linliu-code opened a new pull request, #13946:
URL: https://github.com/apache/hudi/pull/13946
### Describe the issue this Pull Request addresses
For queries that run short, like a few seconds, the impact from latency to
read Hudi metadata could be huge. We have observed that the metadata read
latency could be higher than that of the data read itself. Therefore, reducing
the latency of reading Hudi metadata is important for these queries.
Previously the hfile reader could trigger multiple read for a single hfile
in the MDT, e.g., a read is triggered when a data block is requested. When the
hfile becomes bigger, this could be become a huge problem. An feature that Hudi
introduces to alleviate this problem is caching the entire hfile before reading
it if its size is below a certain threshold.
By default the threshold is 0MB.
### Summary and Changelog
This change Increases the threshold for hfile prefetch threshold from 0 MB
to 50 MB, and really enables hfile prefetching and caching.
The change was tested by using the following query based on a table in 10TB
tpcds dataset. After this change, its latency reduces from ~4s to ~1s.
`spark.read.format("hudi").option("hoodie.enable.data.skipping",
"true").load("s3a://performance-benchmark-datasets-us-west-2/tpc-ds-datasets-hudi/10TB/store_sales/").createOrReplaceTempView("ss1");
spark.time{spark.sql("select count(1) from ss1 where
ss_sold_date_sk='2450818' and ss_sold_time_sk=31367;").count()}`
### Impact
Reduce the MDT read latency and further improve the query performance.
### Risk Level
Low.
### Documentation Update
<!-- Describe any necessary documentation update if there is any new
feature, config, or user-facing change. If not, put "none".
- The config description must be updated if new configs are added or the
default value of the configs are changed.
- Any new feature or user-facing change requires updating the Hudi website.
Please follow the
[instruction](https://hudi.apache.org/contribute/developer-setup#website)
to make changes to the website. -->
### Contributor's checklist
- [ ] Read through [contributor's
guide](https://hudi.apache.org/contribute/how-to-contribute)
- [ ] Enough context is provided in the sections above
- [ ] Adequate tests were added if applicable
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
