29.655 ms
15 162.158.52.2 (162.158.52.2) 30.721 ms 162.158.52.4 (162.158.52.4) 30.366 ms 29.295 ms
16 162.158.52.39 (162.158.52.39) 34.329 ms 31.499 ms 162.158.52.19 (162.158.52.19) 34.633 ms
17 one.one.one.one (1.1.1.1) 33.239 ms 28.380 ms 28.128 ms
when im doing the dig code its not replying anything
 
I ran a few ping commands to 100 or more counts, no packet loss at all. Collectd-mod-ping showed 87.7% loss on cloudflare IPv6 1111 and 10.16% loss on 1.0.0.1 which was running at 60 second interval. Currently running PingInfoView on Windows and only 8.8.8.8 has 100% loss, 8.8.4.4, 1.1.1.1, 1.0.0.1 has no loss so far. Also, I don't think packet loss will have any major affect on your DNS resolution.

You can run the dig command on Windows command prompt with older Bind tools version 9.16 or use Termux on phone.
 
for a tcp connection it wont be a problem but for gaming it will be a problem anything above 0.1% packet loss is noticeable. I have dnsutils running on my homelab from where i ran the command
 
Back