harmony 鸿蒙Comparison Between TaskPool and Worker

2023-10-30
浏览 (466)

Comparison Between TaskPool and Worker

TaskPool and Worker provide a multithread running environment for applications to process time-consuming computing tasks or resource intensive tasks, preventing these tasks from blocking the main thread. This maximizes system utilization, reduces overall resource consumption, and improves overall system performance.

This topic compares TaskPool with Worker from two aspects: implementation features and use cases. It also describes their operating mechanisms and precautions.

Implementation Feature Comparison

Table 1 Comparison between TaskPool and Worker in implementation features

Item	TaskPool	Worker
Memory model	Threads are isolated from each other, and memory is not shared.	Threads are isolated from each other, and memory is not shared.
Parameter passing mechanism	The structured clone algorithm is used for serialization and deserialization. ArrayBuffer and SharedArrayBuffer are used for parameter passing and sharing.	The structured clone algorithm is used for serialization and deserialization. ArrayBuffer and SharedArrayBuffer are used for parameter passing and sharing.
Parameter passing	Parameters are directly passed in, without being encapsulated.	Only one parameter can be carried in a message object. Therefore, you must encapsulate excess parameters.
Method invocation	Methods are directly passed in and called.	Messages are passed in the worker thread and the corresponding methods are called.
Return value	A value is returned by default after asynchronous calling.	Messages proactively sent must be parsed and assigned by calling onmessage().
Lifecycle	The task pool manages its own lifecycle, without considering the load.	You are required to manage the number and lifecycle of worker threads.
Maximum number of task pools	The number is automatically managed, rather than being manually configured.	A maximum of eight worker threads are supported.
Maximum task execution duration	3 minutes (excluding the time used for Promise or async/await asynchronous call, for example, the time consumed by I/O tasks such as network download and file read/write)	There is no restriction.
Task priority setting	Setting the task priority is supported.	Setting the task priority is not supported.
Task cancellation	Tasks that have been initiated can be canceled.	Tasks that have been initiated cannot be canceled.

Use Case Comparison

Both TaskPool and Worker support multithread concurrency. TaskPool worker threads are bound to the system scheduling priority and support load balancing (automatic scaling). Worker threads are manually created and do not support scheduling priority setting. Therefore, TaskPool provides better performance than Worker and is recommended in most scenarios.

TaskPool is oriented to thread-level independent tasks, and tasks running for more than 3 minutes are automatically reclaimed by the system. Worker is oriented to threads and supports thread execution for a long time.

Common use cases are as follows:

Use Worker for a task that runs for more than 3 minutes (excluding the time used for Promise or async/await asynchronous call, for example, I/O tasks such as network download and file read/write). For example, use Worker for a 1-hour prediction algorithm training job in the background.
Use Worker for a series of associated synchronous tasks. For example, use Worker for a series of database operations, since the same handle is required.
Use TaskPool for a task for which the priority needs to be set. For example, in the histogram rendering scenario in Gallery, histogram data calculated in the background is used for UI display on the foreground. This requires high-priority processing. In this case, use TaskPool.
Use TaskPool for a task that needs to be canceled frequently. For example, in the large image browsing scenario in Gallery, both images on the left and right sides of the current image are cached. When the user slides to the next image, a cache task on one side needs to be canceled. In this case, use TaskPool.
Use TaskPool for a large number of tasks or tasks with scattered scheduling points. For example, a large-scale application with multiple modules has multiple time-consuming tasks, and it is inconvenient to use eight worker threads to manage load. In this case, TaskPool is recommended.

TaskPool Operating Mechanism

Figure 1 TaskPool operating mechanism

taskpool

With TaskPool, you can encapsulate tasks in the main thread and throw the tasks to the task queue. The system selects proper worker threads to distribute and execute the tasks, and then returns the result to the main thread. TaskPool provides APIs to execute and cancel tasks, and set the task priority. It also minimizes system resource usage through unified thread management, dynamic scheduling, and load balancing algorithms. By default, the system starts a worker thread and increases the thread quantity as the number of tasks increases. The maximum number of worker threads that can be created depends on the number of physical cores of the device. The formula is max(3, Number of physical cores – 1). If no task is distributed for a long period of time, the system reduces the number of worker threads.

