Remember the Internet Exchange Point (IXP) discussion from about a year ago? Well, looks like we may be making some headway now. We’re at a regulator (POTRAZ) organised event in Harare today where the implementation of the Exchange is being discussion.
If you wondering what an Internet Exchange is, it’s a physical place (usually a room with machines and cable connections coming into it) that all internet providers would connect to in order to exchange internet traffic, essentially to ensure that the movement of internet traffic is as efficient as possible through their networks.
For example, it ensures that when you access a locally hosted website, your request for that website travels the shortest distance possible to that website and content sent to you also travels the shortest distance possible, giving you a faster internet experience.
Zimbabwe already has an IXP which is managed by an association of ISPs called ZISPA (Zimbabwe Internet Services Providers Association). We’re not sure how well things are working there. In the past there have been issues of routing that would cause all kinds of problems for ISP customers. I remember customers of some ISP unable to access our website when it was hosted at YoAfrica some years back.
There were also cases apparently when some ISPs would misconfigure routers resulting fellow IXP members having their international bandwidth inadvertently used by the offending ISP.
Whatever the reasons, there has been talk of an additional IXP to complement the ZISPA one. This initiative is being pushed by the regulators, the Internet Society of Zimbabwe and ISPs. Hence the event to discuss the implementation of this new IXP being held today. We will update with infromation about the progress so far later.
Share
Home
18 comments
Just a correction, an Internet Exchange Point is infrastructure as opposed to a machine. There are several machines and switches e.t.c involved. You make it seem like there’s single machine sitting alone in a room with a hundred ethernet cables coming from it doing all the exchanging. 🙂
A second exchange would be good though. Local traffic is still terribly slow, ironically even when you are accessing sites hosted by your own ISP. Some of the latency issues won’t be alleviated by having a second exhange, there are many misconfigured network appliances that ISPs need to fix as well. I recall failing to access a website till we got the ISP to reconfigure one of it’s DNS servers under threat migrating to another ISP.
That is very true, its not a single machine or router! What tend to happen is that its a single building, or buildings with close proximity connected by a fast self healing fibre network.
Each ISP then put their routers with big guts enough to hold 500 000 plus BGP routes. They then interconnect amongst themselves, absorbing and routing traffic between each.
Any misconfigured router from tone of them can take out others.
Since we have about 4 big ISPs, it means we will have 4 big internet routers in there!
you might find that the ISP’s leave their routers at their own premises and only push a fibre in and then peer over the fibre rather than deploy a router in the IXP – that said putting a dedicated router in the IXP point is a good idea
if its done the way of a route reflector then the IXP router doesn’t have to be huge as all its handling is the BGP table and nothing else – needs to be managed competently though to avoid some of the issues like the famous Pakistan accidental take down of youtube a few years ago
my bad. initially wanted to go with a simplified explanation, then thot, the heck let me just say what it is, but somehow left the word machine in there.
Its possible for it to be one huge machine. With SDN’s of course.
SDN introduces latency, even if its minimal, it does. One things you do not want in ISP network is latency in any form.
Hardware is always better in this case!!
I thought that was what ZINX is for ? and why create a second exchange and not just fix the one currently working ,every ISP is currently using it
Not exactly most ISP (am looking at you LIQUID) do not route local traffic locally. Why ? because idiots that why
ZINX is an opt-in solution that is to say if your ISP wants to use it they can otherwise you are going for a round trip to London :-p
I think you’re wrong. I’m on a ZOL connection and I can trace a direct route to the few local ISP I tested just now.
can you qualify that with facts please? example trace routes and source/destination ISP’s etc so be careful not to blame Liquid only when peering is a 2 way street – it might well be that your source ISP doesn’t peer with ZINX properly or some other issue that only proper diagnostics and reporting of issues would locate and hopefully resolve
btw this new IXP would also be “Opt In” i can’t see anything that “Compels” or obliges an ISP or IAP to peer with ZIXA – its obviously in their interests at the end of the day but thats a different perspective than being compelled to
That can be your ISP, if you are using PowerTel or TelOne, they interlink with Econet or Liquid at Linx in London
No Telone we peer locally with our global AS exchanging ALL AFRICAN traffic with them.
We peer with other ISPs on our local AS exchanging Zimbabwean traffic
We keep local traffic local
#AfricanDataInAfrica
ZINX is an Opt-in System your ISP has to actively send traffic through it most (like LIQUID) dont – and there technically is NOTHING wrong with ZINX – its the ISP’s where the Problem is
The fact that Zinx didnt go for the tender or considered should explain a lot about it
ZINX is an Opt-in System your ISP has to actively send traffic through it most (like LIQUID) dont – and there technically is NOTHING wrong with ZINX – its the ISP’s where the Problem is
can you qualify that with facts?
You do not understand the IXP at all, do you?
You’re all a bunch of nerds and you make me sick.