Hello all! I just started using splynx for my bandwidth monitoring about 2 weeks ago. Prior to that I was shaping on the CPE’s, as I hadn’t had time to work on setting up the basics to let Splynx do the shaping.
I now have it setup, and have several questions/issues. But first some background:
I have an OSPF routed network. I have a router at my edge, that connects to my fiber circuit. It does IPV4 Nat. All internal addresses are private (10.1.x.x addressing). I have a CCR1016 at my core. The tower routers are RB2011’s. I’m using static-dhcp with radius to Splynx to assign IP addresses at each tower router. I’m using Mikrotik API to do shaping. My core router (The CCR1016) has an interface that servers some customers on the main tower, but most of the data is coming in from other towers with RB2011’s on them.
I’ve noticed that the traffic accounting is not at all accurate. For instance, last night I had a customer that hooked a new XBox up. It’s been running at 25mbps non-stop for the last 12 hrs. Yet, in splynx it only shows 2.6 GB of traffic. Also, at all of the sites with RB2011’s, my customers show much larger traffic in the “Upload MB” field than “Download MB”. On the customers that are located right at the core, directly connected to the CCR1016, the traffic shown on the accounting tab seems to be correct (more download traffic than upload)
Also, I initially only wanted my RB2011’s to handle DHCP and routing, since they don’t have very powerful CPU. I wanted to do all shaping for all customers on the CCR at the core. However, when I went into the router preferences for my tower routers and chose the Shaper as being my core router, I found that every customer that I set up created 2 queues, one on the CCR and one on the RB2011, instead of just on the CCR.
I’m wondering if I need to plan on just doing PPPOE tunneling back to my core and having all customers connected directly to the core router. (Use VPLS tunnels). Would this solve the issue? I could then use Radius AAA accounting. I was trying to avoid this for network overhead reasons.