Jul
7

More and more I've been receiving inquiries about the costs per square mile for a CUWiN-based mesh wireless system. Interest in this number really took off with the announcement by Tropos that it would cost $68,000 to wireless one square mile using their gear. At the time, I found this number extremely strange -- more plucked from thin air than empirically derived to suite the multitude of different geographies and topologies. Due to large increase in the number of inquires I was receiving, I wrote up this quick numbers-crunching on CUWiN networking costs.
Which brings me to today -- I just read a piece by Eric Griffith where he gets this information from Bert Williams, vice president of marketing for Tropos Networks:
- "We usually talk in terms of $75,000 per square mile for our gear," says Williams, who estimates that the Tropos hardware accounts for about 60 to 70 percent of the total cost of installation. Adding the ongoing costs for backhaul and system management, he believes the number to be closer to $100,000 per square mile.
Now wait a second -- first $68,000? Then $75,000?? Then $100,000??? This is why I'm skeptical about the firmness of numbers when they're provided. A critical consumer of municipal networks needs to be extremely careful about the total costs of their network. So, to add to the debate, and provide a counterpoint to proprietary systems -- here's some estimates for what the total cost is to build a CUWiN-based network (which is a non-proprietary, open source software and open architecture hardware system) in five easy steps:
- First, one needs to estimate the density of nodes needed per square mile. Click here for information on how I computed my node density numbers. This number will let you estimate how many nodes you'd need to cover a specific geographical area.
- Second, nail down your up-front, per-node networking costs. Here are estimates for a CUWiN-based system they are purposefully at the high end of the range (e.g., one can buy nodes for under $350, hourly wages are set at $50/hour, one doesn't need any server equipment on the network -- though why not throw in Intranet services and applications instead of just focusing on Internet Service Provision?):
- $350 -- Per Node Cost of Hardware & Software
- $250 -- Setup, Construction, & Installation (5hrs/per node@$50/hour)
- $5,000 -- Server Equipment (per network)
- $60 -- Node Upkeep Costs (10% Node Failure Rate Equivalent)
- $15 -- Electricity for One Year
- $600 -- Bandwidth for one year (per 10 nodes)
- Third, identify your network size (square mileage/node density) and your start-up costs for the network (i.e., your first year costs) -- here's some quick benchmarks for variously sized wireless networks for quick reference:
- $8,750 -- Node Costs of Hardware & Software
- $6,250 -- Setup, Construction, & Installation (5hrs/per node @ $50/hour)
- $5,000 -- Server Equipment (per network)
- $750 -- Node Upkeep Costs (10% Node Failure Rate Equivalent)
- $375 -- Electricity for One Year
- $1500 -- Bandwidth for one year
- TOTAL: $22,625
25 Nodes
- $17,500 -- Node Costs of Hardware & Software
- $12,500 -- Setup, Construction, & Installation (5hrs/per node @ $50/hour)
- $5,000 -- Server Equipment (per network)
- $1500 -- Node Upkeep Costs (10% Node Failure Rate Equivalent)
- $750 -- Electricity for One Year
- $3000 -- Bandwidth for one year
- TOTAL: $40,250
50 Nodes
- $35,000 -- Node Costs of Hardware & Software
- $25,000 -- Setup, Construction, & Installation (5hrs/per node @ $50/hour)
- $5,000 -- Server Equipment (per network)
- $3000 -- Node Upkeep Costs (10% Node Failure Rate Equivalent)
- $1500 -- Electricity for One Year
- $6000 -- Bandwidth for one year
- TOTAL: $75,500
100 Nodes
- $87,500 -- Node Costs of Hardware & Software
- $62,500 -- Setup, Construction, & Installation (5hrs/per node @ $50/hour)
- $5,000 -- Server Equipment (per network)
- $7,500 -- Node Upkeep Costs (10% Node Failure Rate Equivalent)
- $3,750 -- Electricity for One Year
- $15,000 -- Bandwidth for one year
- TOTAL: $181,250
250 Nodes
- $350,000 -- Node Costs of Hardware & Software
- $250,000 -- Setup, Construction, & Installation (5hrs/per node @ $50/hour)
- $5,000 -- Server Equipment (per network)
- $30,000 -- Node Upkeep Costs (10% Node Failure Rate Equivalent)
- $15,000 -- Electricity for One Year
- $60,000 -- Bandwidth for one year
- TOTAL: $710,000
1000 Nodes
- Fourth, estimate your yearly maintenance and upkeep (recurring) costs for your network:
- $3,875/year -- 25 Node Network
- $7,750/year -- 50 Node Network
- $15,500/year -- 100 Node Network
- $38,750/year -- 250 Node Network
- $155,000/year -- 1000 Node Network
Estimated at $50/year per node + 10% upkeep cost + electrical cost:
- Fifth, estimate monthly costs per node/house based on different network sizes and break-even points:
- $33.75 -- Over 3 Years
- $25.42 -- Over 5 Years
- $19.17 -- Over 10 years
25 Node Network
- $30.97 -- Over 3 Years
- $23.75 -- Over 5 Years
- $18.33 -- Over 10 years
50 Node Network
- $29.58 -- Over 3 Years
- $22.92 -- Over 5 Years
- $17.92 -- Over 10 years
100 Node Network
- $28.75 -- Over 3 Years
- $22.42 -- Over 5 Years
- $17.67 -- Over 10 years
250 Node Network
- $28.33 -- Over 3 Years
- $22.17 -- Over 5 Years
- $17.54 -- Over 10 years
1000 Node Network
These are high estimates for this type of network -- if one can get nodes for under $350, or labor for under $50/hour, one can bring down the costs substantially. But clearly, the costs are far cheaper for an open source, open architecture system than are currently being estimated for proprietary networks.
