Printer FriendlyEmail Article Link

Spirent TestCenter: How do I determine the Best Method to test Latency (FIFO, FILO, LIFO and LILO) and Jitter?

Answer

 

There is difference between the method TestCenter measures latency in the basic test and the RFC2544 test. 

The default for Spirent TestCenter is FIFO and the default for the RFC2544 test is LILO.
 
How TestCenter calculates latency:
 
TestCenter has a signature with a timestamp to determine the latency between frames; The signature is at the end of the payload next to the FCS; The signature has a timestamp, sequence numbers and stream ID.
 
The default for Spirent TestCenter is FIFO. This is set in the Advanced section of the Generator and Analyzer and can be set independently.
 
Please reference the Spirent Customer Service Center found at http://support.spirentcom.com, and Login and Select -> Knowledge Base and search for SOL10521 that explains. 
  • On the generator side, if we define our timestamp reference location to be the start of the frame, then that means that we are marking the transmit time of the frame at the first bit the leaves the generator. After traversing the wire, this will be the first bit that goes into the DUT. So, generator “Start of Frame” timestamp equals FI (First In to the DUT). Similarly, if we define our generator timestamp reference location at the end of the frame, then that means that we are marking the transmit time of the frame at the last bit that leaves the generator. This would be the last bit that enters the DUT. So, generator “End of Frame” timestamp equals LI (Last In to the DUT) 
  • On the analyzer side, if we define our timestamp reference location to be at the start of the frame, then that means we are marking the receive time of the frame at the first bit that we receive. This would be the first bit out of the DUT. Analyzer “Start of Frame” = FO (First Out of the DUT)Similarly, if we define our analyzer timestamp reference location to be at the end of a frame, then that means we are marking our receive time of the frame at the last bit we receive. This would be the last bit out of the DUT. Analyzer “End of Frame” = LO (Last Out of the DUT)
 
Spirent TestCenter can do any combination of these since transmit and receive timestamp reference locations can be set independently. The default is start of frame for both generator and analyzer and thus FIFO is the default.
 

LILO is the BEST choice for measuring latency!  There are cases where a user might want a different method (and that’s why we support them), but in those cases the user should understand what the different latency modes tell you and pick the correct one for their situation.

LILO is the only latency mode that actually gives you what most users consider to be latency: the time it takes a packet to get from point A to point B, regardless of what media/devices/impairments/etc are between A and B.

 
Additionally, LILO is actually a better choice for testing most cut-through devices than FIFO.  Under congestion, a switch CAN’T possibly be a pure cut-through device, since congestion is forcing it to store frames in a buffer.  So under congested conditions frames could either be immediately forwarded like a cut-through device OR they could be placed in a buffer and forwarded later like a traditional store-and-forward device OR they get dropped, in which case you don’t have a latency value to worry about.  At least with LILO you will get consistent measurement results for every case.
 
Finally, if you’re testing multiple devices, e.g. STC <-> [DUT1] <-> [DUT2] <-> STC, then LILO is the ONLY option you should even consider, as both LIFO and FIFO are inappropriate in this case, especially if your DUT is changing media types somewhere along the path.
 
Now, If you are testing a single store-and-forward device, then LIFO is a reasonable choice to use, but again, LILO and LIFO tell you different things
 
LILO = The amount of time required for your frame to be traverse the network, e.g. to get from A to B.
LIFO = The amount of time the test packet sat in the DUT’s buffer. 
  • Latency results are viewed in the Results Browser -> Streams -> Detailed Stream results -> Basic Counters >> Short Term Average Latency, Average Latency, Min Latency, and Max Latency.
  • TestCenter Short Term Avg Latency is calculated over the last second and the Avg/Min/Max Latency is calculated over all received packets.
  • Reference attached Latency and Jitter Whitepaper

Product : Spirent TestCenter,L2-3,Windows GUI,RFC 2544,Scenarios