Wednesday, October 27, 2010

Superfast broadband in Herefordshire

I did some work earlier this year to see if the model set out in JANET(UK)'s local loop unbundling (LLU) reports could be replicated to model connectivity options and costs for other regions and local authorities.

Coincidentally, one of the authorities I selected to try this out was Herefordshire, which has now been selected as one of the four superfast broadband pilots announced as part of last week's comprehensive spending review (alongside North Yorkshire, Cumbria and the Highlands and Islands). A long post this, so you might want to get a cup of coffee before you read any further.

A word of warning: all product and cost information below is based on publicly available Openreach information as of April 2010. I haven't reviewed this data, so the information below should be regarded as indicative rather than definitive.

My first step was to create a list of all schools in Herefordshire from the Department for Education's EduBase database, listing all primary, secondary and special schools, together with their postcodes. The result was a list of 88 primary schools, 20 secondary schools and 4 special schools. The ever helpful folks at SamKnows then very kindly used this data to generate a new list matching school postcodes to their serving exchanges, together with the radial distance in metres between the school and its exchange.

The key JANET LLU report is the Stockport LLU/Access Locate Feasibility Study. This describes an approach for modelling an LLU deployment across the authority, based on provisioning secondary schools with 100Mbps connections and primary/special schools with 10Mbps connections. All costs are based on Openreach's open book pricing.

The Stockport study based its approach on using copper to deliver symmetric 10Mbps connections to schools within 3km of their exchange:
"Given the requirement to provide 10Mbit/s symmetrically to all Primary schools it is proposed to use EFM (Ethernet in the first mile 802.3ah) G991.2 (SHDSL) technology running over copper circuits for those schools within range...Basically G991.2 will deliver 5.6Mbit/s over a single copper pair out to about 1.5km and 2.5Mbit/s out to about 3km. To achieve 10Mbit/s reliably we have adopted a conservative policy of using 4 copper pairs bonded together for all sites out to a calculated distance, as the copper goes, of 3km. For sites beyond the 3km limit BT WES10 local access will be used...For 10Mbit/s schools below 3km this is 4xMPFs indicating the intention to use 4 bonded pairs to deliver the 10Mbit/s. These can be delivered into the LLU space in the serving exchange. For schools over 3km we revert to fibre-based circuits."
MPF stands for metallic path facility, or copper phone line. Note that the 3km figure relates to line length, not radial distance, as the key determining factor is the actual length of the line. The study suggests that multiplying the radial distance by 1.4 gives a reliable estimate of actual line length, so the next step was to apply this to the radial distances for primary and special schools to determine which could be connected via 4 copper pairs and which by WES10-LA connections (where LA means local access). WES100-LA connections would be used to deliver 100Mbps for all secondary schools.

The next step is to examine the exchange loading, i.e. how many schools are served by each exchange, to determine a core network design to interconnect the exchanges appropriately. Basically, this involved totalling the number of 4xMPF, WES10-LA and WES100-LA connections per exchange. It's worth noting that in rural areas like Herefordshire you end up with a "long tail" of exchanges serving a single primary school.

The map below shows the locations of all exchanges connecting two or more sites, created using Google Earth:

Yellow pins indicates exchanges serving four or more sites (the Hereford exchange serves the most, 5 secondaries and 19 primaries), red shows exchanges serving three sites and green two. 5 exchanges serve 4 schools or more, 5 serve 3 schools, 13 serve 2 schools and 22 serve a single primary school (these aren't shown above to keep it simple). The greatest distance between a school and its serving exchange is just over 7km; 64 of the 92 primary/special schools are 3km or less from their exchange and 28 are further away.

The numbers in brackets next to each exchange code indicate either the total bandwidth requirement of the exchange (so 120Mbps means the exchange serves 1 secondary and 2 primary schools) or the number of secondary/primary schools served by the exchange (so 5/19 means 5 secondaries and 19 primaries are served, an aggregated bandwidth requirement of 690Mbps, or 5x100Mbps + 19x10Mbps). The total aggregated traffic amounts to 2920Mbps.

Connecting all this up to ensure uncontended, symmetrical bandwidth (employing BES100 and BES1000 circuits of either 100Mbps or 1Gbps respectively) results in a core network that looks like this:

Red lines are BES1000 and green lines are BES100 links. For resilience, a ring round the perimeter could be provided using additional BES1000 circuits. However, that may well not give resilience, as the routes that BT use to connect the exchanges may well involve coming back to Hereford. So there would be an illusion of resilience, but the reality might well be a single fibre duct break could take out the primary and secondary routes. The remaining 22 exchanges connecting 1 school would be connected to their nearest neighbouring exchange via a 100Mbps connection (BES100), these aren't shown on the above map for clarity.

The pricing of these core links is dependent on distance, so the next step was to measure the length of each link using Google Earth. I then calculated the cost of all the above core network links, which gives an install cost of just over £67K and a recurrent cost of around £192K per annum. Note that these costs don't include the construction charges to lay the necessary fibre, we'll come back to those later.

I then needed to add in the cost of connecting the 22 exchanges serving a single primary school to their nearest exchange (the ones not shown above) via BES100 links (to give headroom for growth in usage and capacity to connect additional sites if necessary). I did this by averaging the cost of all the BES100 links in the core and then multiplying by 22 to give an additional cost of £43K for installation and £119K recurrent per annum, bringing the total cost of the core to around £110K for installation and £311K recurrent per annum.

Then I calculated the cost of the last mile links for schools to connect them to their serving exchange. This was much easier as these costs (for MPFs and local access products) aren't dependent on distance. There are some additional costs to consider too, as set out in the Stockport study, such as ethernet over copper modems for schools connecting via 4xMPFs. This gave a grand total of £768K for installation and £554K recurrent, or a total cost of £1.32million in year one to provide every school with either a 10Mbps or 100Mbps uncontended, symmetrical connection, with sufficient additional capacity to connect additional sites as well. Not bad really, even taking into account that there would be additional costs involved in operating and maintaining a network.

However, this is where the wheels start to come off. None of the above costs include construction charges to lay the necessary fibre. I estimated that the network would require 470km of fibre to be laid, which at a cost of £30 per metre (which is probably too low) gives a total construction cost of just over £14million. A lot of money to be sure, but a one-off cost to develop a future-proofed and fully scalable infrastructure. 

The spreadsheet documenting all of this is available here and I've also created a summary of the process steps involved. For me, there are two important aspects to this: firstly, it enables you to understand what the input costs are in any network approach based on Openreach infrastructure (which, let's face it, may well be the only commercially available infrastructure in many rural areas). But secondly, and perhaps more importantly, it gives you the opportunity to grasp the nettle of local provision, creating an infrastructure over which you have full visibility and control - you're commissioning from the private sector, rather than being sold a solution. A very important distinction I think.

The Stockport study (the author of which I owe a huge debt, in more ways than one) should have the last word here I think:
"Finally it should be remembered that whilst we have looked at the education community we have ignored the rest of local government and services such as health and blue light services. It is obvious that all of these could be carried on the same network needing only MPLS clouds to segregate traffic into private VLANs as needed. So, the investment can be spread ever more widely with associated savings."
What better fit with the opportunity to re-use existing public infrastructure could there be?

No comments:

Post a Comment