@Michel thanks, and understand your feedback well. I’ll try to discuss our overall approach from a networking perspective, as there is a bit more method to this than it may appear.
Fundamentally with the speeds that we’re now achieving, and quickly moving to 1 Gbps subscriber speeds in 2017, to achieve line-speed L3 and above networking features, the chip architectures have moved towards needing companion processing to efficiently perform the networking features you’re talking about.
The tradeoff here was to either:
(a) add additional processing at significant cost outdoors, which of course will be in duplication to a Ethernet connected router in the home for Wi-Fi and home networking, or
(b) add a number of features to handle much of the controls you need within the A5 Access Point and through more advanced provisioning features.
We’ve seen important advantages of adding subscriber and application traffic intelligence directly in the A5, which was designed for this task with a Gigabit+ Network Processor, so that we’re able to individually manage subscriber impact to Airtime and QoS, giving significantly better control and network/air health versus tunneled PPP which eliminates critical visibility at that central AP point, or a simple routing client.
With this tradeoff in mind, we chose to keep the C5 as a cost effective managed bridge, and extend a large number of features into the A5 to support a high quality Layer 1/2 network, as well as to automate provisioning of in-home routing (provided by the WISP OR managing 3rd party routers cleanly).
So for example, as of 2.0.2 coming in the next week, in addition to client isolation, multicast/broadcast and unicast flood prevention, we’ve also added DHCP rogue server prevention into the A5. So this will prevent ill-connected routers from impacting upstream and the wireless network.
Furthermore for VLAN settings, we are adding those into the A5 for applying VLAN settings. Obviously all of this was not completed in the first release obviously.
I assure you this is not due to lack of effort, for example we implemented some very sophisticated router functions as you may have seen with the new Mimosa G2 announced recently. But simply putting routing features into the outdoor CPE for edge control was going to be a cost addition that still required cost in-home, and also was not going to give us the degree of control we want to manage airtime quality at the A5 network layer.
Regarding the L3 termination (since the C5 will remain as a bridge), if you prefer NOT to add a router of your own (MikroTik, Linksys or Mimosa G2, etc.) for cost reasons, the best solution is to control the customer 3rd party router which typically is unknown MAC address, etc. which can be a bit of a pain of course to setup static IP and manage in scale. To make this seamless to manage, we are adding Option 82 on the A5 to protect the network DHCP server, and properly serve the correct subscriber IP address to the customer router by injecting the known MAC of the C5 into the DHCP relay request.
This of course assumes the C5 MAC/customer relationship has been associated and information available in the DHCP server tables or via Radius (Radius is the typical Option 82 lookup solution with DHCP servers). We have a lot of customers waiting on both Option 82 and Radius direct implementations in the A5 that we are adding support for.
We understand that some WISPs will not prefer this approach as a controlled L3 termination made their lives easiest. Also comparing it purely to a retail Ethernet/Wi-Fi router is also not really a good comparison as it’s not part of a comprehensive wireless ecosystem concerned with balancing the health of the subscribers across the AP. I assure you the technique we’re taking on system integrating the subscriber provisioning into the A5 and end-to-end airtime optimization approach is feedback we’ve received from the larger scaling ISPs as their more advanced preference driving to these much higher speeds.
For some things, like flexible channel size/bandwidths, we have more limited control in this generation of technology due to the chip, those are certainly things we’re pushing to improve in the next generation chip as we get opportunities to add features into this new line of chip technology.
Hope this helps explain our approach, it actually is quite a bit more advanced in the kinds of subscriber and traffic controls we can offer long term. Obviously we’re not entirely there yet with each piece of functionality so the early days are more manual and challenging for sure, but we’re very committed to continually adding features to make this very easy to deploy and make it well-tied in with larger scaling ISP operations and practices.
