Jump to content

Recommended Posts

Posted (edited)

Hi,

 

This is a custom implementation of a thread pool management system, designed to efficiently manage the execution of tasks in a multi-threaded environment. Automatically adjusts the size of the thread pools based on the system's CPU load and the number of available cores. This dynamic adjustment ensures that the thread pools are always optimized for the current system load, resulting in efficient resource utilization and improved performance. If the calculated sizes are equal to the current core pool sizes, this means that there's no need to update the pool sizes, as they are already optimized for the current system load. I haven't conducted extensive testing yet; however, the code can be thoroughly tested, and if any issues are encountered, I can make the necessary modifications. Comments were added throughout the code.

 

For testing purposes you can do the following:

 

		scheduleAtFixedRate(() ->
		{
			try
			{
				purge();
				updateThreadPoolSizes();
			}
			catch (Exception e)
			{
				LOGGER.severe("Error during scheduled ThreadPool update: " + e.getMessage());
			}
		}, 10, 10, TimeUnit.SECONDS);

 

	private static void updateThreadPoolSizes()
	{
		double systemLoad = getSystemCpuLoad();
		int availableCores = Runtime.getRuntime().availableProcessors();
		int newScheduledThreadPoolSize = calculateThreadPoolSize(availableCores, systemLoad, true);
		int newInstantThreadPoolSize = calculateThreadPoolSize(availableCores, systemLoad, false);
		
		if (newScheduledThreadPoolSize != scheduledExecutor.getCorePoolSize())
		{
			scheduledExecutor.setCorePoolSize(newScheduledThreadPoolSize);
			LOGGER.info("Updated scheduled thread pool size to " + newScheduledThreadPoolSize + ", CPU: " + getSystemCpuLoadPercentage());
		}
		else
		{
			LOGGER.info("Scheduled thread pool size remains to " + newScheduledThreadPoolSize + ", CPU: " + getSystemCpuLoadPercentage());
		}
		if (newInstantThreadPoolSize != instantExecutor.getCorePoolSize())
		{
			instantExecutor.setCorePoolSize(newInstantThreadPoolSize);
			LOGGER.info("Updated instant thread pool size to " + newInstantThreadPoolSize + ", CPU: " + getSystemCpuLoadPercentage());
		}
		else
		{
			LOGGER.info("Instant thread pool size remains to " + newInstantThreadPoolSize + ", CPU: " + getSystemCpuLoadPercentage());
		}
	}

 

P.S.: This is the lowest it can get:
 

return Math.max(isScheduledPool ? 16 : 8, threadPoolSize);

 

The initial code:
 

 * This file is part of the L2Gold Classic project.
 * 
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 * 
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 * General Public License for more details.
 * 
 * You should have received a copy of the GNU General Public License
 * along with this program. If not, see <http://www.gnu.org/licenses/>.
 */
package gold.lineage2.commons.threads;

import java.lang.management.ManagementFactory;
import java.util.concurrent.LinkedBlockingQueue;
import java.util.concurrent.ScheduledFuture;
import java.util.concurrent.ScheduledThreadPoolExecutor;
import java.util.concurrent.ThreadPoolExecutor;
import java.util.concurrent.TimeUnit;
import java.util.logging.Logger;

import com.sun.management.OperatingSystemMXBean;

/**
 * A custom implementation of a thread pool management system, designed to efficiently manage the execution of tasks in a multi-threaded environment.
 * Automatically adjusts the size of the thread pools based on the system's CPU load and the number of available cores. This dynamic adjustment ensures
 * that the thread pools are always optimized for the current system load, resulting in efficient resource utilization and improved performance.
 * @author Trance
 */
public class ThreadPool
{
	// Logger to log information and errors.
	private static final Logger LOGGER = Logger.getLogger(ThreadPool.class.getName());
	// Maximum delay to be used to validate the delay.
	private static final long MAX_DELAY = TimeUnit.NANOSECONDS.toMillis(Long.MAX_VALUE - System.nanoTime()) / 2;
	
	// ScheduledThreadPoolExecutor for scheduled tasks.
	private static ScheduledThreadPoolExecutor scheduledExecutor;
	// ThreadPoolExecutor for instant tasks.
	private static ThreadPoolExecutor instantExecutor;
	
