Performance monitoring at edge of communication networks using hybrid multi-granular computation with learning feedback
Example implementations disclosed herein are directed to monitoring communications networks in real time to ensure that they are performing in a desired fashion. Metrics such as utilization and latency are monitored to ensure that customer SLAs (Service Level Agreements) are met in a timely fashion. To detect problems in advance, example implementations predict the future values of metrics and detect events if the future value violates certain conditions. Example implementations build models that can predict the future values of metrics and analyze historical, near real time and real time streaming data so as to build predictive models.
1. A management server comprising:
a memory configured to manage a first prediction model of traffic to a first type of storage configured to store data for a first time period in a storage system, a second prediction model of traffic to a second type of storage configured to store data for a second time period in the storage system, and a third prediction model of traffic to a third type of storage configured to store data for a third time period in the storage system, wherein the second time period is longer than the first time period, and wherein the third time period is longer than the second time period;
a processor, configured to:
compute a prediction model of traffic to the storage system based on the application of weights to the first prediction model, the second prediction model, and the third prediction model; and
update the application of weights based on traffic results;
wherein the first type of storage is a memory buffer type storage and is different from the second type of storage, and wherein the third type of storage is a dedicated storage for the storage system;
wherein the memory is further configured to store one or more time intervals;
wherein the first type of storage is configured to move the stored data from the first type of storage to the second type of storage according to one of the one or more time intervals;
wherein the second type storage is configured to move the stored data from the second type of storage to the third type of storage according to the one or another one of the one or more time intervals.
2. The management computer of claim 1 , wherein the processor is configured to:
incorporate recent stream data to the first storage type in the first prediction model;
incorporate the stored data moved from the first type of storage to the second type of storage in the second prediction model; and
incorporate the stored data moved from the second storage type to the third storage type in the third prediction model.
3. The management server of claim 2 , wherein the processor is configured to:
update the first prediction model based on the recent stream data;
for data not moved from the second type of storage to the third type of storage, update the second prediction model based on the stored data moved from the first type of storage to the second type of storage, and update the third prediction model based on the stored data moved from the second type of storage to the third type of storage;
for data moved from the second type of storage to the third type of storage, construct a new model on the stored data of the second type of storage that was moved from the first type of storage, and update the third prediction model based on the stored data moved to the third type of storage.
4. The management server of claim 1 , wherein the processor is configured to compute the prediction model of traffic to the storage system and update the application of weights in response to a detected event.
5. The management server of claim 1 , wherein the processor is configured to obtain an overall prediction of the storage system from the prediction model based on applying the weights to the first prediction model, the second prediction model, and the third prediction model.
6. A method, comprising:
managing a first prediction model of traffic to a first type of storage configured to store data for a first time period in a storage system, a second prediction model of traffic to a second type of storage configured to store data for a second time period in the storage system, and a third prediction model of traffic to a third type of storage configured to store data for a third time period in the storage system, wherein the second time period is longer than the first time period, and wherein the third time period is longer than the second time period;
computing a prediction model of traffic to the storage system based on the application of weights to the first prediction model, the second prediction model, and the third prediction model; and
updating the application of weights based on traffic results;
wherein the first type of storage is a memory buffer type storage and is different from the second type of storage, and wherein the third type of storage is a dedicated storage for the storage system;
wherein the memory is further configured to store one or more time intervals;
wherein the first type of storage is configured to move the stored data from the first type of storage to the second type of storage according to one of the one or more time intervals;
wherein the second type storage is configured to move the stored data from the second type of storage to the third type of storage according to the one or another one of the one or more time intervals.
7. The method of claim 6 , further comprising:
incorporating recent stream data to the first storage type in the first prediction model;
incorporating the stored data moved from the first type of storage to the second type of storage in the second prediction model; and
incorporating the stored data moved from the second storage type to the third storage type in the third prediction model.
8. The method of claim 7 , further comprising:
updating the first prediction model based on the recent stream data;
for data not moved from the second type of storage to the third type of storage, updating the second prediction model based on the stored data moved from the first type of storage to the second type of storage, and updating the third prediction model based on the stored data moved from the second type of storage to the third type of storage;
for data moved from the second type of storage to the third type of storage, constructing a new model on the stored data of the second type of storage that was moved from the first type of storage, and updating the third prediction model based on the stored data moved to the third type of storage.
9. The method of claim 6 , wherein the computing the prediction model of traffic to the storage system and updating the application of weights is conducted in a continuous fashion.
10. The method of claim 6 , further comprising obtaining an overall prediction of the storage system from the prediction model based on applying the weights to the first prediction model, the second prediction model, and the third prediction model.
11. A non-transitory computer readable medium, storing instructions for executing a process, the instructions comprising:
managing a first prediction model of traffic to a first type of storage configured to store data for a first time period in a storage system, a second prediction model of traffic to a second type of storage configured to store data for a second time period in the storage system, and a third prediction model of traffic to a third type of storage configured to store data for a third time period in the storage system, wherein the second time period is longer than the first time period, and wherein the third time period is longer than the second time period;
computing a prediction model of traffic to the storage system based on the application of weights to the first prediction model, the second prediction model, and the third prediction model; and
updating the application of weights based on traffic results;
wherein the first type of storage is a memory buffer type storage and is different from the second type of storage, and wherein the third type of storage is a dedicated storage for the storage system;
wherein the memory is further configured to store one or more time intervals;
wherein the first type of storage is configured to move the stored data from the first type of storage to the second type of storage according to one of the one or more time intervals;
wherein the second type storage is configured to move the stored data from the second type of storage to the third type of storage according to the one or another one of the one or more time intervals.