we’re starting the “live” system deploy tomorrow, uh, today

Cross posted from Amy. I love it when she does all the writing for me...

For the past 3 weeks or so, Nipun, Nick and others have been running about Nairobi scouting and getting permissions to mount FabFi equipment on towers and roofs as well as connections into the internet. They’ve settled on some sites which broadly speaking are the Mountain View area (which includes some of Kangemi) and Loresho (and some of Kangemi).

Mountain View and Loresho are the two possible areas for the pilot. The orange circles are internet uplinks for the network.

We wanted to make the backhaul connections along schools and examined the communities near Nairobi School and Kangemi School. In both cases we could see different socioeconomic strata pressed up against each other and somewhat well mixed.

Planned community (upper left corner), slums, and estates. The community near Kebete.

There's some distance between Nairobi school and where people live.

We had to choose one of the areas to build out first. Making the 5 GHz backhaul among the main repeater sites and the internet uplinks as well as making local omnidirectional 2.4 GHz wireless AP’s at each repeater site consumes approximately 1/3 of the devices we budgeted. We estimate that building out any one area will consume the remainder of the devices. So it seems we must chose one area to build first and use the profits from the first to buy the devices to build out the second.

For many reasons the Mountain View area wins. We’re now prioritizing building up expanded access around Kangemi school and Mountain View areas. The technical team now mostly turns to getting equipment up and working.

Nipun has a homework assignment to work the numbers to figure out what the price(s) should be for the paid “Premium Service Level” – that’s how we make our nut. Here’s his assignment, I’m sure you b-school folks will get all kinds of excited.

The analysis has two components:

1) Determine long-term “sustainable” pricing for the system based on going tier1 bandwidth rates. This is for the conservative case where we’re collecting no revenue other than from selling bandwidth.

2) Assuming free bandwidth, what do we have to price at in order to finance duplicating our current deployment in three months? (see hardware cost below)

Here are the inputs:

- Residential market price for bandwidth @1Mbps (assuming 1/3 of operator cost is bandwidth this is about a 20:1 contention ratio) = 4,999KES/mo (zuku)

- Going rate for backhaul (tier 1) = 32,730KES/(Mbit*mo)


- We should have a contention ratio the same or better than other providers

- Deployment to Kabete / Mountain View with backhaul will consume ~810,250KES worth of hardware (attached is a potential coverage map).

- We assume about 10% of the bandwidth will be consumed by “free” customers (but we don’t know. Maybe your analysis will dictate QoS rules…)

Assume the average device must be replaced once every three years

Required outputs are prices for access-cards of the following duration:

- 1 day

- 1 week

- 1 month

In long-term case fees will have to cover:


- Bandwidth costs

- Staff (150,000KES/mo)

- Local Transport (black van with no windows…)

- Overhead (legal, g&a, etc)

- Maintain physical system

- Expand physical footprint on a reasonable timescale

Short term, we’re only concerned with making 810,250 KES above our costs, after staff in 3mo.

Our model is that you can always connect for free using the unused part of the bandwidth. Wikipedia and some domains such as .edu, .edu.ke, .gov, some locally mirrored content such as MIT’s Open Course Ware, and similar resource information type sites are exceptions – they aren’t ever limited. You can also pay for Premium service which is faster since you are guaranteed some level of service. If that sounds impossible, the intrepid folks at Afrimesh showed that given free, slower service, people would pay for the faster service in an earlier implementation in Scarborough, Cape Town, South Africa, reaching profitability about a year ago!

Organizing the FabFi equipment for tomorrow's deploy in Lower Kebete.

Tomorrow we’ll be doing a three-point backhaul deployment at the UoN Lower Kebete School of Business tomorrow (connecting their Admin building, library, and student center). It’s mostly a training and deployment preparedness shakedown and each 5 GHz backhaul site will require something unique, for example a solar panel at the student center, hacking a Linksys to provide POE to the Ubiquiti devices at the admin building, and some challenging penetration at the library. And if we still have the energy, we’ll haul a large FabFi up the 12 story radio tower which will ultimately point to Nairobi school and a site in Mountain View.

Shopbot Tom is bleary-eyed after spending all night modifying and cutting out a large FabFi reflector to work with the Ubiquity 5 GHz NanoStation.

Yeay! Now back to my long, long to do list…

shaping the Nairobi Join Africa backhaul network

Ok, look, what can I say, I'm impossibly behind on the blog thing. There's just too much... a million things from Fab 6 and then immediately after a million more things from Maker Faire Nairobi, plus another million from the Join Africa deploy. By my math that makes a "bajillion" and I've been only collapsing into bed without the willpower to tell you about it.

Connection diagram for a Ubiquity device when using AC power

But tonight we're pushing some documentation on to the wiki and I want you to see. This is an image that was made by Tom Okite, Hansel Omondi and Laurence Ombuki from ARO FabLab Kenya West in Kisumu. They're in Nairobi this week to help deploy the Nairobi networks in anticipation of doing the same in Kisumu.

mapping out the back haul sites and internet connection points in Nairobi

You can find the (in work) gallery at: http://www.fablab.is/w/index.php/Power-over-ethernet_diagrams.

And go here a little more on the FabFi and Afrimesh project we're calling Join Africa.

Fabfi and Afrimesh: Building a Wireless Africa

Fabfi, the open source platform for building large-scale mesh wireless network infrastructure, and Afrimesh, the platform-agnostic, open dashboard for network management, have joined forces to build a turnkey platform for deploying, managing and monetising high-speed data networks under the codename JoinAfrica.
Over the last few weeks, we have completely overhauled the Fabfi platform in anticipation of a pilot project near Nairobi Kenya. After some feverish last-minute work, we're also providing wireless at Maker Faire (it's beta, so please use it and tell us if you notice anything that's not working right)

SSID Makernet by Fabfi
User: pamoja
Password: JoinAfrica Note: no proxy is required


For those of you reading from afar, this page is being used as our info page at MakerFaire Africa, so it might seem strangely like an advert... because it is.

Kenya, Week 1

The first week of the Keyna deploy was, in a word, intense. The Nairobi lab has a great group of talented [super] users, and it was easy to fill the roles of linux geek, project manager, and JOAT (jack of all trades) in the first couple days. It was a good thing as well, because we might have bitten off a little more than we could do for a first week. After very positive meetings with the Permanent Secretary for ICT and university officials, it became clear, that "just getting something up" would not be enough to fully capture the opportunities available. The network not only had to work, but it had to be "hish-performance" in a way that fabfi had never really considered before.


The demand for reliable, high-speed connectivity to end users very high, and in the university/Education environment the desire for locally hosted content is strong enough that sitting behind the uplink speed as an upper requirement for network performance is effectively a cop-out. It turns out, however, that adhoc networks don't support N-speeds, so we spent the week reworking fabfi to support AP/STA operation, debugging the new devices, and upgrading to WPA. It was a great experience in understanding the software, but as of Friday the number of nodes deployed in the field by the crew was exactly 4 -- enough to learn cable making, battery power and basic pointing, but leaving a little too much to the imagination for my comfort. Only time will tell how the list of "needs done" will fare in my absence, but in the meantime off to fab6 for a couple days of much-needed geek R&R.

Best Hotel Spread EVER

Sorry Nipun, you're sleepin' on the floor--cuz if it doesn't all fit in the picture, it certainly doesn't fit in the hallway. Let's go build a series of tubes...

Unrelated travel note, courtesy of the WC at Amsterdam Centraal:

Pay toilets are a disservice to humanity. The idea is noble enough: You want to have a nice bathroom so you charge people some to use it and use the money to keep it from getting crapped up (pun fully intended). The problem with this logic is that when people gotta go, they gotta go. After walking the hundred yards to the WC at the end of the Amsterdam Central train station platform to be thwarted by an electro-mechanical gate that ONLY TAKES EXACT CHANGE, the last thing you're inclined to do is feel respect for city ordinances against public urination. By the smell outside from 30yds away more people have let go on wall of the WC building than inside. Surprised the gate isn't shorted out now.

The moral of the story here is that everyone benefits when certain things are free. Keeping the whazz of the train platform is a pretty clear example, but I don't think certain types of internet access aren't that far behind in today's world. That's sort of why we're here...

Kenya Ho!

On the plane to the Kenya Fablabs with 275lb of wireless gear (most of which can be found in Kenya, for future expansion). Fabfi v4 is gettin' coded in country! Let's just hope Kenyan customs doesn't follow this blog. [Edit: it's all perfectly legal, but it's an easy target for the odd entrepreneurial airport security guy...]

Teaser of things to come

May dad tells me I haven't been blogging much, so it must mean I'm busy. He couldn't be more right...

Fabfi is going to Kenya in August, and there's more to do than I can keep in my head all at once. Amy and I did a little late-night planning session on Friday, and the output was "you're not sleeping 'till September". This looks like it could be the biggest deploy yet, but you'll have to wait for the details. As a wise man once said: "Under-promise. Over-deliver".

Stay tuned.