	/**
	 * Initialize the ThreadPool with the appropriate sizes.
	 */
	public static void init()
	{
		LOGGER.info("ThreadPool: Initialized");
		
		int availableCores = Runtime.getRuntime().availableProcessors();
		int scheduledThreadPoolSize = availableCores * 4;
		int instantThreadPoolSize = availableCores * 2;
		
		scheduledExecutor = new ScheduledThreadPoolExecutor(scheduledThreadPoolSize, new PriorityThreadFactory("ScheduledThreadPool", Thread.NORM_PRIORITY), new ThreadPoolExecutor.CallerRunsPolicy());
		scheduledExecutor.setRejectedExecutionHandler(new RejectedExecutionHandlerImpl());
		scheduledExecutor.prestartAllCoreThreads();
		LOGGER.info("Scheduled thread pool size: " + scheduledThreadPoolSize);
		
		instantExecutor = new ThreadPoolExecutor(instantThreadPoolSize, Integer.MAX_VALUE, 5, TimeUnit.SECONDS, new LinkedBlockingQueue<Runnable>(), new PriorityThreadFactory("ThreadPoolExecutor", Thread.NORM_PRIORITY), new ThreadPoolExecutor.CallerRunsPolicy());
		instantExecutor.setRejectedExecutionHandler(new RejectedExecutionHandlerImpl());
		instantExecutor.prestartAllCoreThreads();
		LOGGER.info("Instant thread pool size: " + instantThreadPoolSize);
		
		scheduleAtFixedRate(() ->
		{
			try
			{
				purge();
				updateThreadPoolSizes();
			}
			catch (Exception e)
			{
				LOGGER.severe("Error during scheduled ThreadPool update: " + e.getMessage());
			}
		}, 5, 5, TimeUnit.MINUTES);
	}
	
	/**
	 * Purge both thread pools.
	 */
	public static void purge()
	{
		scheduledExecutor.purge();
		instantExecutor.purge();
	}
	
	/**
	 * Validate the delay ensuring it's within acceptable bounds.
	 * @param delay The delay value to be validated.
	 * @param timeUnit The time unit of the delay.
	 * @return A validated delay value.
	 */
	private static long validate(long delay, TimeUnit timeUnit)
	{
		long delayInMilliseconds = timeUnit.toMillis(delay);
		long validatedDelay = Math.max(0, Math.min(MAX_DELAY, delayInMilliseconds));
		
		if (delayInMilliseconds > validatedDelay)
		{
			return -1;
		}
		
		return timeUnit.convert(delayInMilliseconds, TimeUnit.MILLISECONDS);
	}
	
	/**
	 * Schedule a Runnable with a specific delay.
	 * @param r The Runnable to be scheduled.
	 * @param delay The delay before the Runnable is executed, in milliseconds.
	 * @return A ScheduledFuture representing the pending result of the task.
	 */
	public static ScheduledFuture<?> schedule(Runnable r, long delay)
	{
		return schedule(r, delay, TimeUnit.MILLISECONDS);
	}
	
	/**
	 * Schedule a Runnable with a specific delay and TimeUnit.
	 * @param r The Runnable to be scheduled.
	 * @param delay The delay before the Runnable is executed.
	 * @param timeUnit The time unit of the delay.
	 * @return A ScheduledFuture representing the result of the scheduling.
	 */
	private static ScheduledFuture<?> schedule(Runnable r, long delay, TimeUnit timeUnit)
	{
		delay = validate(delay, timeUnit);
		if (delay == -1)
		{
			return null;
		}
		
		return scheduledExecutor.schedule(new RunnableWrapper(r), delay, timeUnit);
	}
	
	/**
	 * Schedule a Runnable at a fixed rate with an initial delay.
	 * @param r The Runnable to be scheduled.
	 * @param initial The initial delay before the Runnable is executed for the first time.
	 * @param delay The delay between the execution of the Runnable in subsequent runs.
	 * @return A ScheduledFuture representing the result of the scheduling.
	 */
	public static ScheduledFuture<?> scheduleAtFixedRate(Runnable r, long initial, long delay)
	{
		return scheduleAtFixedRate(r, initial, delay, TimeUnit.MILLISECONDS);
	}
	
