Database-like ops benchmark (original) (raw)

The code for this benchmark can be found at https://github.com/duckdblabs/db-benchmarkand has been forked from https://github.com/h2oai/db-benchmark.

This page aims to benchmark various database-like tools popular in open-source data science. It runs whenever a PR is opened requesting an update, provided the PR author has ran the benchmark themselves. We provide this as a service to both developers of these packages and to users. You can find out more about the project in Efficiency in data processing slides and talk made by Matt Dowle on H2OWorld 2019 NYC conference.

We also include the syntax being timed alongside the timing. This way you can immediately see whether you are doing these tasks or not, and if the timing differences matter to you or not. A 10x difference may be irrelevant if that’s just 1s vs 0.1s on your data size. The intention is that you click the tab for the size of data you have.

Machine Type

Small (16 cores / 32GB memory)

Task

groupby

0.5 GB

basic questions

advanced questions

5 GB

basic questions

advanced questions

50 GB

basic questions

advanced questions

join

0.5 GB

basic questions

5 GB

basic questions

50 GB

basic questions

X-Large (128 cores / 256GB memory)

Task

groupby

0.5 GB

basic questions

advanced questions

5 GB

basic questions

advanced questions

50 GB

basic questions

advanced questions

join

0.5 GB

basic questions

5 GB

basic questions

50 GB

basic questions

Details

groupby

Timings are presented for a single dataset case having random order, no NAs (missing values) and particular cardinality factor (group size question 1 k=100). To see timings for other cases go to the very bottom of this page.

groupby timings

join

Timings are presented for datasets having random order, no NAs (missing values). Data size on tabs corresponds to the LHS dataset of join, while RHS datasets are of the following sizes: small(LHS/1e6), medium (LHS/1e3), big (LHS). Data case having NAs is testing NAs in LHS data only (having NAs on both sides of the join would result in many-to-many join on NA). Data case of sorted datasets tests data sorted by the join columns, in case of LHS these are all three columns id1, id2, id3 in that order.

join timings

Errors

Since adding the new machine type, many updates need to be made to report generation in order to accommodate the extra dimension of the benchmark. One area where this maintenance has not yet occured is with error reporting. Please be patient while this is worked on. Many error “Not Tested” errors are placeholders for “Out of Memory” errors. I can re-run certain benchmarks on request, but please double check that they will not run out of memory. If you see an error that doesn’t make sense and you are sure is not an out of memory error feel free to file an issue.

Requesting an updated run

The benchmark will now be updated with PR requests. To publish new results for a solution(s), you can open a PR with changes to solutions scripts or VERSION files, with updates to the time.csv and log.csv files of a run on a c6id.metal machine or c6id.4xlarge. If using a c6id.4xlarge, we recommend a dedicated instance, to avoid variable performance due to noisy neighbors. To facilitate creating an instance identical to the one with the current results, the script_setup_utils/mount.sh and_setup_utils/setup_small were created. The script does the following

1. Formats and mounts an nvme drive so that solutions have access to instance storage. This prevents variability in performance due to network storage.
2. Creates a new directory db-benchmark-metal on the nvme drive. This directory is a clone of the repository

Once the db-benchmark-metal directory is created, you will need to 1. Install the solutions you wish to have updated 2. Update the solution(s) groupby or join scripts with any desired changes 3. Modify the run.conf file so that your desired solution/task is run. 4. Run the benchmark on your solution by calling _run/partitioned_run.sh (note this will download the data and then delete it after the run. If you wish to have the benchmark data persist, comment out the lines removing the dta.) 5. Generate the report to see how the results compare to other solutions 6. Create your PR! (make sure the new time.csv and logs.csv files are included!)

The PR will then be reviewed by the DuckDB Labs team where we will run the benchmark ourselves to validate the new results. If there aren’t any questions, we will merge your PR and publish a new report!

Notes

• You are welcome to run this benchmark yourself! All scripts related to setting up environment, data and benchmark are in the repository repository in the_setup_utils directory. Please remember to run_setup_utils/mount.sh so you know you are running on the instance storage and not network backed storage.
• Data used to generate benchmark plots on this website can be obtained from time.csv (together with logs.csv). See _report/report.Rfor quick introduction how to work with those. You can also download them from aws s3 without requiring any credentials.
• Solutions are using in-memory data storage to achieve best timing. In case a solution runs out of memory (for the smaller machine), it will use nvme storage if correctly set up. In these cases, the solution name is denoted by a * suffix on the legend.
• ClickHouse and DuckDB queries areCREATE TABLE ans AS SELECT ... to match the functionality provided by other solutions in terms of caching results of queries, see#151.
• We ensure that calculations are not deferred by solution.
• Because of the above, as of current moment, join timings of python datatable suffers from an extra deep copy. As a result of that extra overhead it suffers additionally with out of memory error for 1e9 join q5 big-to-big join.
• We also tested that answers produced from different solutions match each others, for details see _utils/answers-validation.R.

Environment configuration

• R 4.3.2
• python 3.10
• Julia 1.9.3

Scope

We limit the scope to what can be achieved on a single machine. Laptop size memory (8GB) and server size memory (250GB) are in scope. Out-of-memory using local disk such as NVMe is in scope. Multi-node systems such as Spark running in single machine mode is in scope, too. Machines are getting bigger: EC2 X1 has 2TB RAM and 1TB NVMe disk is under $300. If you can perform the task on a single machine, then perhaps you should. To our knowledge, nobody has yet compared this software in this way and published results too. Some solutions are not run on smaller machines. You can request a run for your solution, but there is no guarantee when the run will take place.

Why db-benchmark?

Because we have been asked many times to do so, the first task and initial motivation for this page, was to update the benchmark designed and run by Matt Dowle(creator of data.table) in 2014here. The methodology and reproducible code can be obtained there. Exact code of this report and benchmark script can be found at h2oai/db-benchmarkcreated by Jan Goreckifunded by H2O.ai. H2O.ai stopped supporting the benchmark in 2021. In 2023, DuckDB Labs decided to start maintaining the benchmark, and code can be found at duckdblabs/db-benchmark.

Explore more data cases

Benchmark run took around 58.9 hours.

Report was generated on: 2025-02-17 12:20:42 UTC.