Elephant in the room here -> There's no money to be made there (yet)
Why the Internet in Sub-Saharan Africa Sucks
11–20 of 127 posts
Re: Why the Internet in Sub-Saharan Africa Sucks
#12Earlier quoted context omitted.
Also, QUIC helps a lot.
Is QUIC protocol out of experimental phase?
It's only in Chrome and only supported by Google's web server. They are using it though. You can see the stats by going to chrome://net-internals/#quic
They are "working" to get it into Apache Traffic Server.
Re: Why the Internet in Sub-Saharan Africa Sucks
#13Re: Why the Internet in Sub-Saharan Africa Sucks
#14Summary: Nairobi needs a datacenter, with Akamai and Cloudflare.
A local Cloudflare would definitely help with the TLS setup since that's their magic (fast SSL setup without requiring your private key).
Akamai could help, depending on the use case. If the content isn't cacheable, there's not much you can do. They do have a service that keeps open TCP sessions from certain POPs back to your origin, thus cutting down on TCP 3-way handshake time. Clients connect quickly, locally and then send the request because the origin connection is already established. That service is pretty expensive compared to CDN and they don't push it very hard. I used it when I ran an ad network where latency mattered (and could pay for this feature in an explicit increase in conversions and impressions).
Re: Why the Internet in Sub-Saharan Africa Sucks
#15Re: Why the Internet in Sub-Saharan Africa Sucks
#16Elephant in the room here -> There's no money to be made there (yet)
The question then: Is there sufficient first-mover advantage that a far-minded company setting up a datacenter there now could reap the benefits when there is money to be made?
Re: Why the Internet in Sub-Saharan Africa Sucks
#17In one of Google's lobby there is a 3-D animation of a spinning globe, with spikes that appear wherever and whenever searches are performed, with the height of the spikes being proportional to the rate of query submissions.
The recruiter who showed me this was very proud of google's accomplishment and indeed it was impressive to see the spikes erupting from the east and west coast of the US, Europe, Japan and India.
But I watched quietly, sadly and solemnly South America and Africa. There were modest spikes in south africa and brazil. Not many anywhere else.
Both continents are quite populous.
Among the reasons for the choices I've made in my life and my career the last few years is that I have come to regard the vast majority of the code I've written in my long career as a developer as being of no real lasting value to anyone.
Instead I wander around downtown and oldtown portland seeking out crazy people then finding some way to lift their spirits somehow. It can be a little frightening as I specialize on the particularly violent sort. Those are also the most cruelly isolated:
"LEAVE ME ALONE! LEAVE ME ALONE! WHY DON'T YOU LEAVE ME THE FUCK ALONE?"
The gent he was accusing had no clue.
"Hi! My name is Mike. What's yours?"
"Jerry! Pleased to meet you."
Does it have to be so hard?
My point is that those google search spikes would be erupting all over South America and Africa were someone to take responsibility to enable them.
I intend to contribute somehow, someday, in part by setting up some nameservers in cybercafes. There are actually lots of them, cybercafes are quite popular in the developing world.
Re: Why the Internet in Sub-Saharan Africa Sucks
#18Northern China talks to Southern China through Japan, to the extent that they talk at all. All of China at first had one state telecom monopoly, but then some party official set his son up with Northern China. What was left of the state monopoly won't route to it.
My take on what to do about that is to set up servers in both north and south that have website mirrors, as well as a geographically-sensitive nameserver. I don't have a clue how to do the GeoDNS but surely it is a solved problem as big sites require it for load balancing.
Re: Why the Internet in Sub-Saharan Africa Sucks
#19That is suprising to me.
I lived in Tanzania for a year, and while the internet connection there was not perfect by any stretch of immagination, I had way less ping than this (100 ms to London, where my company's servers are, 120-150 to Boston).
For reference, that is: aerial to my ISP two kilometers away, then fiber to Dar Es Salaam, then Seacom to Mombassa, then Seacom again to Marseille, then terrestrial fiber to London via the channel).
This smells a lot like contention on the seacom link (the big fiber optics carrier that goes up the east coast of Africa through the Suez canal and accross the mediterranean to Marseille). I would be interested to see a traceroute.
Most of the times ISPs in west Africa don't buy enough upstream traffic on undersea cables, creating contention. I can't blame them, it's expensive, but it's mostly an economic argument, not because the servers are too far away (also, most ISPs run pretty agressive caches of everything under the sun).
Re: Why the Internet in Sub-Saharan Africa Sucks
#20Earlier quoted context omitted.
Is QUIC protocol out of experimental phase?
Yes and no. It is a ways off from being solidified as an RFC AFAIK, but it serves a healthy portion of Google's traffic. So, more than an experiment but less than a standard.
I was looking for a reliable audio streaming method for a project, I came across QUIC and was in process of studying it. Its definitely seems to be the future, but I am still unsure though about using it for the project.