Thread: Hayai Broadband: How it all began!

Originally Posted by admin

frankly speaking... 8Mbits/80GB cap is not going to work for most users on this forum at least.

the faster the speeds you get, the more you end up downloading. the more you would want to download. 80GB is nothing when you are downloading at 8mbps.

Agreed, I know exactly what you mean. As mentioned, prices and whatnot are purely speculation at the moment because of all the new things coming in - I had to throw all my old prices out the window. All I can say as of now is "stay tuned".

Originally Posted by admin

and about the local network free downloads offering... it's worthless unless you have at least a thousand users on the same network in the same city.

if a locality has just 10 users on your network... it is very unlikely that any of these customers would be downloading the same file using a p2p network.

Agreed. However it's not limited to any particular locality - it is, literally, anyone connected to our network, anywhere. The beauty of having our own fibre. Given that we assume that the service will attract more than 10 users, we're confident that, although it might be a tough first few months, once our user-base hits 5,000 or so, the aforementioned advantages will become very obvious.

The size of our first line suggests that upon launch, we'll be ready to support about 31,000 users on the "regular internet" according to the regulations, and if I have my math right, most parts of the internal network will support all 31,000 users transferring at about 16Mbits synchronously, simultaneously and as a non-stop stream (which would be about 4TB/user/month).

This is assuming everything is run through only 1 switch for the whole network - actual throughput will be higher, of course, because we're not running just one switch - that would be a recipe for disaster. Speaking of which, I shouldn't be making these calculations at this time of the day. Thank you for all your questions and feedback thus far - keep it coming.

After a while (but in the nearest future) maybe you could look at providing IPv6, at least to your fixed line customers?

Originally Posted by admin

and i guess local network only comes into the equation with a p2p application.

http downloads are simply not going to gain anything if another user is or has downloaded the same file in the past. so that's painful as well.

It depends. As mentioned in my original post, we plan to mirror a lot of commonly downloaded files from a lot of common sites, but of course we can't do that for each and every file in each corner of the internet.

Naturally, we'll be monitoring traffic usage and trying to optimize the network appropriately, but of course, even HTTP downloads should still be fast - we're purchasing many gigabits of international bandwidth, so this shouldn't be a problem.

The fact that we are (somewhat) concentrating on p2p traffic is because statistically speaking it constitutes one of the largest areas of usage for your average ISP.

Originally Posted by coimby

After a while (but in the nearest future) maybe you could look at providing IPv6, at least to your fixed line customers?

IPv6 isn't really difficult to implement: it's a matter of the client being enabled to handle it. On our end, it is something we can implement immediately, because when you buy IPv4 addresses, you can buy v6 addresses at the same time. As such, IPv6 would be able to be provided to those who want it, though strictly speaking there isn't much benefit to it at this point because to surf most of the net, one still needs a corresponding IPv4 address.

Originally Posted by admin

mmm. lets take this example.

updating ubuntu distro.

it picks up packages from one of the many servers.

even if you replicate all the latest files at your end, how would you direct the user's machine to bypass the international gateway and use the local repositories?

i believe the user would have to manually configure your local servers...

same case with windows updates.

and any other http download i believe.

i mean... http is not designed to be smart enough to locate a more efficient route to download the same file from the web. it points to a fixed location from what i know

Originally Posted by saurabh28

Agreed

Actually, we would be implementing the same sort of system that Akamai (and indeed most CDNs) use - it's all about DNS and CNAME entries - it was explained to me in great detail by the folks at Akamai, but unfortunately I was in traffic at the time.

The way around this is the HOSTS file or using an IP address directly - but even then, companies including Google do this slightly weird thing using the same IP address for most of their servers so that, even though a given IP address is reported to be in LA or London or whatever, the actual server serving India might be located somewhere like Singapore. This sort of thing can be done with clustering also.

Anyway HTTP can be made to be smarter than you think. As for things like Ubuntu, it wouldn't be so hard to have an official mirror (which would be then implemented in an update) or by the geo-location feature at Ubuntu.

*edited*