Worker Operating Mechanism

Figure 2 Worker operating mechanism

worker

The thread that creates the worker thread is referred to as the host thread (not necessarily the main thread, since a worker thread can also create a worker subthread). A worker thread is also named an actor thread. Each worker thread has an independent instance from the host thread, including the infrastructure, object, and code segment. The worker thread communicates with the host thread by means of message exchange. They use the serialization technique to exchange commands and data.

Precautions for TaskPool

A task function must be decorated with \@Concurrent and can be used only in .ets files.
A task function must be a common function or async function, but not a class member function or anonymous function.
A task function can use imported variables and input parameter variables only in a project created on the stage model. In a project created on the FA model, it can use input parameter variables only.
A task function in the TaskPool worker thread must finish the execution within 3 minutes (excluding the time used for Promise or async/await asynchronous call, for example, the duration of I/O tasks such as network download and file read/write). Otherwise, it forcibly exits.
Input parameter types in a task function must be those supported by serialization. For details, see Common Objects.
Parameters of the ArrayBuffer type are transferred in TaskPool by default. You can set the transfer list by calling setTransferList().
Context objects vary in different threads. Therefore, the worker thread of TaskPool can use only a thread-safe library, rather than a non-thread-safe library, for example, UI-related non-thread-safe library. For details, see Precautions for Multithread Safe.
A maximum of 16 MB data can be serialized.

Precautions for Worker

The rules for passing in the Worker.ts path during the worker creation vary in different API versions. For details, see Precautions for File Paths.
After a worker thread is created, you must manually manage its lifecycle. A maximum of eight worker threads can run simultaneously. For details, see Lifecycle Precautions.
Modules of the ability type support Worker, but modules of the library type do not support Worker.
When creating a worker thread, the Worker.ts file of another module cannot be used. This means that a worker cannot be called across modules.
Context objects vary in different threads. Therefore, a worker thread of Worker can use only a thread-safe library, rather than a non-thread-safe library, for example, UI-related non-thread-safe library. For details, see Precautions for Multithread Safe.
A maximum of 16 MB data can be serialized.

Precautions for File Paths

Before calling an API of the Worker module, you must create a Worker instance. The constructor function varies in different API versions.

// Import the module.
import worker form '@ohos.worker';

// Use the following function in API version 9 and later versions:
const worker1: worker.ThreadWorker = new worker.ThreadWorker('entry/ets/workers/MyWorker.ts');
// Use the following function in API version 8 and earlier versions:
const worker2: worker.Worker = new worker.Worker('entry/ets/workers/MyWorker.ts');

The Worker.ts file path (specified by scriptURL) must be passed in the constructor function. By default, the workers directory (upper-level directory of the Worker.ts file) is at the same level as the pages directory.

Stage Model

The following is an example of scriptURL in the constructor function:

// Import the module.
import worker form '@ohos.worker';

// Method 1
// In the stage model, the workers directory is at the same level as the pages directory in the entry module.
const worker1: worker.ThreadWorker = new worker.ThreadWorker('entry/ets/workers/MyWorker.ts', {name:"first worker in Stage model"});
// In the stage model, the workers directory is a child directory of the pages directory in the entry module.
const worker2: worker.ThreadWorker = new worker.ThreadWorker('entry/ets/pages/workers/MyWorker.ts');

// Method 2
// In the stage model, the workers directory is at the same level as the pages directory in the entry module, and bundlename is com.example.workerdemo.
const worker3: worker.ThreadWorker = new worker.ThreadWorker('@bundle:com.example.workerdemo/entry/ets/workers/worker');
// In the stage model, the workers directory is a child directory of the pages directory in the entry module, and bundlename is com.example.workerdemo.
const worker4: worker.ThreadWorker = new worker.ThreadWorker('@bundle:com.example.workerdemo/entry/ets/pages/workers/worker');

