Memory footprint of a Java process by Andrei Pangin

Рет қаралды 34,570

4 жыл бұрын

Does JVM option '-Xmx4g' mean that the process will consume up to 4 GB RAM? Certainly not. What else can take virtual memory and how much of it?
The question becomes especially important when running Java in a shared environment or in a container with limited resources. Excessive memory consumption may occur everywhere: from the application code and the libraries to the Java virtual machine and the operating system. While Java memory leaks are usually easy to discover from a heap dump, native memory leaks can be quite confusing.
During this session we will discuss what structures contribute to the JVM footprint. We will study the real cases of native memory leaks and explore the tools useful in memory allocation analysis.
Andrei Pangin leads the development of Odnoklassniki social network, specializing in high performance Java servers. He previously worked on HotSpot JVM, which became his favorite topic and area of expertise. Andrei is also known as top #JVM answerer on Stack Overflow and the author of Async-profiler.

Пікірлер: 24

@SirWolf2018 4 жыл бұрын

This is so useful, I can't thank you enough!

@FernandoRacca 3 жыл бұрын

that's what a deep dive looks like. EXCELLENT presentation

@amez44 2 жыл бұрын

This is one of the best presentations I’ve seen, great work, thank you very much

@leozilla Жыл бұрын

The best video about java memory consuption I have seen so far! Thanks a lot for preparing this great talk!

@sgerardpandian 3 жыл бұрын

We had been facing issues in production instance of Oracle Service Bus and Weblogic server. Omg this post cleared up so many misconceptions and doubts that we have been going through in our troubleshootings. I really cant thank you enough, this might look simple but it is such deep stuff that application server admins rarely have exposure to. Kudos sir!

@YousufZiaTheOneAndOnly 2 жыл бұрын

Incredible! Loved this talk!

@cheriviralakiran 7 ай бұрын

This is very good coverage in one hour

@sameersarmah1446 3 жыл бұрын

Immensely informative session. After watching one would be able to perform in-depth analysis of different sections of native memory. Grateful for such awesome session.

@marcellolorenzi8401 11 ай бұрын

Amazing presentation! Well done!

@ravipasumarthy8732 2 жыл бұрын

This is one of the excellent talks

@dhawaljoshi Жыл бұрын

Extremely informative, thanks!

@kevinwsewell 2 жыл бұрын

Thank you soooo much for this video, it was a life saver!!!

@Natalicha1234 4 жыл бұрын

Круто, спасибо!

@Capibara1 2 жыл бұрын

Great talk!

@marceloortizdesantana9952 3 жыл бұрын

Thanks! Help a lot!!!

@thunkwaltz5571 2 жыл бұрын

Wonderful presentation with great insights, thanks a lot. Regarding adaptive policy, I was reading at some place that in case of Parallel GC, enabling adaptive policy change heap generation sizes dynamically to meet latency and throughput Goals. Can you comment on that?

@Gabriel-sc3mg 2 ай бұрын

i'm facing a huge problem of memory increasing and it does not seen to be the heap i'll use this material to try to find ou what could be. thank you very much

@desigooglycom 2 ай бұрын

Hi Andrei our application on Kubernetes is consuming lot of page cache, the page cache continues to grow and ultimately container hits memory limit, what should we do?

@berndeckenfels Жыл бұрын

Another big footprint (at least for large heaps without large pages is OS Pagetables. This only shows up in Linux slab info (fortunately it’s very linear to application page count).