MMPP analysis of network traffic using a transition window
View Patent ↗Data communication in network traffic is modeled in real time and is analyzed using a 2-state Markov modified Poissen process (MMPP). The traffic inter-arrival times for bursty and idle states define a transition window [λ 1 max , λ 2 min ] represented by the boundary values λ 1 max for the inter-arrival time for bursty traffic, and λ 2 min for the inter-arrival time for idle traffic. Changes in the values of λ 1 max and λ 2 min are tracked over time, and the size of the transition window is enlarged or decreased based upon relative changes in these values. If the inter-rival times for the bursty state and the idle state become approximately equal, the model defaults to a single state model. The modeling is applicable to the synchronization of polling and blocking in a low-latency network system. This permits the adoptive selection of poll or block to maximize CPU utilization and interrupt latency.
1. A computer implemented method for dynamically tracking and controlling the transfer of packets of data in computer network traffic using a 2-state Markov modulated Poisson process (MMPP) model, comprising the steps of:
a) measuring the inter arrival time between individual data packets;
b) determining whether the packet arrival rates are ‘bursty’ (λ 1 ) representing heavy traffic conditions or ‘idle’ (λ 2 ) representing light traffic conditions;
c) establishing a transition window with the values [λ 1 max , λ 2 min ] represented by the boundary values λ 1 max for the upper inter-arrival time for bursty traffic, and λ 2 min for the lower inter-arrival time for idle traffic;
d) utilizing the transition window values, determining the probability ρ 1 that a packet inter-arrival time is occurring in bursty traffic or the probability ρ 2 that the arrival time is occurring in idle traffic, and
e) based upon this determination, either blocking or polling the transfer of data.
2. The method according to claim 1 wherein the analysis of network traffic is carried out in real-time.
3. The method according to claim 1 wherein the bursty packet inter-arrival time has a mean value of λ 1 mean , the idle packet inter-arrival time has a mean value of λ 2 mean , and the traffic characteristics are such that λ 2 mean normally is substantially smaller than λ 1 mean .
4. The method according to claim 3 wherein as λ 2 mean approximates λ 1 mean , the tracking defaults to a single state MMPP model.
5. The method according to claim 1 further including the step of tracking changes in the values of λ 1 max and λ 2 min over time and enlarging or decreasing the size of the transition window based upon relative changes in these values.
6. The method according to claim 5 wherein the changes are tracked by computing the mean inter-arrival time of a plurality of packets.
7. The method according to claim 1 wherein the burst length during the bursty state is determined by the interrupt latency time wherein the model is used for the synchronization of a user process with the completion of the receive operation at the system to network interface.