Based on the directory structure of the stage model project, the field meanings in method 1 are as follows:
- entry: value of the name attribute under module in the module.json5 file.
- ets: directory for storing the ArkTS source code. It is fixed.
- workers/MyWorker.ts: path of the worker source file in the ets directory.
Based on the directory structure of the stage model project, the field meanings in method 2 are as follows:
- \@bundle: fixed label.
- bundlename: bundle name of the current application.
- entryname: value of the name attribute under module in the module.json5 file.
- ets: directory for storing the ArkTS source code. It is fixed.
- workerdir/workerfile: path of the worker source file in the ets directory.

FA Model

The following is an example of scriptURL in the constructor function:

// Import the module.
import worker form '@ohos.worker';

// In the FA model, the workers directory is at the same level as the pages directory in the entry module.
const worker1: worker.ThreadWorker = new worker.ThreadWorker('workers/worker.js', {name:'first worker in FA model'});
// In the FA model, the workers directory is at the same level as the parent directory of the pages directory in the entry module.
const worker2: worker.ThreadWorker = new worker.ThreadWorker('../workers/worker.js');

Lifecycle Precautions

Creating and terminating worker threads consume performance. Therefore, you are advised to manage available workers and reuse them. The worker threads keep running even when they are idle. Therefore, when a worker thread is not required, call terminate() interface or parentPort.close() to destroy it. If a worker thread is destroyed or being destroyed, an error is thrown when it is called.
A maximum of eight worker threads can co-exist.
- In API version 8 and earlier versions, when the number of worker threads exceeds the upper limit, the error “Too many workers, the number of workers exceeds the maximum.” is thrown.
- Since API version 9, when the number of worker threads exceeds the upper limit, the error “Worker initialization failure, the number of workers exceeds the maximum.” is thrown.

Precautions for Multithread Safe

Multithread safe ensures the correctness and reliability of applications when multiple threads access or modify shared resources at the same time.

When you use TaskPool or Worker for multithread development, ensure that the APIs and modules imported to the worker thread of TaskPool and Worker support multithread safe. Otherwise, multithread data competition may occur, causing application exceptions or breakdown.

The following modules can be used in the worker thread of TaskPool and Worker. If other modules are used, you must check whether the thread security requirements are met.

console
setInterval
setTimeout
clearInterval
clearTimeout
@ohos.buffer
@ohos.convertxml
@ohos.file
- @ohos.file.backup
- @ohos.file.cloudSync
- @ohos.file.cloudSyncManager
- @ohos.file.environment
- @ohos.file.fileAccess
- @ohos.file.fileExtensionInfo
- @ohos.file.fileuri
- @ohos.file.fs
- @ohos.file.hash
- @ohos.file.photoAccessHelper
- @ohos.file.picker
- @ohos.file.securityLabel
- @ohos.file.statvfs
- @ohos.file.storageStatistics
- @ohos.file.volumeManager
@ohos.fileio
@ohos.hilog
@ohos.multimedia
- @ohos.multimedia.image
@ohos.net
- @ohos.net.http
@ohos.pasteboard
@ohos.systemDateTime
@ohos.systemTimer
@ohos.taskpool
@ohos.uri
@ohos.url
@ohos.util
- @ohos.util.ArrayList
- @ohos.util.Deque
- @ohos.util.HashMap
- @ohos.util.HashSet
- @ohos.util.LightWeightMap
- @ohos.util.LightWeightSet
- @ohos.util.LinkedList
- @ohos.util.List
- @ohos.util.PlainArray
- @ohos.util.Queue
- @ohos.util.Stack
- @ohos.util.TreeMap
- @ohos.util.TreeSet
- @ohos.util
@ohos.worker
@ohos.xml