org.opennms.netmgt.measurements.impl

Class NewtsFetchStrategy

java.lang.Object

org.opennms.netmgt.measurements.impl.NewtsFetchStrategy

All Implemented Interfaces:

MeasurementFetchStrategy

public class NewtsFetchStrategy
extends Object
implements MeasurementFetchStrategy

Used to retrieve measurements from SampleRepository. If a request to fetch(long, long, long, int, java.lang.Long, java.lang.Long, java.util.List<org.opennms.netmgt.measurements.model.Source>, boolean) spans multiple resources, separate calls to the SampleRepository will be performed in parallel. Reading the samples and computing the aggregated values can be very CPU intensive. The "parallelism" attribute is used to set an upper limit on how may concurrent threads can be used to perform these calculations. By default, this is set to the number of cores, but can be reduced if the operator wishes to ensure cores are available for other purposes.

Author:

jwhite

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected static class	NewtsFetchStrategy.LateAggregationParams

Field Summary

Fields

Modifier and Type	Field and Description
static long	DEFAULT_HEARTBEAT_MS
static int	INTERVAL_DIVIDER
static long	MIN_STEP_MS
static int	PARALLELISM

Constructor Summary

Constructors

Constructor and Description
NewtsFetchStrategy()

Method Summary

Modifier and Type	Method and Description
FetchResults	fetch(long start, long end, long step, int maxrows, Long interval, Long heartbeat, List<Source> sources, boolean relaxed) Fetches the measurements for the given sources.
protected static NewtsFetchStrategy.LateAggregationParams	getLagParams(long step, Long interval, Long heartbeat) Calculates the parameters to use for late aggregation.
protected void	setContext(org.opennms.newts.api.Context context)
protected void	setResourceDao(ResourceDao resourceDao)
protected void	setSampleRepository(org.opennms.newts.api.SampleRepository sampleRepository)

Methods inherited from class java.lang.Object

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

Field Detail

MIN_STEP_MS

public static final long MIN_STEP_MS

INTERVAL_DIVIDER

public static final int INTERVAL_DIVIDER

DEFAULT_HEARTBEAT_MS

public static final long DEFAULT_HEARTBEAT_MS

PARALLELISM

public static final int PARALLELISM

Constructor Detail

NewtsFetchStrategy

public NewtsFetchStrategy()

Method Detail

fetch

public FetchResults fetch(long start,
                          long end,
                          long step,
                          int maxrows,
                          Long interval,
                          Long heartbeat,
                          List<Source> sources,
                          boolean relaxed)

Description copied from interface: MeasurementFetchStrategy

Fetches the measurements for the given sources.

Specified by:

fetch in interface MeasurementFetchStrategy

Parameters:

start - timestamp in milliseconds

end - timestamp in milliseconds

step - desired resolution in milliseconds - actual resolution might differ

maxrows - maximum number of rows - no limit when <= 0

interval - duration in milliseconds, used by strategies that implement late aggregation

heartbeat - duration in milliseconds, used by strategies that implement late aggregation

sources - array of sources - these should have unique labels

relaxed - if false a missing source results in a return of null. true on the other hand ignores that source.

Returns:

null when a resource id or attribute cannot be found

getLagParams

protected static NewtsFetchStrategy.LateAggregationParams getLagParams(long step,
                                                                       Long interval,
                                                                       Long heartbeat)

Calculates the parameters to use for late aggregation. Since we're in the process of transitioning from an RRD-world, most queries won't contain a specified interval or heartbeat. For this reason, we need to derive sensible values that will allow users to visualize the data on the graphs without too many NaNs. The given step size will be variable based on the time range and the pixel width of the graph, so we need to derive the interval and heartbeat accordingly. Let S = step, I = interval and H = heartbeat, the constraints are as follows: 0 < S 0 < I 0 < H S = aI for some integer a >= 2 H = bI for some integer b >= 2 While achieving these constraints, we also want to optimize for: min(|S - S*|) where S* is the user supplied step and S is the effective step.

setResourceDao

protected void setResourceDao(ResourceDao resourceDao)

setSampleRepository

protected void setSampleRepository(org.opennms.newts.api.SampleRepository sampleRepository)

setContext

protected void setContext(org.opennms.newts.api.Context context)