	/**
	 * Schedule a Runnable at a fixed rate with an initial delay and TimeUnit.
	 * @param r The Runnable to be scheduled.
	 * @param initial The initial delay before the Runnable is executed for the first time.
	 * @param delay The delay between the execution of the Runnable in subsequent runs.
	 * @param timeUnit The time unit of the delay.
	 * @return A ScheduledFuture representing the result of the scheduling.
	 */
	private static ScheduledFuture<?> scheduleAtFixedRate(Runnable r, long initial, long delay, TimeUnit timeUnit)
	{
		initial = validate(initial, timeUnit);
		if (initial == -1)
		{
			return null;
		}
		
		delay = validate(delay, timeUnit);
		if (delay == -1)
		{
			return scheduledExecutor.schedule(new RunnableWrapper(r), initial, timeUnit);
		}
		
		return scheduledExecutor.scheduleAtFixedRate(new RunnableWrapper(r), initial, delay, timeUnit);
	}
	
	/**
	 * Execute a Runnable instantly.
	 * @param r The Runnable to be executed.
	 */
	public static void execute(Runnable r)
	{
		instantExecutor.execute(r);
	}
	
	/**
	 * Shut down both thread pools.
	 * @throws InterruptedException If the shutdown process is interrupted.
	 */
	public static void shutdown() throws InterruptedException
	{
		try
		{
			scheduledExecutor.shutdown();
			if (!scheduledExecutor.awaitTermination(10, TimeUnit.SECONDS))
			{
				scheduledExecutor.shutdownNow();
			}
		}
		finally
		{
			instantExecutor.shutdown();
			if (!instantExecutor.awaitTermination(1, TimeUnit.MINUTES))
			{
				instantExecutor.shutdownNow();
			}
		}
	}
	
	/**
	 * Update the thread pool sizes based on system load.
	 */
	private static void updateThreadPoolSizes()
	{
		double systemLoad = getSystemCpuLoad();
		int availableCores = Runtime.getRuntime().availableProcessors();
		int newScheduledThreadPoolSize = calculateThreadPoolSize(availableCores, systemLoad, true);
		int newInstantThreadPoolSize = calculateThreadPoolSize(availableCores, systemLoad, false);
		
		if (newScheduledThreadPoolSize != scheduledExecutor.getCorePoolSize())
		{
			scheduledExecutor.setCorePoolSize(newScheduledThreadPoolSize);
			LOGGER.info("Updated scheduled thread pool size to " + newScheduledThreadPoolSize + ", CPU: " + getSystemCpuLoadPercentage());
		}
		if (newInstantThreadPoolSize != instantExecutor.getCorePoolSize())
		{
			instantExecutor.setCorePoolSize(newInstantThreadPoolSize);
			LOGGER.info("Updated instant thread pool size to " + newInstantThreadPoolSize + ", CPU: " + getSystemCpuLoadPercentage());
		}
	}
	
	/**
	 * Calculate the thread pool size based on available cores, system load, and whether it's a scheduled pool.
	 * @param availableCores The number of available processor cores.
	 * @param systemLoad The current system load.
	 * @param isScheduledPool A boolean indicating if the pool is a scheduled thread pool or not.
	 * @return The calculated thread pool size.
	 */
	private static int calculateThreadPoolSize(int availableCores, double systemLoad, boolean isScheduledPool)
	{
		double factor;
		if (systemLoad <= 0.4)
		{
			factor = isScheduledPool ? 4 : 2;
		}
		else if (systemLoad <= 0.6)
		{
			factor = isScheduledPool ? 3 : 1.5;
		}
		else if (systemLoad <= 0.8)
		{
			factor = isScheduledPool ? 2 : 1;
		}
		else
		{
			factor = 0.5;
		}
		
		int threadPoolSize = (int) Math.round(availableCores * factor);
		return Math.max(isScheduledPool ? 16 : 8, threadPoolSize);
	}
	
