hckrnws

Hi HN, I'm the author of this post and a JVM engineer working on OpenJDK.

I've spent the last few years researching GC for my PhD and realized that the ecosystem lacked standard tools to quantify GC CPU overhead—especially with modern concurrent collectors where pause times don't tell the whole story.

To fix this blind spot, I built a new telemetry framework into OpenJDK 26. This post walks through the CPU-memory trade-off and shows how to use the new API to measure exactly what your GC is costing you.

I'll be around and am happy to answer any questions about the post or the implementation!

Thank you for this interface! It will definitely help in tracking down GC related performance issues or in selecting optimal settings.

One thing that I still struggle with, is to see how much penalty our application threads suffer from other work, say GC. In the blog you mention that GC is not only impacting by cpu doing work like traversing and moving (old/live) objects but also the cost of thread pauses and other barriers.

How can we detect these? Is there a way we can share the data in some way like with OpenTelemetry?

Currently I do it by running a load on an application and retaining its memory resources until the point where it CPU skyrockets because of the strongly increasing GC cycles and then comparing the cpu utilisation and ratio between cpu used/work.

Edit: it would be interesting to have the GC time spent added to a span. Even though that time is shared across multiple units of work, at least you can use it as a datapoint that the work was (significantly?) delayed by the GC occurring, or waiting for the required memory to be freed.

Thanks for reading! Your current method, pushing the load until the GC spirals and then comparing the CPU utilization, is exactly the painful, trial-and-error approach I'm hoping this new API helps alleviate.

You've hit on the exact next frontier of GC observability. The API in JDK 26 tracks the explicit GC cost (the work done by the actual GC threads). Tracking the implicit costs, like the overhead of ZGC's load barriers or G1's write barriers executing directly inside your application threads, along with the cache eviction penalties, is essentially the holy grail of GC telemetry.

I have spent a lot of time thinking about how to isolate those costs as part of my research. The challenge is that instrumenting those barrier events in a production VM without destroying application throughput (and creating observer effects) is incredibly difficult. It is absolutely an area of future research I am actively thinking about, but there isn't a silver bullet for it in standard HotSpot just yet.

Something that you could look at there are some support to analyze with regards to thread pauses is time to safepoint.

Regarding OpenTelemetry. MemoryMXBean.getTotalGcCpuTime() is exposed via the standard Java Management API, so it should be able to hook into this.

After writing my previous post I was wondering, do we actually need to instrument the barrier events and other code tied to a GC? Currently we benchmark our application with different GC at different settings and resource constraints and the we pick one sizing and settings combination that we like (read most work/totalcpu that is still fits within the allocation constraints of our clusters). What ultimately matters for production is how the app behaves in production.

This will not help directly when developing new (versions) or GC. On the other hand, if we can have a noop GC including omitting any of the barriers etc required for GC to function we can create a baseline for apps. Provided we have enough total memory to run the benchmark in.

Edit: I guess we can then also use perf to compare cache misses between runs with different GC implementations and settings. Not sure how this works out in real life as it will be very CPU, kernel, and other loads dependent.

The problem is that there is no baseline for measuring GC overhead. You cannot turn it off, you can only replace and compare with different strategies. For example sbrk is technically a noop GC, but that also has overhead and impact because it will not compact objects and give you bad cache behavior. (It illustrates the OP's point that it is not enough to measure pauses, sbrk has no pauses but gets outperformed easily.)

You could stop collecting performance counters around GC phases, but you even if you are not measuring the CPU still runs through its instructions, causing the second order effects. And as you mentioned too-short-to-measure barriers and other bookkeeping overheads (updating ref counters etc) or simply the fact that some tag bits or object slots are reserved all impact performance.

There is a good write-up of the problem and a way to estimate the cost based on different GC strategies, as you suggested, here: https://arxiv.org/abs/2112.07880

The way I found to measure a no-GC baseline is to compare them in an accurate workload performance simulator. Mark all GC and allocator related code regions and have the simulator skip all those instructions. Critically that needs to be a simulator that does not deal with the functional simulation, but gets it's instructions from a functional simulator, emulator or PIN tool that does execute everything. It's laborious, not very fast and impractical for production work. But, it's the only way I found to answer a question like "What is the absolite overhead of memory management in Python?". (Answer: lower bound walltime sits around +25% avg, heavily depending on the pyperformance benchmark)

Hey, noob question, but does OpenJDK look at variable scope and avoid allocating on the heap to begin with if a variable is known to not escape the function's stack frame?

