Thursday, September 30, 2010
Wednesday, September 29, 2010
Sunday, September 5, 2010
Here's happy John. John is happy because the devices have linked and he's strapping in a FabFi reflector which has made a 15dB improvement in the signal.
Great signal. Left devices in place.
Heath Center is a great site. Tons of community focus and welcoming staff and management. Was approached by interested, friendly, and eager random public while working. Wanachi guys and management have been unbelievably helpful and enthusiastic (especially in comparison to the recent days of firewall-ish frustration with UoN ICT).
Linux Tom peaking at the Mountain View health center
FabFi playing with the big boys on the Lavington tower.
Amy tired. More later.
ps - did I mention there's a JoinAfrica Facebook group?
soulstar - (Linux) Tom, from UoN Fablab
OnsomuOchoti - John from UoN Fablab
lusrandi - Nicholas from UoN fablab
antoinevg - Antoine from Afrimesh
Saturday, September 4, 2010
For the last week the Fablab team has been working 14 hour days configuring devices and preparing materials to deploy a wireless network at a UoN campus. The team deployed their network on Wednesday, and by the end of the day Friday had yet to receive an uplink. UoN's IT department cited inadequate communication of plans and exclusive behavior as an excuse for refusing the connection despite failing to attend any work, meeting, or briefing sessions at the fablab, to all of which they were explicitly invited.
The IT dept. must have been fully occupied making a fuss over our project because the internet in the lab was completely down for at least a half hour each day this week, and down for more than ten hours on Friday.
In less than the number of man-hours wasted waiting for UoN IT Wednesday through Friday, the team designed a reflector mount to accept a new device, planned a ptp link and connected a remote site to 5Mbps of generously donated Wananchi bandwidth across 3.5km to Mountain View. Not only did Wananchi donate bandwidth to the project with little more than a handshake on our research data being open-access, they also lent us tower space and two pairs of helping hands on a Saturday to git-er-done. Meanwhile UoN IT continues to frustrate the efforts of its own students to provide wifi on part of their campus.
With some great help from David and Ken of Wananchi, we linked up with a nanostation5M-loco at -56dBm, putting the gain of the big fabfi reflector at 15dBm over the stock device.
Thanks, Wananchi. We all appreciate your support of the Fablab's hard work.
As Amy alluded to her in post today, the remote end of our link is an excellent community hub bringing access to a health center, schools and residential communities across multiple socio-economic strata. The vision for this location is to provide local and educational content from the KENET/UoN system as a free service in parallel with a paid premium service on the commercial link, both serving the KENET outreach mandate and providing a valuable community resource. That is, if someone at UoN will get us an IP that works...
Friday, September 3, 2010
Fnd killer spot in mt view
Near to Lake Victoria in the western part of Kenya, the ARO Fab Lab has a 1024k down, 512k up Linkstar satellite connection that they’d like to distribute to neighboring schools and business areas. Three intrepid labbers have been drinking from the firehose this week in Nairobi learning “everything” to install a network themselves. They’ve been part of the Lower Kebete deployment team all week during the day, and backfilling on theory and processes at night. Late into the night.
It’s just past midnight-thirty and I’ve just sent them off to bed, Tom, Lawrence, and Hansel, so they can return in the early AM and get started on planning then flashing and configuring devices and fabricating the reflector. Tom and Lawrence have been here the whole night figuring out the logistics and planning for the above network. Before they leave on Saturday they owe me a rough schedule of install and I’ve extracted a promise of emailed updates and photos so we can follow along.
Here’s the part that just blows me away. Based on their device count, the approximate total cost of this network is $1,580.
Thursday, September 2, 2010
Wednesday, September 1, 2010
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).
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.
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!
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.
Yeay! Now back to my long, long to do list…