	/**
	 * Get the system CPU load.
	 * @return A double value representing the system CPU load.
	 */
	@SuppressWarnings("deprecation")
	private static double getSystemCpuLoad()
	{
		OperatingSystemMXBean osBean = ManagementFactory.getPlatformMXBean(OperatingSystemMXBean.class);
		return osBean.getSystemCpuLoad();
	}
	
	/**
	 * Get the system CPU load as a percentage.
	 * @return A string representing the system CPU load percentage.
	 */
	private static String getSystemCpuLoadPercentage()
	{
		double cpuLoad = getSystemCpuLoad();
		return String.format("%.2f%%", cpuLoad * 100);
	}
}

 

Edited by Trance
  • Like 3
  • Thanks 2
  • Upvote 2
Posted
14 minutes ago, Kishin said:

what of  ? 

PriorityThreadFactory


You don't need to copy and paste everything. Instead, you can use the same logic to implement it in your Thread Pool Manager. However, I am providing you with the files: https://files.lineage2.gold/threads.zip

Posted
1 minute ago, Trance said:


You don't need to copy and paste everything. Instead, you can use the same logic to implement it in your Thread Pool Manager. However, I am providing you with the files: https://files.lineage2.gold/threads.zip

ye was about to do that , but was curious of its purpose . since even if i did implement it how it was , i would have to rework the whole src  .
though ty for the file . i'll look around and work around on these 

