A concept on Port Utilization (a.k.a Throughput) , Incoming Pkt Burst and port egress Line rate –

Sharing for light reading .. :)

?

*******************************

*******************************

?

A concept on Port Utilization (a.k.a Throughput) , Incoming Pkt Burst and port egress Line rate –

?

The concept below is drawn to explain you how NE in general (MPLS R, SW, BTS, MW, ++) measures the port utilization and what the difference b/w port utilization counters and Port Line Rate ..

?

Let me take an example of one MW link with LAN and WAN ports (IDU and full outdoor radio).. we see port utilization @ Radio out i/f, but what it signify or how it calculates the throughput? It measures the total number of bits every sec and divides that by 1 sec .. so this gives us rate i.e. bps/Kbps/Mbps .. basically it is measuring the total number of bits that are coming/accumulating at o/p i/f of radio per sec .. this accumulation will depend upon the system buffer (bucket size), burst of incoming traffic (volume of bytes or bits coming at any instance ‘t’) .. so if volume is small i.e. burst received at radio o/p i/f is small (could be bec user traffic volume received from MPLS itself is very small, ex 5 ping pkts of 64 Bytes only passed o the MW link during 1 sec or burst absorbed by huge IDU & Radio buffer before burst reached radio’s out i/f, ++), then radio throughput report will not show you high throughput .. example if only 5 x 64B ping pkts were passed in 1sec and all of them reaches radio egress port in 1sec then, it will display (5*64*8/1s) = 2.56Kbps as a measured throughput for that second .. and if no traffic is passed, utilization report will measure 0 Mbps ..

?

We have to understand what we are referring (i.e. radio throughput counter) , how system is measuring or performing the calculation and the variables involved in the calc ?- 1) Data volume/burst/buffer/total bits received @ o/p i/f of radio and 2) time delta ‘t' over which measurements are done i.e. 1sec ..

?

Now, you might have a question, traffic goes out @ line rate only, right? for example, if MPLS Router egress i/f (connected to MW IDU @ hub) is shaped @ 32Mbps or say MW radio link (WAN i/f) is configured for 330Mbps, then why we don’t see consistent 32 Mbps or 330Mbps in MW throughput reports (i.e. flat line of 32Mbps or 330Mbps for throughput vs time graph) ?

?

Understand, traffic always exits the i/f @ line rate only .. if i/f is shaped to 32Mbps or 1Gbps or 10Gbps or 100Mbps/330Mbps WAN speed, ++ traffic existing the i/f will follow corresponding line rate only .. and this doesn’t depend upon the size of burst or volume of traffic .. even if 5 x 64Bytes ping pkts are send from IDU1 towards IDU2 (connected w 1GE ODU-IDU cable and 330Mbps Radio WAN Speed i/f), then @ each respective out i/f (i.e. IDU 1GE i/f connected to radio or 330Mbps WAN i/f of radio), traffic will follow egress port line rate only ..

?

So, if this is correct, then why we don’t see always 330Mbps on radio out i/f or 1Gbps on IDU Egress i/f i.e. 100% LAN / WAN port utilization? Answer is that, we are not measuring or say NE is not measuring & sharing the line rate in your MW utilization reports .. the active line rate can always be checked under port config status (1Gbps or 100Mbps or 10Gbps or 330Mbps or 600Mbps, ++) ..

?

The utilization reports we see from Radio o/p port is based on the fact that how many bites or bytes are getting accumulated @ o/p i/f of that radio at any time t and reported in a 1sec measurement interval .. this information should not be correlated w i/f line rate (shaped or w/o shaping) .. It doesn’t give line rate information .. however line rate is one of the variables based on which pkt buffering , pkt scheduling & pkt accumulation on o/p i/f will take place ?& thus egress port utilization will be calculated .. Incoming burst (i.e. volume of incoming burst) ; NE internal Buffer size ; configured scheduling criteria SP, WRR; Egress port line rate (100Mbps, 330Mbps, 1Gbps, 10Gbps,++) are some of the QOS variables based on which pkt or bytes accumulation @ egress port will take place and thus egress port utilization will be calculated for measurement interval delta t (1sec in general but depends upon the NE)

?

In fact not only MW radio utilization reports, but all products that show bits in/out and throughput or port utilization reports are actually performing same or similar task .. i.e. total bits @ out i/f divided by measurement interval .. this measurement interval can be different for different nodes/vendors/products .. For MW NE in reference, it is 1sec ..