Storage area network queue depth profiler
View Patent ↗A system for improving performance of a network includes a monitor that obtains data sets corresponding to network performance, infers characteristics from the data set and generates visualizations and recommendations. In one aspect, the network is a storage area network and the system generates a box plot showing response time based on inferred queue depth, and recommending a queue depth setting if possible.
1. A computer-implemented method of improving performance characteristics of a data network, comprising:
receiving from a user a time window;
automatically monitoring network activity to obtain recorded data sets, wherein a first recorded data set comprises response times for storage requests corresponding to a first number of pending exchanges and wherein a second recorded data set comprises response times for storage requests corresponding to a second number of pending exchanges;
identifying, from the recorded data sets, a plurality of data sets associated with the time window received from the user;
determining inferred queue depth settings from a number of pending exchanges associated with the identified plurality of data sets;
generating a box plot of the identified plurality of data sets with the number of pending exchanges plotted according to the inferred queue depth settings;
performing a curve fitting analysis on the number of pending exchanges plotted in the box plot to select, from a plurality of queue depth profiles, a queue depth profile applicable to the identified plurality of data sets;
determining, by a processing device and based on the selected queue depth profile and the identified plurality of data sets, whether to recommend a setting for a queue depth different than the inferred queue depth setting; and
responsive to determining to recommend the setting, presenting a recommendation corresponding to the setting.
2. The computer-implemented method of claim 1 , wherein presenting the recommendation comprises providing a box plot visualization that shows the recommendation.
3. A system for improving performance characteristics of a data network, comprising non-transitory computer storage media storing programming instructions executable by at least one processor for:
receiving from a user a time window;
automatically monitoring network activity to obtain recorded data sets, wherein a first recorded data set comprises response times for storage requests corresponding to a first number of pending exchanges and wherein a second recorded data set comprises response times for storage requests corresponding to a second number of pending exchanges;
identifying, from the recorded data sets, a plurality of data sets associated with the time window received from the user;
determining inferred queue depth settings from a number of pending exchanges associated with the identified plurality of data sets;
generating a box plot of the identified plurality of data sets with the number of pending exchanges plotted according to the inferred queue depth settings;
performing a curve fitting analysis on the number of pending exchanges plotted in the box plot to select, from a plurality of queue depth profiles, a queue depth profile applicable to the identified plurality of data sets;
determining, by the at least one processor and based on the selected queue depth profile and the identified plurality of data sets, whether to recommend a setting for a queue depth different than the inferred queue depth setting; and
responsive to determining to recommend the setting, presenting a recommendation corresponding to the setting.
4. The system of claim 3 , wherein presenting the recommendation comprises providing a box plot visualization that shows the recommendation.
5. The computer-implemented method of claim 1 , wherein the selected queue depth profile comprises a queue depth versus response time pattern.
6. The system of claim 3 , wherein the selected queue depth profile comprises a queue depth versus response time pattern.