  • Upvote 1

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


  • Posts

    • L2jBayev Chronicle 3: Rise of Darkness – AiEngine Edition In short: this is a C3 build with a full-fledged AI engine, live mercenaries, a built-in quiz, a “personal account” in the Community Board, and server logic neatly distributed across thread pools. The project is about a living world without lags : bots farm, communicate, gather parties, teleport along routes, and the server remains cold and stable.   What's inside (the most delicious) 1) Full-fledged AI engine for characters Behavior types: farming ( FarmAI ), combat ( CombatAI ), party logic ( PartyAI ), trading/walking ( TraderAI / WalkerAI ), support roles (healer, etc.). Class profiles: for mages/archers/daggers, etc., “smart” skill rotations, distance control, sleep/save skills, healing, loot pickup, etc. are implemented (see examples of classes like SpellSingerAI , NecromancerAI , etc.). Self-healing and teleports: when dying, the bot goes through a sequence of steps without sleep()- via AITaskSequence + AITeleportToLocTask , searches for the nearest gatekeeper and teleports via TeleportationManager with routes depending on the level. Auto-support: auto-nipples, arrows/bones, smart auto-proceduring of buffs and auto-banks CP/HP/MP with thresholds - all sewn into the auxiliary EtcPlayersAi . Chat context: ChatManagerAi processes mentions, makes responses with delays (anti-flood), supports party chat and “human” reaction. Understanding: ChatManagerAi system  processes the dialogue, bots remember your aggression and insults, they start to respond less often to modern users, stop accepting or inviting to a group (party) and when it goes beyond the peak they will simply merge you, and every time they see you on the PC, there is an opportunity to measure more often, communicate respectfully and beautifully, in general, a “human” reaction. Why a player/admin needs this: bots actually “live”, farm and interact, and don’t just stand on macros. This is a great background for online and PvE action.   2) Mercenaries (Mercenary system) Full-fledged companion character : L2MercenaryInstance with its own MercenaryAI (movement, attack, support, consumables, shots). Behavior modes: DEFENDER / SUPPORT / PASSIVE - switchable to suit your playing style. Progress and trust: the mercenary's trust/exp/level grows , skills are learned according to the MercenarySkillTree (conditions are based on the trust or level of the owner). Templates and equipment: via MercenaryTemplateTable and spawner - model/weapon/type are selected. Social: MercenarySpeechManager - a set of speeches; the mercenary "comes to life" in the chat. Premium Link: Premium account owners give the mercenary additional trust (faster progress). Why: This is not a dummy pet, but a playful companion with modes, training and “character”.   3) Quiz (event viktorina ) Rounds according to schedule: pre-launch with announcements (minutes/seconds before start), registration .reg, auto-opening of the window. Multiple choice questions: question + set of answer buttons; fair processing, timings, question change. Tops and history: results table, statistics, neat UI via HTML assembly. Flexible control: you can start immediately or set a delayed start (notification package 5/2/1 min, etc.). Why: regular activity for players, “social entertainment” module right in the build.   4) Personal account in Community Board KB managers: buff cabinet, teleports, clans/forums/mail/friends, tops (PK/PvP/wealth/players), character repair, viewing skill trees , etc. Premium logic: some services/mail are limited by premium; premium also affects the visual (nickname color) and bonuses (see effect on mercenary). Single sign-on: all in one place, no team chaos. Why: conveniently manage your character and services without going into the console or installing third-party mods.   Why is the system technically valuable? Minimum load and stability Separated thread pools: AI logic, hunting, teleports, chat - on separate onesScheduledExecutorService ( AI_THREAD_POOL , MONSTER_HUNT_POOL , TELEPORT_POOL , CHAT_POOL ). No "freezing": task sequencers (teleport/recovery) work through the scheduler, not Thread.sleep(). Bot limitation: protection against overload via thresholds/counters - “extra” bots do not start. One bot - one sequence: AITaskManager ensures that the character does not have parallel conflicting tasks. Smoothing out peaks: starting tasks with offsets so that there are no simultaneous “ticks” of hundreds of bots. Monitoring/logs: own loggers (separate files for info/errors/processes/chats), CPU load monitoring. Bottom line: the build is designed for “thick online” and mass activities without TPS failures .   Additional Features Auto-alliances for farming: party logic invites suitable players (checking level/equipment/clan flags), there are “human” responses to requests. Sub/class management: out of the box helpers for changing class/subclass, auto-learning of necessary skills and selection of equipment by level. Security/protection: secondary PIN/picture password support (used in KB/voiced commands; optional). Premium accounts: privileges in KB/mail/visual and synergy with mercenary progress. Ready-made services: tops, auctions/mail, teleports from KB, buff rooms, repairs, viewing skill trees, etc.   Who is this build for? Freeshare/project admins who want a living world “from the pack”: bots and mercenaries provide a constant background of activity. Players who value convenience: personal account, premium services, events and a mercenary companion. Developers who want a clean, predictable backend with thread pools and a neat task model without “magic”.   How it differs from standard assemblies Not macros - AI profiles with “brains”: rotations, positioning, healing, decision making. Not a decoration pet - a mercenary with his own modes, progress, skill tree and lines. Not a faceless gamemod - an event quiz with UI, schedule, tops. No chaos in flows - strict pools, planning and task managers designed for online and growth. No separate scripts - a single personal account in KB for most activities.   TL;DR (one paragraph for the project card) AiEngine C3 is a build with live AI, smart bots, mercenaries (modes/progress/skills), built-in quiz, premium logic and a convenient personal account in KB. Under the hood are distributed thread pools and task managers without sleep(), so even with a dense online the server remains stable and responsive.   Additionally add - there is still a lot of interesting things command .assassin or shift+target (order murder), shift+target for admins on AI characters for control, admin panel is completely rewritten, many additional functions, mercenaries change their appearance depending on trust, deepseek and chatGPT system is connected for communication of characters like real players, GPT - for newer java, there is still a very large list of fixes after the last versions, a lot has been fixed, including height coordinates (Z) geo-Squares, pathfinding, visibility through obstacles, fix pet summons, trade packages, shop packages, many effects, quests (including the original ones like nipples, etc.), Ai behavior of NPC and RB monsters, absolutely all epics have been transferred to AiLoader no longer in python scripts. Attention! The server is suitable for both classic mode and PvP format, as well as with various mods. Absolutely everything is configured in the configurations to suit your taste and purposes of use. It is recommended to launch the server through L2ServerControl (simplifies management and control of processes). Download Servers: Chronicle 3 Server Chronicle 4 Test Upgraded Server Full Desc & screens: Post & Screens c3 Post & Desc c4    
    • 🎃 HALLOWEEN EVENT 🎃   ‼️ Information and details: https://forum.l2harbor.com/threads/halloween-event-fall-harvest-30-10-07-11.8265/post-168620
  • Topics

×
×
  • Create New...

AdBlock Extension Detected!

Our website is made possible by displaying online advertisements to our members.

Please disable AdBlock browser extension first, to be able to use our community.

I've Disabled AdBlock