I am testing simple queues to tune/limit bandwidth on a network. The client has a dedicated router with bandwidth limits set at 1Mbps. My Mikrotik RB2011L/5.19.1 sees one IP address from this client.

I am using the RB2011L_ipaddress/accounting/ip.cgi feature to capture traffic levels. This afternoon I noticed that the client traffic level was showing at ~1.3Mbps with no queue enabled. I enabled a simple queue on this client address with 1M down/up, no burst settings - all other settings = default. At this point the Byte Graph within the queue page showed 1Mbps. However, the accounting page continued to show ~1.3Mbps.

I think that this discrepancy is somehow related to the type of traffic. In this particular case, the incoming stream was coming concurrently from 3-4 public IPs owned by LimeLight Networks. I have compared the accounting vs. queue traffic levels during other traffic flows and they match very closely.

Does the accounting traffic monitor capture the same information as the simple queue system? Why would these systems be reporting different stats? Which one is more accurate?

Thanks much!

P.S. i am capturing the accounting web page data every 8 seconds. Thus the bit rate corresponds directly to the byte count. I do not suspect a calculation error because the two systems match for most traffic.

P.S. Other people have reported strange issues on queues dealing with LimeLight Networks. Could this discrepancy be related to the way way that LimeLight splits up traffic across different IP source address streams?