Tuesday, 11 January 2011

wholesale dilemma (and solution)

Who wants wholesale access to our community fibre networks?
  1. government want to see wholesale access offered if there's public money spent on their construction
  2. ISPs believe they have a right to offer services on networks where there is a monopoly operator (irresepective of who paid for it)
  3. consumers want a "choice" of ISPs
Sounds reasonable but the downside is that to engineer wholesale access can result in an inferior technical solution.  The standard wholesale architecture sees IP packets to and from a subscriber "tunnelled" to the serving ISP which can be quite some distance away.  Packets can't be processed (routed) at the Digital Village Pump.

Consider this scenario:

A rural area installs their own fiber access network, using the "home run" model to a Digital Village Pump.  This gives gigabit speed access between the home and the DVP.

For political reasons the network is opened for wholesale access delivery of Internet services, so each subscriber goes to the open market and signs up with any of the ISPs that have signed up on the wholesale offer, modelled on BT's Wholesale Broadband service of tunneled sessions to a managed interconnect.

Most ISPs have their core interconnects in London.

The local council wants to have interactive video access to their open meetings, for all the right social reasons.

But when it comes to it, they find that instead of the Gigabit performance the access network promised, people are getting relatively limited speeds, blocky and stuttering video.  traceroute shows the video traffic leaves the town hall and travels over the council's ISP to London on oversubscribed national backbone links.  There it crosses an interconnect to the viewer's ISP, then travels back up the country over more sub-rate backbone links to the viewer.

Everyone is disappointed with the resulting poor-quality experience.

What's gone wrong here?

The big problem with wholesale is the trunking of traffic out of the community.  Routing, like jobs or shopping, is best kept local to avoid commuting.

So what are the options for implementing wholesale access in the community fibre access Ethernet network?  Here are some I've considered:
  1. routed IP

    concept: the local operator runs a routed network from the subscriber to the Internet, including the minimum services required for functional Internet access (ie DNS) but not any extra services (email, web hosting etc.)  This is simple Ethernet over Fibre To The Home connected to Internet transit.  This option is modeled on BT Wholesale's old CENTRAL PLUS service.  The ISP (the wholesale customer) is responsible for billing the subscriber and adding email, etc. via their own servers on their own network elsewhere on the Internet.

    cpe: customer's choice on the open market
    pros: easy to implement and support (and therefore cheap)
    pros: most optimal routing solution - supports local routing between subscribers, between local networks (nearby DVPs) and local Internet users (content providers, other ISPs etc.)
    cons: not to many ISP's liking as they don't have control of the addressing, routing etc.

  2. dark fibre handed off at the DVP

    concept: bare "dark" fibre service, rent each subscriber's fibre connection to the ISP at the DVP
    cpe: ISP's choice of CPE
    pros: easy, no equipment needed by the local community operator
    pros: ISP is free to use any technology (including PON)
    cons: not attractive to most ISPs, requires they install kit, arrange their own backhaul
    cons: no local routing unless the ISP does it (which is unlikely)
    cons: BT don't offer dark fibre services for good reason, it would be like giving away the crown jewels, so should we?  it would be giving up all control on the community built resource

  3. PPPoE

    concept: turn the Ethernet link to the consumer into a PPP tunnel (this is how ADSL is curretly delivered to the wholesale market)
    cpe: ISP's choice of CPE
    pros: attractive to ISPs, looks like what they've currently got
    cons: needs more expensive DVP access control systems
    cons: even less likely to support local routing

  4. IEE802.1q VLAN trunking to the home

    concept: encapsulate subscriber's packets in dot1q VLAN and trunk to ISP (this is what BT offer on 21CN GEA)
    cpe: ISP's choice
    pros: cheap DVP hardware (most switches support dot1q)
    pros: opportunity to run second VLAN to support locally routed services (with enhanced CPE)
    cons: if ISP controls CPE, not much chance of second VLAN support

  5. MPLS

    concept: community network runs MPLS
    cpe: ISP provided router
    pros: the technology de joure, popular with service providers
    cons: DVP switch becomes the MPLS Provider Edge (PE), the service is an MPLS L2VPN
    cons: no local routing option unless MPLS extended to the CPE (which is complex)

  6. L2TPv3

    concept: CPE based wholesale access, the CPE establishes the tunnel to ISP
    pros: simpler than MPLS - doesn't require complex switches at DVP
    pros: opportunity to route locally in parallel with running the wholesale tunnel from the CPE
    cons: more complex CPE configuration and management
    cons: not widely deployed as a wholesale solution

The challenge is to fit a wholesale model into a service where the community's own ISP is one of the wholesale ISPs - ie the community ISP has to buy services on a level playing field with other ISPs.

The bottom line is that wholesale can't be ignored and if done right can help clarify the division between network operation and service management.  So I propose the following model:

ISPs who want wholesale access to any DVP can choose from any of the services listed above.  We're very inclusive, we support all the options.

Option (1)  puts all the IP provisioning and routing in the hands of the DVP operator, and the wholesale ISP is simply responsible for billing and value add services.  Option (1) would be the service the local ISP uses.  This allows the community operator to focus on running the network while the ISP side takes care of customer relationships, billing and added value services like email, web hosting etc.

The other options come with some conditions:
  1. There is a minimum order unit of 20 connections of any one type  (i.e. the ISP has to buy in lots of 20.) This is similar to the commitment BT required in the "ADSL Exchange Activate" programme and reflects the engineering required to implement the solution and represents a demonstration of commitment from the ISP for the service to be viable.

  2. For backhaul, there is a choice:
    • a connection fee for a backhaul port if the ISP wants to bring their own fibre in, or
    • the connection fee plus a backhaul service which gives an amount of bandwidth to the nearest Internet peering location (depending on geography)

  3. Equipment can be hosted at the DVP for appropriate fee.  Peering with the community network is free at the DVP to encourage local routing

All services would be priced according to the level of effort required to design, implement and support the overall service (i.e. a fair share of the entire cost of the community access solution)  We can look to BT Wholesale's pricing for similar services for what the market considers reasonable charges.

It is likely that option (1) would be significantly cheaper than the other options, which is good as it would encourage the technically superior solution to be more widely adopted.

No comments:

Post a Comment