SleepingCellDetector R-APP is a Java SpringBoot application for reading Aggregated PM Metrics from DataCollector R-APP and make sleeping cell prediction based on PM data
SleepingCellDetector R-APP needs several parameters to be defined before start.
TCA Algorithm configuration situated in resources/tca.json file, example:
[ { "name": "latency", "condition": "MORE_OR_EQUAL", "averageThresholdValue": 400, "latestThresholdValue": 150, "latestSize": 2 }, { "name": "throughput", "condition": "LESS_OR_EQUAL", "averageThresholdValue": 10, "latestThresholdValue": 10, "latestSize": 2 } ]
File contains information about names of performance measurement parameters, conditions and values of thresholds for them. "averageThresholdValue"- is a threshold of performance signal average "latestThresholdValue" - is a threshold of last slots of performance signal (number defined in " latestThresholdValue"). Needed to detect correction of performance signal. Conditions available: "LESS", "LESS_OR_EQUAL" , "EQUAL", "MORE_OR_EQUAL", "MORE"'
Actually DataCollector R-APP returns Aggregated Metrics of "latency" and "throughput" parameters, example:
{ "pm": [ { "cellId": "Cell1", "performance": [ { "latency": 50, "throughput": 80 }, { "latency": 50, "throughput": 80 }, { "latency": 50, "throughput": 80 } ] } ], "itemsLength": 1 }
Set parameters of environment variables:
A1PolicyManagementService URL can be set using environment variables:
To customize DataCollector R-APP connectivity you need to set the following:
SleepingCellDetector R-APP configuration fields:
Prefix of high priority user equipment (policy instances will be created only for user equipments with this prefix), example: "emergency_" - policy instances will be created only for UEs with "emergency_" prefix Slot of time in seconds, number of slots for DataCollector R-APP Aggregated Metrics endpoint call. Slot: aggregation period (in seconds) for which an average performance metrics are calculated Count: number of aggregated performance metrics that should be returned by the method, one aggregated performance metric per each slot. The first performance metrics is average metrics for (startTime, startTime +slot). StartTime for DataCollector R-APP Aggregated Metrics endpoint call is generated based on slot and count parameters as "time.now - slot*count"
Example configuration in environment variables in application.yml:
server:
port: 8382
a1:
host: "policy-agent"
port: 8081
dc:
host: "localhost"
port: 8087
version: "v1"
scd:
prefix: "emergency"
slot: 10
count: 12
logging:
level:
org:
springframework: DEBUG
logging.file.name: logs/rapp-sleepingcelldetector.log
pattern:
console: "%d %-5level %logger : %msg%n"
file: "%d %-5level [%thread] %logger : %msg%n"
During start-up, SleepingCellDetector R-APP registers itself as a service in A1PolicyManagmentService(PMS). After that SleepingCellDetector R-APP sends periodic keepalive requests to PMS. PMS exposes the Ric Configuration for SleepingCellDetector R-APP. Ric Configuration contains information about Policy Type and Ric name, example:
[ { "ricName": "ric1", "managedElementIds": [ ], "policyTypes": [ "1000" ], "state": "AVAILABLE" } ]
SleepingCellDetector R-APP uses this information with created UUID are used for Policy Instances creation request to A1PolicyManagementService, example:
{a1policymanagementservicehost}/policy?id=123e4567-e89b-12d3-a456-426614174000&ric=ric1&service=rapp-sleepingcelldetector&type=1000
{ "scope": { "ueId": "emergency_samsung_s10_01" }, "resources": [ { "cellIdList": [ "Cell3" ], "preference": "AVOID" } ] }
SleepingCellDetector R-APP creates the policy resources, which are used to traffic steering, selecting cell for a connection, or scheduling traffic on available cells, in a different way than what would be done through default behaviour.
Like was presented on the example above, created policy instances provided information which cells (mention in the array "cellIdList") the User Equipment (define in the section "scope", key "ueId"") should avoid (section "resources", key "preference").
After start-up SleepingCellDetector R-APP begins to make predictions periodically in intervals defined by CSD_SLOT period based on new PM metrics measurement data received in each interval. When sleeping cell is detected, creation of A1 Policy instance is enforced by SleepingCellDetector R-APP. If cell becomes active again, A1Policy instance deletion request is sent.
Following mvn command (in the current directory) will build SleepingCellDetector R-APP:
mvn clean install
To build docker image add -P docker:build flag.
Following command will run SleepingCellDetector R-APP:
java -jar sleepingcelldetector-0.0.1-SNAPSHOT.jar org.onap.rapp.sleepingcelldetector.SleepingCellDetectorApplication
The log file will be created in the /log path. Parameters of logging are in application.yml file. After SleepingCellDetector R-APP starts successfully, log/rapp-sleepingcelldetector.log should start to contain the logs:
.
└──log
└── rapp-sleepingcelldetector.log