Enabling rubix for trino-iceberg

[rupal-s]

Description

Enabling rubix caching for trino-iceberg connector. With this change, we can use properties like hive.cache.enabled,
hive.cache.location etc to cinfigure the iceberg cache

Is this change a fix, improvement, new feature, refactoring, or other?
Improvements
Is this a change to the core query engine, a connector, client library, or the SPI interfaces? (be specific)
trino-iceberg
How would you describe this change to a non-technical end user or system administrator?
Enabling rubix caching for trino-iceberg connector.

Related issues, pull requests, and links

Documentation

( ) No documentation is needed.
( ) Sufficient documentation is included in this PR.
( ) Documentation PR is available with #prnumber.
( ) Documentation issue #issuenumber is filed, and can be handled later.

Release notes

( ) No release notes entries required.
( ) Release notes entries required with the following suggested text:

# Section
* Fix some things. ({issue}`issuenumber`)

[cla-bot]

Thank you for your pull request and welcome to our community. We require contributors to sign our Contributor License Agreement, and we don't seem to have you on file. In order for us to review and merge your code, please submit the signed CLA to cla@trino.io. For more information, see https://github.com/trinodb/cla.

[findepi]

I remember @raunaqmorarka expressing some strong concerns about this idea https://trinodb.slack.com/archives/CJ6UC075E/p1650894628024159?thread_ts=1650888689.756519&cid=CJ6UC075E

The main problem I see is that upstream Rubix is an unmaintained project at this point. It works fine in tests, however there have been reports of bugs once in a while at scale from actual users.
So if you run into any issues with it which can't be worked around using configs or from Trino side, it's unlikely to receive a code fix.

[rupal-s]

Hi @findepi / @raunaqmorarka ,
Thank you for the quick feedback.
I have gone through the thread referenced, and saw the following:

There are thoughts about building caching into Trino natively, however there is no concrete plan at the moment.
You could look into Varada's caching solution and see if that solves your needs

The above comment was in April, so I wanted to check if there are now any concrete plans of building a native cache, or maybe even moving to alluxio etc.
We need a data cache for the iceberg connector as object stores tends to degrade and even throttle on very high loads. Currently we are facing poor performance with large data sizes due to degradation in s3 latencies (sometimes calls even throttle)

[rukartok]

@rupal-s @findepi
this seems to be a working solution (I did the same change in my private branch and tested it) - what are any other concerns preventing this from check-in?

[raunaqmorarka]

@rupal-s @findepi
this seems to be a working solution (I did the same change in my private branch and tested it) - what are any other concerns preventing this from check-in?

The main problem IMO continues to be :
Upstream Rubix is an unmaintained project at this point. It works fine in tests, however there have been reports of bugs once in a while at scale from actual users.
So if you run into any issues with it which can't be worked around using configs or from Trino side, it's unlikely to receive a code fix.

I would prefer to not increase dependence on an unmaintained external project.

[rukartok]

@rupal-s @findepi
this seems to be a working solution (I did the same change in my private branch and tested it) - what are any other concerns preventing this from check-in?

The main problem IMO continues to be : Upstream Rubix is an unmaintained project at this point. It works fine in tests, however there have been reports of bugs once in a while at scale from actual users. So if you run into any issues with it which can't be worked around using configs or from Trino side, it's unlikely to receive a code fix.

I would prefer to not increase dependence on an unmaintained external project.

The dependency is already there (Rubix is already used in Hive connector).
Assuming that Rubix is buggy also should not be a problem for Iceberg as this is controlled by configuration parameter (so it can be turned off if an issue occurs).

[hackeryang]

The replacement of Rubix: #18719

[wendigo]

Rubix is going to be deprecated and removed once new Alluxio lands

[mosabua]

#20102

[rupal-s]

@mosabua will Alluxio be present on Trino iceberg connector as well?

[mosabua]

Yes - see the code and docs PRs.

• Alluxio cache #18719
• Add docs for filesystem caching #20300