So is this similar to Amazon's Athena? I'm trying to place what a 'realtime distributed OLAP datastore' is, or competes with, in cloudy/naive terms.
https://startree.ai/blog/a-tale-of-three-real-time-olap-data...
LinkedIn originally had been using traditional OLTP systems to power the "Who Viewed My Profile" app; they just simply hit the wall. So LinkedIn looked at other analytics databases at the time. Most just couldn't handle a large number of QPS — which, as a social media platform, they readily anticipated. This is why they defined the category as "user-facing, real-time analytics."
[This is terminology mostly specific to the Apache Pinot crowd, though I see that StarRocks / CelerData also recently started talking about user-facing analytics. So I wrote up an article explaining it here:
https://startree.ai/resources/what-is-user-facing-analytics ]
Other similar extant systems LinkedIn benchmarked at the time just couldn't give them the numbers they needed at the low latency, high concurrency and large scale they anticipated — terabytes to petabytes of data. So they wrote their own solution.
Pinot was originally intended for marketing purposes to capture live intent & action data.
The same Pinot infrastructure eventually grew to other real-time use cases. "Who Viewed My Profile" was followed by "Company Follow Analytics," then sales or recruiting, and even internal A/B testing.
[I wasn't at LinkedIn while this happened; this lore was passed down to me by others. Disclosure: yes I work at StarTree.
https://engineering.linkedin.com/analytics/real-time-analyti... ]
Most of these use cases seem more “pull all data for a specific user/post/company” and less “do analytical queries on millions+ of column values.”
https://engineering.linkedin.com/blog/2019/06/star-tree-inde...
Anecdote: I tested ClickHouse as possible replacement of Trino+DeltaLake+S3 for some of our use cases.
When querying a precomputed flat tables it was easily 10x to 100x faster. When running complex ETL to prepare those tables, I gave up when CTAS with six CTEs that takes Trino 30 seconds to compute on the fly turned into six intermediate tables that took 40 minutes to compute, and I wasn't even halfway done.
The tricky part is, how do you know whether your use case fits? But you have to ask this question about all the "specialized tools", including Pinot.
https://startree.ai/blog/query-time-joins-in-apache-pinot-1-...
[Edit: if users just naively throw query-time JOINs at a problem, they might not get results in the time they want — a non-optimized real-time JOIN took upwards of 39 seconds. With predicate pushdowns, and partition-aware JOINs, suddenly results could be done in 3.3 seconds — faster by an order of magnitude. Still kinda long for a typical page or mobile app refresh but survivable. And yes, even faster subsecond results were possible with additional compute parallelism, but that has a cloud services cost associated with it.
Pre-ingestion JOINs, such as through Apache Flink, would be generally more performant than query-time JOINs. And that's how most people do them right now anyway. So just because you can do query-time JOINs doesn't mean you should if you haven't thought about it ahead of time. If you can, optimize the data partioning to ensure best performance. The good news is that users now have a lot more flexibility in how they want to do JOINs.]
but that benchmark is 250G, plus all established players I think support joins well, I work with BQ closely and it works very well.
> So just because you can do query-time JOINs doesn't mean you should if you haven't thought about it ahead of time.
prebuilding denormalized table can significantly increase your datasize, so it is tradeoff and depends on your data structure and cardinality.
The 1.0 version of Pinot seems to bring a lot of maturity, they seem to have added new engine that can do joins now. I'm not sure how stable it is, but it seems interesting.
As for what is this kind of database usedful for, this is for operational analytics on large data that also update in real time. In my domain that would be things like having insight into large supply chains or manufacturing operations, like power plants or factories, just in general for monitoring stuff. I know it's also used in security and finance (for fraud).
This is a completely different beast in terms of expected query latency (sub-second) and as a result has different constraints about structure of data, amount of data queried and cost naturally.
You can think of Pinot and Druid as solving for the real-time analytics cases, like say for instance you run an ad network and you want to provide your customers with a live dashboard of ad impressions and allow segmentation of that dashboard by cohorts and other metrics on the fly.