NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.
Forum Discussion
Solved
CM3000 on Xfinity with 1200Mbit, onlt getting around 940Mbit with link aggregation
I just purchased a CM3000 for use with my Xfinity 1200mbit plan. It's currently running firmware 6.01.04. It was set up in link aggregation mode and connected to my router via ethernet ports 1 and 2....
The Ookla speedtest, both app and in-browser is multi-connection, so shouldn't this be multiple data flows? I've seen other's Ookla results with link aggregation and it's well over 1000Mbit, so is there something wrong or is the Ookla speedtest not truly multi-connection?
It depends on the load balancing algorithm used by LACP. Layer 3+4 might separate out the connections, layer 2 would not. I don't know what the CM3000 does, but generally devices that support layer 3+4 let you configure the load balancing. Devices that don't let you configure it are using layer 2.
Another factor (specific to speed tests) is that they aren't directly measuring throughput. They are looking at packet delivery timing.
You could try (briefly) hooking the Mac up directly to the CM3000 and testing that.
No, I use link aggregation enabled on the CM3000 with LACP active mode on the router. Connection to the Mac is though LAG using the Mac ethernet port and a USB ethernet adapter plugged into a Cisco switch in active LACP mode to another switch, also the same, with a active LACP configuration to the router. So, this should work.
No, I use link aggregation enabled on the CM3000 with LACP active mode on the router. Connection to the Mac is though LAG using the Mac ethernet port and a USB ethernet adapter plugged into a Cisco switch in active LACP mode to another switch, also the same, with a active LACP configuration to the router. So, this should work.
See my reply.
LACP won't deliver more than 1 gbps per dataflow. Each flow is mapped independently by each transmitter on the path to one of the two downstream NICs. That eliminates out-of-order delivery. Speedtest only has one dataflow, so you won't see more than 1 gbps to the mac. LACP active mode doesn't change this behavior.
Generally this works best with a single client when you can use layer 3+4 load balancing on all hops. But I don't see that as an option with the CM3000. Layer 2 load balancing will treat all the traffic going inbound as a single flow, since it only looks at the source and destination MAC addresses when identifying a data flow. That will be mapped onto one NIC by the CM3000, so it is limited to 1 gbps. But even with layer 3+4 load balancing, speedtest still would be treated as one flow, since the packets all have the same IP address and ports.
The easiest way to get full speed is to use multigig ethernet instead of a LAG.
The Ookla speedtest, both app and in-browser is multi-connection, so shouldn't this be multiple data flows? I've seen other's Ookla results with link aggregation and it's well over 1000Mbit, so is there something wrong or is the Ookla speedtest not truly multi-connection?
As a test, Do you get 1200Mbps directly connected behind the CM modem with out the router being connected and LAG disabled on the CM modem?
If you have a 2.5Gb or higher capable ethernet PC or Ethernet adapters, you should see what your pay for from your ISP from the CM modem with out the switches in line.
This would be first thing to check and test.
The Ookla speedtest, both app and in-browser is multi-connection, so shouldn't this be multiple data flows? I've seen other's Ookla results with link aggregation and it's well over 1000Mbit, so is there something wrong or is the Ookla speedtest not truly multi-connection?
It depends on the load balancing algorithm used by LACP. Layer 3+4 might separate out the connections, layer 2 would not. I don't know what the CM3000 does, but generally devices that support layer 3+4 let you configure the load balancing. Devices that don't let you configure it are using layer 2.
Another factor (specific to speed tests) is that they aren't directly measuring throughput. They are looking at packet delivery timing.
You could try (briefly) hooking the Mac up directly to the CM3000 and testing that.
The router is a Cisco 8200 and it's LAN link aggregation connects to a Cisco 2960-X and 2960CX switches, both via link aggregation, then to my Mac through a LAG interface. Both the router (WAN and LAN) and both switches use the default options of LACP active mode. These are modern switches and a modern Cisco enterprise router. VLANs are not used and neither switches are trunked together. I'm not sure what is going on, but I don't see how this configuration would not work for parallel flows in this case. Any Cisco-related options for LACP active mode?
I will be getting a 2.5G USB3 adapter shortly just to verify modem provisioning.
I was able to do some testing and I can indeed verify it is the load balancing algorithm. The CM3000 directly connected to my Mac with link aggregation results in ~1300mbit on Ookla speedtest. Thank you for pointing me in the right direction! I appreciate your help.
In summary, it was improper load balancing through my chain, starting with the 8200 router. That's why I'm getting single flows, or at least it appears that way.
Again, thank you!
