software.amazon.kinesis.worker.metricstats

Class WorkerMetricStats

java.lang.Object

software.amazon.kinesis.worker.metricstats.WorkerMetricStats

public class WorkerMetricStats
extends Object

DataModel for a WorkerMetric, this data model is used to store the current state of a Worker in terms of relevant WorkerMetric(CPU, Memory, Network). workerId : unique worker identifier, this is equivalent to the owner attribute from the lease table. lastUpdateTime : wall epoch in seconds when the entry was last updated metricStats : Map of WorkerMetric to last N values for it. e.g. entry "CPU" : [10,20,12,10] etc operatingRange : Map of WorkerMetric to its operating range. First item in the list of values defines the max limit. metricStatsMap : runtime computed WorkerMetric name to its average value map. This field is not stored in ddb and is used during Lease assignment only

Constructor Summary

Constructors

Constructor and Description
WorkerMetricStats()

Method Summary

Modifier and Type	Method and Description
double	computePercentageToReachAverage(Map<String,Double> workerMetricToFleetLevelAverage) Determines percentage of load to reach the mean for the worker.
boolean	containsMetricStat(String workerMetricName) Returns true if given is available for the current worker else false
void	extrapolateMetricStatValuesForAddedLease(Map<String,Double> workerMetricToFleetLevelAverage, int averageLeaseCount) Increase the metric stat value corresponding to the added single lease.
void	extrapolateMetricStatValuesForAddedThroughput(Map<String,Double> workerMetricsToFleetLevelAverageMap, double averageThroughput, double increaseThroughput, double averageLeaseCount) Increase the WorkerMetricStats value by given increaseLoadPercentage.
static List<software.amazon.awssdk.services.dynamodb.model.AttributeDefinition>	getAttributeDefinitions()
static List<software.amazon.awssdk.services.dynamodb.model.KeySchemaElement>	getKeySchema()
double	getMetricStat(String workerMetricName) Returns the value for given WorkerMetricStats name.
boolean	isAnyWorkerMetricAboveAverageUtilizationOrOperatingRange(Map<String,Double> workerMetricToFleetLevelAverage)
boolean	isAnyWorkerMetricFailing() Returns true if any of the metric stat values has -1 in last index which represents that the metric stat value was not successfully fetched in last attempt by worker.
boolean	isUsingDefaultWorkerMetric() If a worker is not using an explicit WorkerMetric such as CPU, Memory, or Network, then it is said to be using the default WorkerMetric.
boolean	isValidWorkerMetric() WorkerMetricStats entry is invalid if any of the field from lastUpdateTime, operatingRange, resourcesStats are not present or if resourcesStats is empty or if any of the WorkerMetrics having resourceStats does not have operatingRange or if operating range values are not present or if maxUtilization is 0 for any WorkerMetric
boolean	isWorkerMetricAboveOperatingRange(String workerMetricName) Evaluates if the given metric stat is above operatingRange for the given WorkerMetric name.
boolean	willAnyMetricStatsGoAboveAverageUtilizationOrOperatingRange(Map<String,Double> workerMetricsToFleetLevelAverageMap, double averageThroughput, double increaseThroughput, double averageLeaseCount)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

WorkerMetricStats

public WorkerMetricStats()

Method Detail

containsMetricStat

public boolean containsMetricStat(String workerMetricName)

Returns true if given is available for the current worker else false

getMetricStat

public double getMetricStat(String workerMetricName)

Returns the value for given WorkerMetricStats name.

extrapolateMetricStatValuesForAddedThroughput

public void extrapolateMetricStatValuesForAddedThroughput(Map<String,Double> workerMetricsToFleetLevelAverageMap,
                                                          double averageThroughput,
                                                          double increaseThroughput,
                                                          double averageLeaseCount)

Increase the WorkerMetricStats value by given increaseLoadPercentage. This is done during execution of LAM and as assignments are happening the current metric stat value is increased based on increaseLoadPercentage.

willAnyMetricStatsGoAboveAverageUtilizationOrOperatingRange

public boolean willAnyMetricStatsGoAboveAverageUtilizationOrOperatingRange(Map<String,Double> workerMetricsToFleetLevelAverageMap,
                                                                           double averageThroughput,
                                                                           double increaseThroughput,
                                                                           double averageLeaseCount)

extrapolateMetricStatValuesForAddedLease

public void extrapolateMetricStatValuesForAddedLease(Map<String,Double> workerMetricToFleetLevelAverage,
                                                     int averageLeaseCount)

Increase the metric stat value corresponding to the added single lease. This is done during execution of LAM and as assignments are happening the load is increase for LAM to determine workers for assignment. The increase is done considering that for a WorkerMetric the fleet level average would be met when fleet level average leases are assigned to a worker and thus 1 lease addition increases the metric stat value by fleet level average of metric stat by averageLeaseCount

computePercentageToReachAverage

public double computePercentageToReachAverage(Map<String,Double> workerMetricToFleetLevelAverage)

Determines percentage of load to reach the mean for the worker. In case of multiple worker metrics the metric stat value closest to mean is used to determine the percentage value. This value is indication of how much load in percentage to current load the worker can take to reach mean value.

Parameters:

workerMetricToFleetLevelAverage - : WorkerMetric to fleet level mean value.

Returns:

percentage to reach mean based on the WorkerMetric closest to its corresponding average.

isAnyWorkerMetricFailing

public boolean isAnyWorkerMetricFailing()

Returns true if any of the metric stat values has -1 in last index which represents that the metric stat value was not successfully fetched in last attempt by worker.

Returns:

true if any metric stat value has -1 in last index, false otherwise.

isValidWorkerMetric

public boolean isValidWorkerMetric()

WorkerMetricStats entry is invalid if any of the field from lastUpdateTime, operatingRange, resourcesStats are not present or if resourcesStats is empty or if any of the WorkerMetrics having resourceStats does not have operatingRange or if operating range values are not present or if maxUtilization is 0 for any WorkerMetric

Returns:

true if the entry is valid false otherwise.

isAnyWorkerMetricAboveAverageUtilizationOrOperatingRange

public boolean isAnyWorkerMetricAboveAverageUtilizationOrOperatingRange(Map<String,Double> workerMetricToFleetLevelAverage)

isUsingDefaultWorkerMetric

public boolean isUsingDefaultWorkerMetric()

If a worker is not using an explicit WorkerMetric such as CPU, Memory, or Network, then it is said to be using the default WorkerMetric. Load management then falls back to throughput.

Returns:

true if the worker is not using an explicit WorkerMetric.

isWorkerMetricAboveOperatingRange

public boolean isWorkerMetricAboveOperatingRange(String workerMetricName)

Evaluates if the given metric stat is above operatingRange for the given WorkerMetric name. If the WorkerMetric does not exist returns false

Parameters:

workerMetricName - WorkerMetric name to evaluate

Returns:

true if metric stat exists and is above operatingRange for the WorkerMetric

getKeySchema

public static List<software.amazon.awssdk.services.dynamodb.model.KeySchemaElement> getKeySchema()

getAttributeDefinitions

public static List<software.amazon.awssdk.services.dynamodb.model.AttributeDefinition> getAttributeDefinitions()