Not strictly related to this post, but I figured it'd be helpful to get an authoritative answer from you on this.

Yes, Hotspot performs Escape Analysis to avoid heap allocation. This is a nice article: https://shipilev.net/jvm/anatomy-quarks/18-scalar-replacemen...

I just want to say this is an incredibly detailed, well written, and beautifully illustrated article. Solid work.

Thanks! I really appreciate that. I spent a lot of time trying to nail the illustrations so I'm really glad it landed well. :-)

Great article!

Will the new metric be exposed in JFR recordings as well?

Thanks!

It is not currently exposed in JFR for JDK 26, but I agree that it would be the logical next step. Now that the underlying telemetry framework (cpuTimeUsage.hpp) is in place within HotSpot, wiring it up to JFR events would be a natural extension.

I built this 15 years ago and it got fairly popular, but is long dead now...

https://github.com/jmxtrans/jmxtrans

Kind of amazing how people are still building telemetry into Java. Great post and great work. Keep it up.

I think a very serious issue with GC is that:

- The number of edges in a graph tend to scale superlinearly with heap size, as the number of edges possible in a graph are quadratic wrt no of objects.

- Memory bandwidth hasn't been scaling very much during the past decade and a half, even compared to memory size. It's also not a thing people think about or even easy to display in any performance monitoring tool.

But considering if you had a machine 15 years ago with 4GB or ram that could be read at 15GB/s, and now you have one with 32GB that can be read at 60GB/s, it means that your bandwidth compared to heap size has halved. Considering the quadratic nature of references, the 'amplification factor', the number of times you have to revisit an already visited block of memory is higher as well.

This is in addition to the cache trashing issues mentioned in the post.

If you need to read the whole heap, this sets a lower bound on how much time the GC will take ~0.25s on the old machine, ~0.5s on the new one.

Suppose your GC triggers a memory bandwidth issue - how do you even profile for that? This is kind of an invisible resource that just gets used up.

At my work, one thing that I've often had to explain to devs is that the Parallel collector (and even the serial collector) are not bad just because they are old or simple. They aren't always the right tool, but for us who do a lot of batch data processing, it's the best collector around for that data pipeline.

Devs keep on trying to sneak in G1GC or ZGC because they hyper focus on pause time as being the only metric of value. Hopefully this new log:cpu will give us a better tool for doing GC time and computational costs. And for me, will make for a better way to argue that "it's ok that the parallel collector had a 10s pause in a 2 hour run".

Every GC algorithm in HotSpot is designed with a specific set of trade-offs in mind.

ZGC and G1 are fantastic engineering achievements for applications that require low latency and high responsiveness. However, if you are running a pure batch data pipeline where pause times simply don't matter, Parallel GC remains an incredibly powerful tool and probably the one I would pick for that scenario. By accepting the pauses, you get the benefit of zero concurrent overhead, dedicating 100% of the CPU to your application threads while they are running.

Gotta be honest, I have a hard time arguing for G1 over ZGC. It seems to me like any situation you'd want G1 you probably want ZGC instead. That default 200ms target latency is already pretty long. If you've made that tradeoff for G1 because you wanted lower latency, you probably are going to be happier with ZGC.

I also find that the parallel collector is often better than G1, particularly for small heaps. With modern CPUs, parallel is really fast. Those 200ms pauses are pretty easy to achieve if you have something like a 4gb heap and 4 cores.

The other benefit of the parallel collector is the off heap memory allocation is quiet low. It was a nasty surprise to us with G1 how much off heap memory was required (with java 11, I know that's gotten a lot better).

We have many apps that run on <1 core just fine for the business logic and run on K8S. If we then use a parallel or concurrent garbage collector it will eat through the cpu limit of the app in a blink causing the process not to be scheduled for several ticks. This introduces more latency than the GC cycles themselves would when using a serial GC than runs frequently enough.

> This freed programmers from managing complex lifecycle management.

It also deceived programmers into failing to manage complex lifecycles. Debugging wasted memory consumption is a huge pain.

Sorry if this is obvious to Java experts, but much as parallel GC is fine for batch workloads, is there a case for explicit GC control for web workloads? For example a single request to a web server will create a bunch of objects, but then when it completes 200ms later they can all be destroyed, so why even run GC during the request thread execution?

