spark guide for Project Infinity 0.1 0.0.48.2 to 0.0.49.3

Same as in the All the Mods 10: To the Sky 2.0.4 guide

spark is different from most mods you'll meet in this pack: there's nothing to craft and no machine to place. It's a performance profiler for Minecraft clients, servers and proxies, so it helps you find out why the game is lagging.^[1]

The mod page says spark needs no configuration or setup, you just install it. That makes your first steps more about knowing what it can do than building anything.^[1]

What you can do with it

1. Start with the CPU profiler when you notice lag, a low tick rate or high CPU usage. Those are the problems it's meant to diagnose.^[1]
2. Let the profiler run for about 30 seconds. The mod page says that's enough to give useful insight into the problem areas.^[1]
3. Read the results as a call graph. spark samples what the system is doing, builds the graph from that, and shows it in an online viewer for you to analyse.^[1]
4. Check overall health. spark can report TPS, tick durations (min, max and average), CPU usage, memory usage and disk usage.^[1]
5. If you suspect a memory problem, use the memory tools: a heap summary, a full heap dump, and garbage collection monitoring.^[1]

The mod page also mentions a /tps command for checking ticks per second more accurately. It is the only command named in the sources I have.^[1]

Unverified This section describes spark as its project page does. The sources I have show no pack-specific changes, extra commands or permission settings for spark in this pack, so check in-game before relying on any of those.

Same as in the All the Mods 10: To the Sky 2.0.4 guide

spark works differently from most mods you'll meet in a modpack. It's a performance profiler for Minecraft clients, servers and proxies, so this section covers how its diagnostic tools work, not machines, power or automation.^[1]

It's made up of three separate components: a CPU profiler, memory inspection tools and server health reporting. Install it by adding the jar to the mods directory.^[1]

CPU profiler

The profiler is for diagnosing performance problems such as lag, low tick rate and high CPU usage. It works by sampling statistical data about system activity and building a call graph from it. That graph is then shown in an online viewer, where you can analyse it as a simple tree structure. The viewer can also apply deobfuscation mappings.^[1]

The project page describes the profiler like this:^[1]

• Lightweight: it can be run in production with minimal impact.^[1]
• Easy to use: no configuration or setup is necessary, you just install it.^[1]
• Quick: running for about 30 seconds is enough to give useful insight into problem areas.^[1]
• Customisable: you can target specific threads, sample at a specific interval, or record only "laggy" periods.^[1]

Profiler engines

There are two profiler engines. The first uses native AsyncGetCallTrace and perf_events through async-profiler, and it's only available on Linux x86_64 systems. The second uses the built-in Java ThreadMXBean, an improved version of the WarmRoast profiler by sk89q.^[1]

Memory inspection

For memory problems, spark includes three tools:^[1]

• Heap Summary: a basic snapshot of the server's memory. It shows memory usage and instance counts for each class in the JVM's heap. It's not meant to replace proper memory analysis tools.^[1]
• Heap Dump: a full HPROF snapshot of the JVM's heap, optionally compressed. You can inspect it with conventional analysis tools.^[1]
• GC Monitoring: watches garbage collection activity so you can relate it to server hangs and see how long collections take and how much memory they free. You can also observe how often young and old generation collections happen and how long they last, which helps you pick GC tuning flags.^[1]

Server health reporting

spark can report metrics that summarise the overall health of the server:^[1]

• TPS: ticks per second, with the /tps command giving a more accurate figure.^[1]
• Tick durations: how long each tick takes (min, max and average).^[1]
• CPU usage: how much CPU the server process uses, and how much the overall system uses.^[1]
• Memory usage: how much memory the process is using.^[1]
• Disk usage: how much disk space is free or in use on the system.^[1]

Besides tick rate averages, spark can monitor individual ticks and send a report whenever a single tick's duration goes over a set threshold. That helps you spot trends and the nature of a performance problem relative to other system or game events.^[1]