Throughput Confusion – How To Measure Radio Capacity
Author Chris Gustaf, VP Engineering
When you’re out evaluating radio performance, understanding how manufacturers represent their throughput is very important. With continually increasing demands for capacity from bandwidth intensive devices and users, understanding what you are deploying, and how much room to scale you have is critical.
When comparing datasheets across manufacturer’s, confusion begins to take hold, making evaluating radio performance challenging. It’s important to understand the underlying criteria of all data points to properly evaluate radio performance.
Issues that are commonly debated are:
- Testing methodology
- Type of traffic tested (Layer 1 vs Layer 2 results)
- FDD vs TDD measurement numbers
How are the radios being tested?
RFC2544 is a standard test method for measuring capacity, latency and burst performance. It is widely used to obtain performance numbers for Layer 1(Physical), Layer 2(MAC), and Layer 3(IP) traffic.
Many dedicated products are available that can perform fully automated RFC2544 throughput tests at up to full line rate. These are preferred over software solutions running on standard PCs or Linux Hosts which may not be accurate depending on the loading of the CPU.
Tools such as Iperf are acceptable provided they are used on dedicated fast CPUs and are tested with no radio link first to ensure the capacity is high enough.
Understanding what you’re testing – Apples to Apples comparison
Some confusion between stated performance can be attributed to comparing Layer 1(L1) with Layer 2(L2) test results. L1 numbers will always look more impressive than L2 results because the preamble and Frame Check Sequence (FCS) are included in the calculation – this will especially be noticeable with small packet L2 test results since the preamble and FCS are a larger percentage of the overall packet.
At 64 byte L2 frame sizes, the L1 capacity of a full Gigabit system is 1 Gbps, but the corresponding L2 rate is only 761.9 Mbps. This is the physical limit of 1000BaseT or 1000BaseX fiber.
Layer 2 throughput is what most equipment manufacturers specify and what is generally accepted by service providers.
FDD vs TDD
Licensed microwave equipment is almost always Frequency Division Duplexed (FDD), meaning that upstream traffic is sent on a dedicated frequency and downstream traffic is sent on a SEPARATE dedicated frequency – two frequencies are used simultaneously.
Capacity numbers for FDD systems are traditionally provided as one-way capacities. For example, a stated capacity of 1 Gbps Full Duplex means that 1 Gbps can be sent upstream at the same time that 1 Gbps can be sent downstream.
For service providers familiar with Time Division Duplex (TDD) systems, this can create some confusion. TDD systems send upstream and downstream traffic on the same RF channel, allocating time slots for each direction. Typically TDD capacities are given as aggregate numbers which is the sum of upstream and downstream capacity.
To remove confusion, many microwave suppliers specify both the Full Duplex numbers and the aggregate numbers on their datasheets. For symmetric FDD microwave systems, the aggregate numbers are twice the stated FDD numbers.
When evaluating the capacity of a microwave consider the following:
1) Make sure the RFC2544 test durations are more than 1 second. Large switch buffers can absorb bursts of traffic and make the calculation of real throughput higher than reality for a continuous stream of traffic.
2) Compare apples to apples: Compare FDD capacity for Layer 2 frame sizes of 64 to 9600 bytes using RFC2544 test methods.
3) Make sure that all frame sizes are tested. Equipment that passes traffic through a CPU instead of using a dedicated switch will most likely have severe reduction in capacity for smaller packet sizes.