即将推出适用于单人及多人模式的 Fabric 模组。
GrimLag
自适应延迟控制,而非惊慌失措。
GrimLag 是一款专为 Paper 服务器设计的、具备延迟感知能力的自适应性能管理插件。
它不会频繁执行清除操作,不会一见农场就“核爆”式清理,也不会假装延迟是玩家的错。它会持续监控服务器状况,判断实际问题的严重程度,并做出相应响应。
专注于损害控制,而非焦土政策。
GrimLag 的功能
自适应延迟阶段
GrimLag 持续监控服务器性能,并在四个内部阶段之间切换:
HEALTHY(健康)——一切正常运行
WARNING(警告)——开始轻度干预
DEGRADED(降级)——采取积极但公平的限速措施
CRITICAL(危急)——进入生存模式
没有生硬的切换,也没有瞬间的惩罚。服务器会平缓地施加限制,并在情况好转后平缓解除。
实体 AI 节流(含恢复机制)
当出现延迟尖峰时,选择性禁用怪物 AI 以降低 CPU 负载。
当性能恢复时,AI 会自动重新启用。
怪物不会永远失去智能。
这里不是博物馆。
刷怪笼速率缩放
刷怪笼的输出量会根据当前的延迟阶段动态调整。
高负载下刷怪笼速度减慢
服务器恢复后恢复正常行为
无需全局“关闭开关”导致的愤怒退游
实体出生控制
GrimLag 按区块限制实体过度繁殖,而非全局限制。
首先采用软性上限
仅在必要时启用硬性上限
绝不会移除玩家
高优先级实体存活时间更长
拥挤的区块会得到处理,而不会导致整个世界范围内的活动停止。
延迟感知的红石节流
在延迟尖峰期间,使用概率性节流来抑制红石活动。
红石仍然有效
只是当服务器“着火”(卡顿严重)时,它不会在每个刻都触发
减少延迟,减少故障装置
这是故意的干扰,但很有礼貌。
物品生命周期管理
与其直接清空掉落物:
尽可能合并物品
在延迟期间,旧物品的老化速度加快
然后才移除多余物品
你的服务器获得的是清理,而不是灭绝事件。
物品生命周期管理
与其直接清空掉落物:
尽可能合并物品
在延迟期间,旧物品的老化速度加快
然后才移除多余物品
你的服务器获得的是清理,而不是灭绝事件。
软性恢复逻辑
随着性能改善,GrimLag 会撤销自身的限制:
AI 恢复
刷怪笼恢复正常
红石恢复全功能行为
消失规则放宽
没有任何限制会比必要的更久。
设计哲学
适应变化,而非惊慌
进行节流,而非摧毁
自动恢复
保持可预测,而非激进
GrimLag 专为长期运行的生存服务器打造,而非短期基准测试。
要求
Paper(推荐)
Java 17+
SOON THERE WILL BE A FABRIC MOD FOR WHICH IT WILL BE FOR BOTH SINGLEPLAYER AND MULTIPLAYER COMPATIBLE.
GrimLag
Lag control that adapts instead of panicking.
GrimLag is a lag-aware, adaptive performance management plugin for Paper servers.
It doesn’t spam clears, doesn’t nuke farms on sight, and doesn’t pretend lag is the players’ fault. It watches the server, decides how bad things really are, and responds accordingly.
Think damage control, not scorched earth.
What GrimLag Does
Adaptive Lag Stages
GrimLag continuously monitors server performance and shifts between four internal stages:
HEALTHY – Everything runs normally
WARNING – Light intervention begins
DEGRADED – Aggressive but fair throttling
CRITICAL – Survival mode
No hard switches. No instant punishment. The server eases into restrictions and eases back out.
Entity AI Throttling (With Recovery)
When lag spikes, mob AI is selectively disabled to reduce CPU load.
When performance recovers, AI is restored automatically.
Mobs don’t stay brain-dead forever.
This isn’t a museum.
Spawner Rate Scaling
Spawner output is dynamically scaled based on current lag stage.
Spawners slow down under load
Resume normal behavior when the server recovers
No global “off switch” rage-quits
Entity Birth Control
GrimLag limits entity overpopulation per chunk, not globally.
Soft caps first
Hard caps only when needed
Players are never removed
Priority entities survive longer
Crowded chunks get handled without killing activity across the entire world.
Lag-Aware Redstone Throttling
Redstone activity is dampened using probabilistic throttling during lag spikes.
Redstone still works
It just doesn’t fire every single tick when the server is on fire
Less lag, fewer broken contraptions
This is intentional sabotage. But polite.
Item Lifecycle Management
Instead of nuking dropped items:
Items are merged where possible
Old items age faster under lag
Only then are excess items removed
Your server gets cleanup — not an extinction event.
Item Lifecycle Management
Instead of nuking dropped items:
Items are merged where possible
Old items age faster under lag
Only then are excess items removed
Your server gets cleanup — not an extinction event.
Soft Recovery Logic
As performance improves, GrimLag reverses its own restrictions:
AI returns
Spawners normalize
Redstone resumes full behavior
Despawn rules relax
Nothing stays throttled longer than necessary.
Design Philosophy
Adapt, don’t panic
Throttle, don’t nuke
Recover automatically
Be predictable, not aggressive
GrimLag is built for long-running survival servers, not short-lived benchmarks.
Requirements
Paper (recommended)
Java 17+
评论 0
登录后参与评分
还没有评论,来抢沙发吧