search

apache / incubator-hugegraph-computer

HugeGraph Computer - A distributed graph processing system for hugegraph (OLAP)
https://hugegraph.apache.org/docs/quickstart/hugegraph-computer/
Apache License 2.0
42 stars 41 forks source link

[Feature] support for Java version >11 #309

Open moonbaseDelta opened 1 month ago

moonbaseDelta commented 1 month ago

Feature Description (功能描述)

On newest Debian and other systems it sometimes impossible to install Java 11/8. But current compilation fails to build on Java 17 due to the dependencies.

imbajin commented 1 month ago

On newest Debian and other systems it sometimes impossible to install Java 11/8. But current compilation fails to build on Java 17 due to the dependencies.

Thanks for your suggestion, we plan to support java17 in the 2.0 version (Also, we provide a in-memory graph-computer implement (written by Go) -- and it will be public soon, it will be more user friendly to use with the small and medium-sized datasets (less than 10 billion vertexes & edges) )

moonbaseDelta commented 1 month ago

On newest Debian and other systems it sometimes impossible to install Java 11/8. But current compilation fails to build on Java 17 due to the dependencies.

Thanks for your suggestion, we plan to support java17 in the 2.0 version (Also, we provide a in-memory graph-computer implement (written by Go) -- and it will be public soon, it will be more user friendly to use with the small and medium-sized datasets (less than 10 billion vertexes & edges) )

Sounds cool, I've read the contest page and will be quite excited to see 2.0 in action!

As for billions, it seems that cluster setup with pd-store will be the way. At least current docs\demos for cluster mode and roles is not so handy as large-single-node setup - but for hundreds of billions edges and more it's hardware bounded.

imbajin commented 1 month ago

On newest Debian and other systems it sometimes impossible to install Java 11/8. But current compilation fails to build on Java 17 due to the dependencies.

Thanks for your suggestion, we plan to support java17 in the 2.0 version (Also, we provide a in-memory graph-computer implement (written by Go) -- and it will be public soon, it will be more user friendly to use with the small and medium-sized datasets (less than 10 billion vertexes & edges) )

Sounds cool, I've read the contest page and will be quite excited to see 2.0 in action!

As for billions, it seems that cluster setup with pd-store will be the way. At least current docs\demos for cluster mode and roles is not so handy as large-single-node setup - but for hundreds of billions edges and more it's hardware bounded.

Yes, indeed the graph-computer could upgrade to support Java17/21 easily due to there isn't too much dependency burden.

U can try upgrading its compilation and runtime versions on your own. I don't think there should be many issues with the adaptation, but the community plan is to support Java17 as a whole until version 2.0, so there won't be a single component to be upgraded ahead of time (Not a technical obstacle)

In addition, the upcoming graph computing can be understood as a Go version of a new computing model (GAS model, where data is stored/computed in memory first, but also supports storing excess data to disk, similar to the combination like Redis+RocksDB. Its code is basically ready and waiting for complete documentation and other preparation to be used.

It provides simple binary boot and multiprocessing clustering (independent of k8s/yarn, easily to start), which is more suitable for the environment and needs of most ordinary users

imbajin commented 4 days ago

FWD: The new in-memory/high-performance computer system code in #311