LTE/

directory
v0.0.0-...-94b13b5 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Jan 9, 2022 License: Apache-2.0

README

Ledger Traffic Engine

This readme explains the working and the usage of Ledger Traffic Engine (LTE test tool.

  • What is the LTE Test Tool
  • How to Run the Tests
  • Where to View the Results

What is the LTE Test Tool

The LTE test tool is test harness that engages the Ledger APIs and benchmarks the performance of the Ledger component. It contains the framework for creating and managing chains, for transaction submission, block creation, and block commit. It conducts benchmark tests for insert and read-write transactions (transactions per second) and contains scripts for launching the benchmarks. An insert benchmark followed by a readwrite benchmark on the same dataset is considered to constitute a single test-run.

How to Run The tests

To run all the available tests, run:

cd fabric/test/tools/LTE/scripts
./runbenchmark.sh -f parameters_daily_CI.sh all

where the file parameters_daily_CI.sh has all the necessary test parameters.

You can run individual tests without running all the available tests by giving the name of the test as parameter, instead of all. You can get the available test names by:

./runbenchmark.sh help
What the Tests Do

Each test reads test parameters from the provided parameter file and conducts several test-runs by varying one or two of the parameters. The parameters are:

  • number of chains (ledger),
  • number of parallel transactions in each chain,
  • number of Key-value pairs,
  • number of transactions,
  • number of keys in each transaction,
  • size of batch for ledger,
  • size of Key-value

For example, the varyNumChains test reads the parameters and varies the number of chains for each test-run while keeping the other parameters constant, and generate result. Varying a specific parameter for each test-runs gives insight into the parameter's influence on the Ledger performance.

Each test-run consists of two phases: benchmarking of ledger insert phase and benchmarking of ledger read-write phase.

Insert Phase

The insert benchmark starts fresh chains and inserts the Key-values by simulating writes-only transactions For each of the chains. It launches the parallel clients (based on the configuration) and the clients simulate and commit write-only transactions. The keys are divided among clients such that one key is written only once and all the keys are inserted. For instance, if this benchmark is invoked with the following parameters:

Number of chains=2,
Number of parallel transactions per chain=2,
Number of key-value pairs=100

then client_1 on chain_1 will insert Key_1 to key_25 and client_2 on chain_1 will insert Key_26 to key_50. Similarly client_3 on chain_2 will insert Key_1 to key_25 and client_4 on chain_2 will insert Key_26 to key_50 where, client_1 and client_2 run in parallel on chain_1 and client_3 and client_4 run in parallel on chain_2.

Read-write Phase

Subsequently, the Read-write benchmark step opens the existing chains and modifies the Key-values by simulating read-write transactions For each of the chains. It launches the parallel clients (based on the configuration) and the clients simulate and commit read-write transactions. This test assumes the pre-populated chain by previously running insert benchmark step (described above). Each transaction simulated by this benchmark randomly selects a configurable number of keys and modifies their values. For instance, if this benchmark is invoked with the following test parameters:

Number of chains=2,
Number of parallel transactions per chain=2,
Number of key-value pairs=100,
Number of total transactions=200

then client_1, and client_2 both execute 50 transactions on chain_1 in parallel. Similarly, client_3, and client_4 both execute 50 transactions on chain_2 in parallel In each of the transactions executed by any client, the transaction expects and modifies any key(s) between Key_1 to key_50 (because, total keys are to be 100 across two chains).

Running with Custom Parameters

The tests can be run with user-defined parameters by creating a new file that has all the necessary parameters to run and using that file as the input (see the section on how to run the tests) . The names of necessary parameters can be found in the file parameters_daily_CI.sh.

How to View the Test Results

The test results can be viewed as in the stdout where it shows how long each single operation took to complete in a test. These results are also saved in a .csv file in the following directory: /tmp/experiments

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL