opinion This week, Exetel chief executive John Linton made the audacious claim that all ISPs reselling National Broadband Network services would deliver the exact same performance to customers. However, I believe the claim to be broadly wrong – and in this article I’ll attempt to demonstrate why.
According to Linton, all ISPs using the NBN will perform the same because (I’m paraphrasing a little here) because everyone will be using several key elements of the NBN infrastructure. Specifically, he mentions the various links between a customer’s house (or business premise) and the hand-off point to an ISP’s own network (NBN Co’s much discussed Points of Interconnect).
The Exetel chief’s argument is that there is no way for an ISP to influence the performance of this highly important network segment, short of underprovisioning bandwidth to the Connectivity Virtual Circuit component which supplies bandwidth to each customers’ connection. He also appears to state that it might be possible to underprovision the entire bandwidth made available from an ISP’s network to that of the NBN, but writes that off as something “which only the truly paranoid would consider possible”.
Now to a certain extent, Linton’s right in his statements here. I would suggest that most ISPs, with their limited experience provisioning customers on the NBN so far, will attempt to deliver more than adequate CVC bandwidth to their customers, to ensure that the new NBN connections deliver on their promises. We know from Internode’s experiences in Tasmania that each CVC connection to an NBN Point of Interconnect should be at least 200Mbps to satisfy customer demand, and at least in the short to mid-term, we can expect other ISPs to follow that guideline.
However, I also believe that Linton is ignoring both principles of dynamic traffic management here with respect to that aspect of ISPs’ networks, and also other segments of their networks.
Internode chief Simon Hackett wrote recently that once deployed, 200Mbps of CVC connection could actually support “quite a lot of customers” in a Point of Interconnect geographical area (of which there will be 121 located around Australia). However, he also made clear that ISPs could choose different contention ratios, and would each have their own policy in terms of the desired ratio of customers to CVC megabits.
“Once deployed, 200 megabits on a CVC can actually support quite a lot of customers, based on whatever contention ratio an RSP chooses, until the RSP ultimately upgrades beyond 200 megabits in accordance with its individual policy in terms of customer POI access point contention ratio,” Hackett said.
What this implies is that – just as with the current ADSL2+ networks – ISPs will be able to dynamically monitor their broadband networks under the NBN and provision supporting bandwidth up and down for certain regions as they need it. Cut-rate ISPs will obviously want to get away with as poor contention rations as possible, while higher-priced ISPs will want to deliver a standard quality experience, so will set up their networks for better ratios.
Secondly, I think Linton has ignored the reality of how ISP networks work. Let’s take a look at this picture of an ISP’s internal network on the NBN, supplied by Internode networking engineer Mark Newton:
The picture starkly displays the fact that the NBN is what it is advertised to be – primarily an access network which connects the core networks of ISPs to customers’ premises. The ISPs then go through a complicated dance to connect their own networks to the broader Internet. In an NBN world, according to Newton’s diagram, customer traffic will flow from customers’ premises to a fibre access node, and then to one of the 121 Points of Interconnect. From there it will enter their ISP’s network through other fibre networks and the ISP’s back-haul switching setup.
I don’t understand all the details in as high a level as I would like (being a former systems administrator rather than a network administrator), but at that point it looks like the traffic goes through what is termed a broadband network gateway (BNG), a broadband remote access server (BRAS) and a L2TP network server (LNS) – with this being a point on the ISP’s network where traffic is aggregated and customer’s tunneled connections are terminated as the ISP makes ready to send traffic on to the broader Internet. Perhaps someone else can clarify in the comments what impact is seen here on network performance.
After that comes an ISP’s core network, and then its edge network bordering on the broader Internet, which is also where it may peer traffic with other ISPs, such as with PIPE Networks’ famous Internet Exchanges located around Australia.
So you can see, from the diagram above, there are many places where an ISP may skimp on the services it provides to customers. Even though using a wholesaler like the NBN takes a fair degree of control away from ISPs in some regions of the entire network they are using, they still have enough control to influence outcomes – and it feels to me as if they have more control in some senses – or at least the same — than they would have over a similar ADSL network.
And this isn’t even taking into account the connections which an ISP has to the rest of the world — it’s just their own internal network. I’m sure some Australian ISPs have many redundant links to other ISPs and internationally — while some get by with less, and focus on the cheaper options.
In addition, before people start jumping up and down about how all of this is different under the NBN, please check out the pictures Newton has also supplied about the network topography of an ISP using a wholesale ADSL network or its own ADSL network. You’ll notice that much of the diagram is exactly the same — for good reason. Again, the NBN is an ISP access network.
Another senior ISP industry representative we asked about the issue this week had this to say:
“The extent to which a given RSP adequately provisions the size of a CVC into the network in each service area of the NBN is only one of many constraints on performance for an ISP.
To frame that as the only, or even the most important, differentiator between providers is to completely ignore the other 99 percent of the data path between a customer and their chosen sources of Internet hosted information. The rest of the national and international data path in use, and the extent to which those are redundant, adequately provisioned in terms of capacity, and served by adequately high performance routing equipment are all at least as important.
The NBN is going to highlight the difference in quality between RSPs far more clearly than is possible with ADSL2+; We’ll see ‘unlimited’ (and/or very low cost/high quota) providers necessarily overselling their upstream Internet capacity in order to make ends meet. They do this today, but its mostly masked by the lower line speeds of ADSL2+ services. In the NBN realm, on 100 megabit services, there will be nowhere to hide for a cheap ISP during the busy hour.
Framing the NBN as something which magically makes the rest of all ISP networks exactly the same as each other is just nonsensical.”
We couldn’t agree more.
So what will this mean for end customers? A lot. In an NBN world, as we’re already seeing being played out through the release of early pricing plans, there will be a great deal of differentiation between ISPs when it comes to the quality of the service they are providing. Just as with ADSL, some providers will gain a reputation for strong and fast networks, while others in comparison will go the low-quality path and attract customers of a different kind.
No doubt business-focused NBN providers will also spring up and provide a whole host of more complicated and even higher quality services over the network. And of course value-add services such as content, ISP servers (for example, the gaming networks operated by BigPond, Internode and iiNet), bundles with mobile services and more will also provide reasons why you would choose one ISP over the other.
The breadth of that competition is still unclear; with debate still existing about the extent to which the NBN will support new national providers with its current pricing model; however, we can’t see any of the current half-dozen major players going out of business any time soon, and we’re sure there are quite a few others in the next tier down which will make a strong NBN play.
Overall, there is strong reason to believe ISPs’ networks will not all perform the same in an NBN world. ISP networking in 2011 is complex; and while we have no doubt not been able to truly do it justice in this brief examination (keep in mind that I’m a journalist and not a network admin), there is no doubt that the situation is a great deal more complex than Exetel chief John Linton’s comments would suggest.