Skip to main content

Test-DkvLatency

Synopsis​

Measures read latency statistics for a Clustron Distributed Key-Value (DKV) store.

Syntax

Test-DkvLatency -StoreName <string> -Iterations <int> [-ObjectSizeBytes <int>]

Description

Test-DkvLatency performs a latency benchmark against a Clustron store by:

  1. Pre-populating the store with test objects.
  2. Performing a short warm‑up phase.
  3. Measuring read latency for each request.
  4. Calculating latency statistics including min, average, P95, P99, and max.

This command is useful for:

  • evaluating cluster performance
  • validating deployment latency
  • comparing hardware or network environments
  • diagnosing slow responses

The benchmark performs sequential read operations and reports latency in milliseconds.

Parameters

-StoreName​

Specifies the target store used for the benchmark.

Type: String
Mandatory: True

Example:

Test-DkvLatency -StoreName TestStore -Iterations 1000

-Iterations​

Specifies the number of operations to measure.

Type: Int32
Mandatory: True
Range: 1 – 1,000,000

Example:

Test-DkvLatency -StoreName TestStore -Iterations 5000

-ObjectSizeBytes​

Specifies the approximate size of each stored object.

Type: Int32
Default: 100
Range: 32 – 1 MB

Example:

Test-DkvLatency -StoreName TestStore -Iterations 10000 -ObjectSizeBytes 512

Output

Returns an object containing latency statistics.

Property Description

StoreName Target store Iterations Number of operations measured ObjectSizeBytes Size of objects used MinMs Minimum latency AvgMs Average latency P95Ms 95th percentile latency P99Ms 99th percentile latency MaxMs Maximum latency

Example output:

StoreName : TestStore Iterations : 1000 ObjectSizeBytes : 100 MinMs : 0.21 AvgMs : 0.45 P95Ms : 0.72 P99Ms : 1.10 MaxMs : 1.48

Examples

Basic latency benchmark​

Test-DkvLatency -StoreName TestStore -Iterations 1000

Test latency with larger objects​

Test-DkvLatency -StoreName TestStore -Iterations 5000 -ObjectSizeBytes 1024

Notes

  • The command writes benchmark data into the store during pre‑population.
  • The benchmark performs read latency measurement.
  • Results depend heavily on network latency and cluster size.
  • Press Ctrl+C to cancel the benchmark safely.