Tag: azure

Testing TPC-DS 10GB Hosted in Fabric OneLake Using Python Data Engines

This is not an official benchmark—just an exercise to experiment with the new Fabric Python notebook.

You can download the notebook and the results here

There is a growing belief that most structured data will eventually be stored in an open table format within object stores, with users leveraging various engines to query that data. The idea of data being tied to a specific data warehouse (DWH) may soon seem absurd, as everything becomes more open and interoperable.

While I can’t predict the future, 2024 will likely be remembered as the year when the lakehouse concept decoupled from Spark. It has become increasingly common for “traditional” DWHs or any Database for that matter to support open table formats out of the box. Fabric DWH, for instance, uses a native storage layer based on Parquet and publishes Delta tables for consumption by other engines. Snowflake now supports Iceberg, and BigQuery is slowly adding support as well.

I’m not particularly worried about those DWH engines—they have thousands of engineers and ample resources, they will be doing just fine.

My interest lies more in the state of open source Python engines, such as Polars and DataFusion, and how they behave with a limited resource environment.

Benchmarking Bias

Any test inherently involves bias, whether conscious or unconscious. For interactive queries, SQL is the right choice for me. I’m aware of the various DataFrame APIs, but I’m not inclined to learn a new API solely for testing. For OLAP-type queries, TPC-DS and TPC-H are the two main benchmarks. This time, I chose TPC-DS for reasons explained later.

Benchmark Setup

All data is stored in OneLake’s Melbourne region, approximately 1,400 km away from my location, the code will check if the data exists otherwise it will be generated, the whole thing is fully reproducible.

I ran each query only once, ensuring that the DuckDB cache, which is temporary, was cleared between sessions. This ensures a fair comparison.

I explicitly used the smallest available hardware since larger setups could mask bottlenecks. Additionally, I have a specific interest in the Fabric F2 SKU.

While any Python library can be used, as of this writing, only two libraries—DuckDB and DataFusion—support:

  • Running the 99 TPC-DS queries (DataFusion supports 95, which is sufficient for me).
  • Native Delta reads for abfss or at least local paths.
  • Python APIs, as they are required to run queries in a notebook.

Other libraries like ClickHouse, Databend, Daft, and Polars lack either mature Delta support or compatibility with complex SQL benchmarks like TPC-DS.

Why TPC-DS ?

TPC-DS presents a significantly greater challenge than TPC-H, with 99 queries compared to TPC-H’s 22. Its more complex schema, featuring multiple fact and dimension tables, provides a richer and more demanding testing environment.

Why 10GB?

The 10GB dataset reflects the type of data I encountered as a Power BI developer. My focus is more on scaling down than scaling up. For context:

  • The largest table contains 133 million rows.
  • The largest table by size is 1.1GB.

Admittedly, TPC-DS 10GB is overkill since my daily workload was around 1GB. However, running it on 2 cores and 16GB of RAM highlights DuckDB’s engineering capabilities.

btw, I did run the same test using 100GB and the python notebook with 16 GB did works just fine, but it took 45 minutes.

OneLake Access Modes

You can query OneLake using either abfss or mounted storage. I prefer the latter, as it simulates a local path and libraries don’t require authentication or knowledge of abfss. Moreover, it caches data on runtime SSDs, which is an order of magnitude faster than reading from remote storage. Transactions are also included in the base capacity unit consumption, eliminating extra OneLake costs.

It’s worth noting that disk storage in Fabric notebook is volatile and only available during the session, while OneLake provides permanent storage.

You can read more about how to laverage DuckDB native storage format as a cache layer here

Onelake Open internet throughput

My internet connection is not too bad but not great either, I managed to get a peak of 113 Mbps, notice here the extra compute of my laptop will not help much as the bottleneck is network access.

Results

The table below summarizes the results across different modes, running both in Fabric notebooks and on my laptop.

  • DuckDB Disk caching yielded the shortest durations but the worst individual query performance, as copying large tables to disk takes time.
  • Delta_rs SQL performance was somewhat erratic.
  • Performance on my laptop was significantly slower, influenced by my internet connection speed.
  • Mounted storage offered the best overall experience, caching only the Parquet files needed for queries.

And here is the geomean

Key Takeaways

  • For optimal read performance, use mounted storage.
  • For write operations, use the abfss path.
  • Having a data center next to your laptop is probably a very good idea 🙂

Due to network traffic, Querying inside the same region will be faster than Querying from the web (I know, it is a pretty obvious observation)

but is Onelake throughput good ?

I guess that’s the core question, to answer that I changed the Python notebook to use 8 cores, and run the test from my laptop using the same data stored in my SSD Disk, no call to onelake, and the results are just weird

Reading from Onelake using mounted storage in Fabric Notebook is faster than reading the same data from my Laptop !!!!

