|
|
|
- Configure the network interfaces such that the GTP encapsulated
traffic runs on a different
port than the unencapsulated traffic.
- Ensure that all the ports
are on the same NIC card.
- For example, do NOT use port 2 for encapsulated traffic
and port 8 for unencapsulated traffic.
|
- Typically ~10k
UE’s will provide the best data throughput.
- As the number of
UE’s is increased to
the system maximum, the data throughput will gradually
decrease .
|
|
|
- Before running the test with 10k UE’s configured, try running
the test with just 1 UE.
- Capture a PCAP log
and see how large the data packets are. If there are any fragmented data packets
in the PCAP log, you will need to adjust the packet payload
size configured in the DMF to insure that the 1500
byte MTU size is not being violated.
- Verify that both encapsulated and unencapsulated packets
are 1500 bytes or
smaller.
|
- Once the above
steps have been completed you can run the test with the desired
# of UEs and observe the performance of the data throughput
by opening the L5-7
report tab and checking the “%
of Configured Rate Attempted”
OM.
- The goal is to
get this OM in the 95% range.
- This OM will tell
you how hard the test server is working and how much CPU processing
power is left.
- If this OM shows
100%, this is an indication that you are leaving performance
on the table and the Transaction Rate can be bumped up.
- This is a trial
and error process and might take several attempts to dial
in the best possible Transaction Rate.
|
- Check the L5-7
Client tab and L5-7
Server tab for
Basic
Data Messages Lost
and Basic
Data Duplicate Messages.
These OM’s should be 0 or very close to 0 and as the test
is running you do not want to see these OM’s increasing. In
a back-to-back test with Landslide Nodes, these OMs
should be 0.
- If you are running a connection
oriented transport layer such as TCP, you should also
check the L4 Client
and L4 Server tabs,
The “Socket Ack Timeouts”
and “Socket Lost Packets”
OMs should be 0 or very close to 0. In a back-to-back
test with Landslide Nodes, these OMs should be 0.
- When running this test in a real network some loss and
delays are possible thus these OMs might slowly increase as
the test runs, but should be a tiny fraction of the total
data throughput.
|
- The steps above are applicable to symmetric and asymmetric
data as well as advanced data such as HTTP.
|