I have endeavored to break down each of the expenses to it's basic components and hope that this will enable potential community wireless networkers to make more informed decisions about the options available. Clearly, this level of detail should be available to (and demanded by) any organization, business, or municipality interested in deploying these types of networks. I hope that this information will help move the "wireless network costs" debate forward and will be a catalyst for more transparency in the underlying assumptions being used to generate municipal wireless cost estimates.
Download this attachment for all of this information in a handy 1-page spreadsheet -- allowing anyone to change any of the difference parameters to instantly get new estimates. A useful tool for everyone thinking of building wireless networks.
[UPDATE 2] Thanks to everyone who's e-mailed me thoughts, ideas, methodological tweaks, etc. Please remember that this piece provides a heuristic that people can use to generate wireless networking cost estimates. As such, this is not an all-encompassing algorithm for measuring deployment costs and node densities (which would need to take into account a great number of additional variables -- such as spectrum congestion levels, multipath interference, antenna height, the non-circularity of wireless coverage patterns, etc. etc. etc.). The numbers provided are based on real-world implementations, and will prove relatively accurate most of the time. As always, I encourage everyone to remain a critical consumer of any cost estimates you see, especially ones that are given without any context and without being broken down into their constituent parts.
A special thanks to Andrew from www.funferal.org who tracked down this article on optimal circle-packing density which quite indulged the geekier side of me.
| Attachment | Size |
|---|---|
| Network Cost Estimates.xls | 13.5 KB |

Sasha,
How does this model change if the End User buys the CPE in order to be an active participant in the Mesh? What business models are there for creating incentives for the End User to become an active ands engaged member of the Mesh?
To me this is the key competitive business advantage for a community network: The end users become investors and cooperative owners of the infrastructure. This is in clear contradistinction of the encumbents who refuse to allow end users to share or other wise participate in the mesh.
Think of Benjamin Franklin and his strong faith in Mutual Aid Societies. Only ours would not be closed and secret.
"1727: forms the Junto, a group of tradesmen in Philadelphia--including silversmiths, glaziers, cobblers, clerks, ironmasters, and joiners. The Junto was dedicated to the improvement of its members through the exchange of information. Esmund Wright, author of Franklin of Philadelphia, describes it in this way: "It was part mutual aid society, part social fraternity, part academy. Its organization was modeled on Mather's neighborhood Benefit Societies, but it was touched also by Masonic principles: it was intended to be secret and exclusive. The questions the members set themselves included 'queries on any point of Morals, Politics or Natural Philosophy,' but the real motivation was self-improvement, the 'wish to do good' that would also bring them advantages, or even profit" (37-38). Franklin describes the Junto's work in his autobiography, as well as a few essays."
http://www.uncp.edu/home/canada/work/canam/franklin.htm
I look forward to your thoughts.
Jock
Hi Jock,
You've brought up some great points in your comments. Mainly, as active participation increases, network costs tend to drop (in some cases fairly dramatically). For example, in my cost estimates, I assume that it will take 5 person-hours to construct and install each node. I further assume that it will cost $50/hour for this work. CUWiN and other wireless groups have often held "node-building" parties, relying on volunteer labor to build the nodes, which can cut this cost in half. Likewise, if participants install their own nodes, installation costs drop to zero also. This can dramatically effect the costs of the network. And, if anyone thinks this is a crazy idea (that it won't scale, isn't feasible, etc.) -- I encourage you to take a look at the Djursland Wireless Network, a massive, rural, community owned and operated, volunteer-led effort (note the site is in Danish) and which I got to experience last Fall at the Fresh Air, Free Networks Summit.
Telecommunications historian Dan Schiller is fond of the Do-It-Yourself nature of the Home Rule Telephone Movement -- who, 100 years ago, created a massive number of community-owned and operated telephone exchanges -- one of his favorite requisites was from one of the communities which allowed anyone to join if they brought a bunch of wire and their own telephone pole. So yes, there's an incredibly long history of people building their own networks. And yes, they can be built cheaper than for-profit services, and they can increase (often dramatically) access to telecommunications. This is why it's so important to protect the self-determination rights of local constituencies -- and why telecom encumbants are working so hard to eliminate these rights at the state and national levels.
Thanks for the link to Sascha's web site. It provides a useful methodology for cities to begin estimating the costs of building out wireless networks. I have compiled a database of 40 case studies of public owned broadband networks. (The database includes both fiber and wireless networks.) Using this database, I've compiled various types of line items that contribute to the cost of the network. They include:
.Installation
.Backhaul & Connection Charges
.Equipment, Cable & Hardware
.Feasibility Studies
.Legal/Administrative
.Marketing
.Tech Support & Network Operations
.Billing Systems
.Insurance
.Debt Reduction
I presented findings from the database at a Washington Association of Telecommunications Administrators and Officers meeting in June. The powerpoint from the presentation is available at my web site at:
www.appcom.net (It is located in the Document section under the
Telecommunications Planning link).
Kate McMahon
Applied Communications
710 W. 7th St
Whitefish, MT 59937
(406)863-9255
kate@appcom.net
www.appcom.net