This news item is to give our customers a bit more information about our recent and ongoing work we are doing with our LNSs.
We are making good progress in our project to both bring faster services to A&A customers and also to test prototype FB9000 hardware with 'real life' traffic. We have been gradually installing prototype hardware in our datacentres, and also significant development on the firmware is being done, to bring true multi-core CPU functionality to that hardware.
An LNS (L2TP Network Server) is a router that sits between our backhaul carriers (eg BT and TT) and the rest of the A&A network - when your router connects to the Internet it connects to the LNS and the LNS is the first 'hop' of your route to the rest of the Internet. It's our LNSs that take care of your PPP login, IP addresses, they produce the CQM graphs for your line and help manage prioritisation of traffic.
A&A are a little different to many other ISPs in that the LNSs we use are actually designed and built by us here in the UK. A&A is behind the FireBrick range of routers and these are devices that we have designed the circuit layout, PCBs, the physical casing, hardware drivers, operating system and application software. These are used extensively in the A&A network and also at other Internet providers.
We have been working on a new model, the FB9000. This again has been designed from the ground up, requiring a new hardware design, faster CPU, faster ports etc. This involves a lot of work, but since April 2022 we have been using them in our network to manage most of our faster FTTP services. We use over 20 of the older generation FireBrick FB6000 LNSs on our network, they provide services to the bulk of our customers, these are named 'Gormless'. We are using a pool of three FB9000 LNSs for our faster FTTP customers, these are named 'Witless'.
One of the major difference between the FB9000 and previous generations of FireBrick devices is that we are moving from running the operating system and software on single CPU cores to running on multiple CPU cores. Our previous generation of FB6000 devices had two CPU cores, and tasks were split between them - one core was in charge of management and background tasks whilst the other was responsible for heavy-lifting of all the network traffic. With the FB9000 we are using multiple cores for all tasks and traffic with the main outcome being higher throughput.
When we first started using FB9000 in April, the network traffic was still running on a single CPU core. Throughput was still much improved over our previous FB6000 generation, but there was more work to be done to make use of all the CPU cores available on the new hardware. Over the past few months a great amount of work has been done to update our code to run on multiple CPU cores and we are at a stage now (July 2022) where we can start introducing the new code to the LNSs we are using for our FTTP customers.
We are currently running three FB9000 'Witless' LNS; we have two in active use and one as a standby. The process of upgrading firmware on our LNSs involves upgrading the standby LNS and making it active. We then re-shuffle customers on to the new pair of active LNSs. This is done by scheduling overnight work to force customers off their LNS and when they reconnect a few moments later they will connect to a new LNS.
These upgrades and line moves are happening more often than usual but we will post the planned work on our status page.
Unfortunately, we have run in to a couple of problems recently which had impacted a small number of customers, we have revised our procedures to help if this sort of thing happens again. We will always have staff on duty when performing overnight line moves who can react quickly if there are problems.
We also have a small number of customers using our 'test' LNS (this is called B.Doubtless) - this LNS will be updated more often than the Witless set of LNSs.
In the course of continuing to improve the firmware we do expect to perform more updates but also minimise disruption to customers. The codebase is improving in leaps and bounds as work continues. In tandem with the codebase work, we are also now finalising the design and manufacturing aspects of the casework, and also acquiring the bill of materials for the first production run of retailable units. Global supply chain issues have hit the project hard, but drawing on a mix of design changes, brokers and patience, we are breaking through the bottlenecks.