The measurement metrics for the ATM layer include mean cell transfer delay (MCTD), 2-point cell delay variation (2-point CDV), cell loss ratio (CLR), cell misinsertion rate (CMR), cell error ratio (CER), cell throughput, and cell interarrival times.
The cell interarrival times measurement belongs to the traffic statistics collection category, while the rest of the metrics are for performance testing purposes.
The cell interarrival times will be recorded under the mesh topology only. The duration of the measurement might not be very long due to the high speed link and the lack of data storage capability.
Table 
summarizes the scenarios for performance testing in the
ATM layer.
Table: ATM layer Performance Testing Scenarios
The cell error ratio is defined as:
The cell error ratio (CER) performance is not dependent on testing topology. Instead, it depends heavily on the quality of the hardware that the testing bit stream traverses. For this reason, we chose to test the CER metric under the point-to-point configuration only. Also notice that the CER can only be measured by using an out-of-service technique, because the ATM protocol does not support error checking at the cell level.
We expect the mean cell transfer delay, cell delay variation, cell loss ratio and the cell misinsertion rate to be highly dependent on measurement topology. Therefore we chose to test these metrics under several topologies.
The Cell Loss Ratio is defined as:
where 
is the number of information cells transmitted and 
is
the number of information cells received.
The transmitter will
insert OAM cells into a user information stream at suitable intervals. Each OAM
cell will contain a count of the number of information cells transmitted since
the last OAM cell.
In this fashion, we are able to obtain the quantities 
and 
at
the receiver. This method undercounts CLR if misinsertion occurs during the
measurement period.
The Cell Misinsertion Rate is defined as:
where Tm is the duration of the experiment. This method underestimates CMR if cell loss events occur during the measurement period.
Again, different services supported on an ATM network have different attributes.
In Table we list the maximum acceptable
cell loss and misinsertion ratios for five types of services
(from [4]).
Table: Service Attributes for an ATM Network
Notice that the Cell Loss Ratio mentioned in Table
is equivalent to our CLR measurement. The Cell Misinsertion Rate in the
Table would be defined as:
and therefore can also be easily determined from our experiments.
The cell throughput will also depend on the measurement topology. We plan to test this metric under all topologies. Under ideal conditions, the throughput under different topologies should be all close to the link capacity.