[polskibus]

I know there are many techniques that used together give good performance (optimal memory layout, compression, vectorization, etc. etc.), however I'd like to use a package that does a lot of it, same what RocksDB (or SQLite) does for OLTP cases. Is there something like that? If not, what's out there that gives the best foundation for building OLAP functionalities on top of it?

[gianm]

Check out Druid [1], an open-source analytical database with tightly-coupled storage and processing engines designed for OLAP. In particular it implements a memory-mappable storage format, indexes, compression, late tuple materialization, and query engines that can operate directly on compressed data. There is a patch out to add vectorized processing as well, so you should expect to see that show up in a future release.

Its storage format and processing engine aren't designed to be embedded in the same way as RocksDB and SQLite are, but you certainly could if you wanted to, since the code is fairly modular. Or you could use it as a standalone service as it was designed to be used.

[1] http://druid.io/

[tristor]

There's also Clickhouse [1] which seems to scale much better than Druid, and has similar architectural decisions to make it somewhat general as a columnar store for OLAP uses. Cloudflare wrote an article in the past where they compared Clickhouse and Druid and they chose Clickhouse because they could get similar performance on the same workload with 9 nodes in Clickhouse which would require hundreds for Druid. They built all of the DNS analytics at CloudFlare on Clickhouse [2].

Disclosure: I work at Percona, and we've seen a lot of our customers make use of Clickhouse and have begun some of our own services work around it in Consulting. It's now a primary database talked about at our conferences, and we post about it regularly. [3]

[1]: https://clickhouse.yandex/ [2]: https://blog.cloudflare.com/how-cloudflare-analyzes-1m-dns-q... [3]: https://www.percona.com/blog/2018/10/01/clickhouse-two-years...

[neeleshs]

There is a very good article[1] by one of the Druid committers about Clickhouse/Drui/Pinot that goes into some details on why the Cloudflare tests turned out the way they did.

[1]:https://medium.com/@leventov/comparison-of-the-open-source-o...

[eurg]

That article is better than expected, and it matches my own experience with CH (it was a great match for our use-case, and some of those reasons are in the article; and also, we could have used an inverted index, would one have been available; surprisingly survived w/o it).

[polskibus]

Does ClickHouse support fine grained data security (for example role A gives access only to tuples with column X==123)?

[atombender]

No [1]. ClickHouse is a fairly low-level tool. If you need that kind of thing, you build an ACL-aware app on top of it.

[1] https://clickhouse.yandex/docs/en/operations/access_rights/

[manigandham]

That's what Apache Arrow is, you had the right choice. That solves the processing component and you can use any number of on-disk formats like Parquet and ORC.

And the hybrid of OLAP + OLTP is usually called HTAP.

[polskibus]

Is it possible to insert new tuples to arrow model without rebuilding it from scratch?

[ryanworl]

No, I am not aware of any storage engine that provides that out of the box. The techniques are very tied into what your query processing engine can do and expects the data to look like.

For example, do you materialize tuples immediately, or do you fully run it through your processing pipeline and not materialize until the end?

Your storage engine and format needs to be at least somewhat involved in answer that question, because you need to know what data to read and when.

[arjunnarayan]

Unfortunately most of the systems that build what you're describing are closed source (e.g. Snowflake, Microsoft SQL Server, Vertica, Teradata). There isn't an open-source project that does all of those things.

[georgewfraser]

What about Presto?

[FridgeSeal]

Presto is more of a distributed SQL solution: you run it on a cluster of nodes, point them at your storage later, it’s more optimised at querying very large datasets and it’s not built or tuned for high performance (in terms of latency or execution time).

[arjunnarayan]

Apache Arrow is your best bet, but it's still very much a new project without a lot of the things you're looking for.