inactive
369
edits
Performance Optimization Guide: Difference between revisions
m
→Analyzer queues
| Line 33: | Line 33: | ||
== Analyzer queues == | == Analyzer queues == | ||
The '''Allegro Network Multimeter''' has a packet queue between I/O threads and the analyzer threads which | The '''Allegro Network Multimeter''' has a packet queue between I/O threads and the analyzer threads which distribute the incoming packets to the actual processing threads. There are two statistics describing the load of the queues and analyzer threads. If analyzer threads cannot process incoming packets quickly enough, the corresponding queue will eventually overflow and packets must be skipped for processing. | ||
<ol> | <ol> | ||
| Line 40: | Line 40: | ||
Note that high counters for a few seconds at the initial startup of the device are normal when it is started under high network load scenarios. | Note that high counters for a few seconds at the initial startup of the device are normal when it is started under high network load scenarios. | ||
</li> | </li> | ||
<li>Utilization of individual analyzer threads: The load of the analyzer threads can be checked at the '''Info''' → '''System info''' → '''Load'''. Depending on the Allegro model | <li>Utilization of individual analyzer threads: The load of the analyzer threads can be checked at the '''Info''' → '''System info''' → '''Load'''. Depending on the Allegro model there can be two (Allegro 200, Allegro 500) or up to 120 (Allegro 5300 or 5500) analyzer threads. | ||
The load graph gives an indication about the overall utilization of each thread but the important counter is the '''Not processed due to overload''' counter since this is the event when ultimately one or more packets could not be processed due to overload. | The load graph gives an indication about the overall utilization of each thread but the important counter is the '''Not processed due to overload''' counter since this is the event when ultimately one or more packets could not be processed due to overload. | ||
</li> | </li> | ||
| Line 51: | Line 51: | ||
The '''Allegro Network Multimeter''' has reached its processing limit for current traffic when the load of one or multiple analyzers reaches 100%. | The '''Allegro Network Multimeter''' has reached its processing limit for current traffic when the load of one or multiple analyzers reaches 100%. | ||
There are a some options to reduce the analyzer load but | There are a some options to reduce the analyzer load but they come with the penalty of no longer seeing the entire measurement data. You can '''disable''' some features or add a '''NIC filter''' to process only parts of the traffic. | ||
<ol> | <ol> | ||
| Line 74: | Line 74: | ||
=== Skipped packets due to analyzer load imbalance === | === Skipped packets due to analyzer load imbalance === | ||
By default, the Allegro load balances the traffic between the analyzers based on the IP addresses of the client and server. This provides | By default, the Allegro load balances the traffic between the analyzers based on the IP addresses of the client and server. This provides good balancing in most situations. | ||
Network packets cannot distributed equally among all analyzer threads if there are many connections between only few IP addresses. An example is 2 SIP trunks with many RTP connections. | Network packets cannot distributed equally among all analyzer threads if there are many connections between only few IP addresses. An example is 2 SIP trunks with many RTP connections. | ||