There are a few ways of looking at this:

- Purely on the JVM, you probably want ZGC (or Shenandoah) because latency is more important than throughput.

- On Erlang / the BEAM VM, each thread gets its own private heap, so GC is a per thread operation. If the request doesn't spill over the heap then GC would never need to run during a request handler and all memory could be reclaimed when the handler finishes.

- There can still be cases where a request handler allocates memory that is no solely owned by it. E.g. if it causes a new database connection to be allocated in a connection pool, that connection is not owned by the request handler and should not be deallocated when the handler finishes.

- The general idea you're getting at is often called "memory regions": you can point to a scope in the code and say "all the memory can be freed when this scope exits". In this case the scope is the request handler. It's the same idea behind arena or slab memory allocation. There are languages that can encode this, and do safe automatic memory management without GC. Rust is an obvious example, but I don't find it very ergonomic. I think the OxCaml [1] and Scala 3 [2] approaches are better.

[1]: https://oxcaml.org/documentation/stack-allocation/reference/

[2]: https://docs.scala-lang.org/scala3/reference/experimental/cc...

See also arena allocation, for realtime systems. But those systems typically require that any task have a reasonably tight upper bound on memory usage.

Thank you, that’s what I came here to learn!

Most web request cases where you care about performance probably have multiple parallel web requests, so there’s no clean separation possible?

Sure, but each request has its own context. Shared resources like DB connection pools will be longer lived but by definition they aren’t alllcated by the request thread. So why not simply exempt everything allocated by a request thread from GC, and simply destroy it on request completion?

Generational GC assumes that short lived objects tend to come in groups, which is probably the best you can do in an OO language with shared everything.

Go tried that [1], a failed experiment that was a complex NIH version of the generational hypothesis. They currently use a CMS-stye collector.

[1] https://docs.google.com/document/d/1gCsFxXamW8RRvOe5hECz98Ft...

His question is still valid for latency. That parallel GC in Java still seems to pause threads from a quick search. https://inside.java/2022/08/01/sip062/

That's why we got ZGC and Shenandoah, and their generational variants, which have very low pause times (in the order of 1 ms)

Are there plans to elucidate implicit GC costs as well?

Great question! I actually just touched on this in another thread that went up right around the same time you asked this. It is clearly the next big frontier!

The short answer is: It's something I'm actively thinking about, but instrumenting micro-level events (like ZGC's load barriers or G1's write barriers) directly inside application threads without destroying throughput (or creating observer effects invalidating the measurements) is incredibly difficult.

> instrumenting micro-level events (like ZGC's load barriers or G1's write barriers) directly inside application threads without destroying throughput (or creating observer effects invalidating the measurements) is incredibly difficult

I've used a sampling profiler with success to find lock contention in heavily multithreaded code, but I guess there are some details that makes it not viable for this?

Do you think it can be done by adjusting GC aggressiveness (or even disabling it for short periods of time) and correlating it with execution time?

That is spot on. Effectively disabling GC to establish a baseline is exactly the methodology used in the Blackburn & Hosking paper [1] I referenced.

In general, for a production JVM like HotSpot, the implicit cost comes largely from the barriers (instructions baked directly into the application code). So even if we disable GC cycles, those barriers are still executing.

If we were to remove barriers during execution, maintaining correctness becomes the bottleneck. We would need a way to ensure we don't mark a live (reachable) object as dead the moment we re-enable the collector.

[1] https://dl.acm.org/doi/pdf/10.1145/1029873.1029891

Would running an application with chosen GC, subtracting GC time reported by methods You introduced, and then comparing with Epsilong-based run be a good estimate of barrier overhead ?

Thank you for the well written article!

That is a creative idea, but unfortunately, Epsilon changes the execution profile too much to act as a clean baseline for barrier costs.

One huge issue is spatial locality. Epsilon never reclaims, whereas other GCs reclaim and reuse memory blocks. This means their L2/L3 cache hit rates will be fundamentally different.

If you compare them, the delta wouldn't just be the barrier overhead; it would be the barrier overhead mixed with completely different CPU cache behaviors, memory layout etc. The GC is a complex feedback loop, so results from Epsilon are rarely directly transferable to a "real" system.