Looking Ahead to 2025

2024 has been an incredible year for Python engines, evolving from curiosities to tools supported by major vendors. However, as of today, no single Python library supports disk caching for remote storage queries. This remains a gap, and I hope it’s addressed in 2025.

For Polars and Daft, seriously works on better SQL support

Building an Ad Hoc Disk Cache with DuckDB and Fabric Notebook

This weekend, I came up with an idea to speed up query execution when running DuckDB inside a Fabric Notebook—and it actually works! 🎉

You can download the notebook here

Approach

  1. Parse the Query
    • Use SQGLot to parse the SQL query and extract the list of Delta tables that need to be scanned from OneLake.
  2. Track Table Metadata
    • Capture the Delta table version and ID to ensure consistency.
  3. Selective Copy
    • Copy only the necessary tables locally to satisfy the query.
  4. Reuse Cached Data
    • For subsequent queries, check if the Delta table has changed:
      • If unchanged, read data from the local SSD.
      • If new tables are required, repeat the caching process for those tables.

Why It Works

This approach effectively creates a temporary, ad hoc disk cache in the notebook. The cache:

  • Persists only for the session’s duration.
  • Evicts automatically when the session ends.
  • Ensures consistency by validating whether the Delta table in OneLake has changed before reusing cached data.
    • Thanks to the Delta format, this validation is a relatively cheap operation.
  • Leverages the user-level isolation in Fabric notebooks to eliminate risks of data inconsistency.

Despite its simplicity, this method has proven to be highly effective for query acceleration! 🚀

Limitations

Yes, I know—the cache is rather naïve since it loads the entire table. Other systems go further by:

  • Copying only the columns needed for the query.
  • Fetching just the row groups relevant to the query.

However, these optimizations would need to be implemented natively by the engine itself.

Industry Gap

Although virtually all Python engines (e.g., Polars, DataFusion, etc.) support reading formats like Delta and Iceberg, almost none offer built-in disk or RAM caching. This lack of caching support limits performance optimization opportunities.

Hopefully, this will change in the future, enabling more efficient workflows out of the box.

Btw, this is really fast !!! just a hint, this is faster than the results obtained by a state of the art DWH in 2022 !!!

Process 1 Billion rows of raw csv in Fabric Notebook for less than 20 Cents 

The Use case

Data source is around 2200 files with a total of 897 Million rows of weird csv files (the file has more columns than the header) , This is a real world data not some synthetic dataset, it is relatively small around 100 GB uncompressed.

The Pipeline will read those files and extract clean data from it using non trivial transformation and save it as a Delta Table.

we used the smallest Compute available in Fabric Notebook which is 4 cores with 32 GB. to be clear this is a real single node (not 1 driver and 1 executor), Although the Runtime is using Spark, All the Engines interact Directly with the Operating system, as far as I can tell, Spark has a very minimum overhead when not used Directly by the Python code.

You need to pick the Engine

Nowadays we have plenty of high quality Calculation Engines,  but two seems to gain traction (Polars and DuckDB) , at least by the number of package downloaded and the religious wars that seems to erupt occasionally in twitter 🙂

For a change I tried to use Polars, as I was accused of having a bias toward DuckDB, long story short, hit a bug with Polars , I tried Datafusion too but did managed to get a working code, there is not enough documentation on the web, after that I did test Clickhouse chdb, but find a bug, anyway the code is public, feel free to test your own Engine.

So I ended up using DuckDB, the code is published here , it is using only 60 files as it is available publicly, the whole archive is saved in my tenant (happy to share it if interested) 

The results is rather surprising (God bless Onelake throughput), I am using the excellent Python Package Delta Lake to write to Onelake

26 minutes, that’s freaking fast, using Fabric F2, the total cost will be

0.36 $/Hour X(26/60) =  15 Cents

you need to add a couple of cents for Onelake Storage Transactions.

As far as I can tell, this is maybe one of the cheapest option in the Market.

0.36 $/Hour is the rate for pay as you go, if you have a reservation then it is substantially cheaper.

because it is Delta Table Then Any Fabric Engine ( SQL, PowerBI, Spark) can read it.

What’s the catch ?

Today DuckDB can not write directly to Delta Table ( it is coming though eventually) instead it will export data to Delta Lake writer using Arrow Table, it is supposed to be zero copy but as far as I can tell, it is the biggest bottleneck and will generate out of memory errors , the solution is easy ; process the files in chunks , not all at once

#############################################
list_files=[os.path.basename(x) for x in glob.glob(Source+'*.CSV')]
files_to_upload_full_Path = [Source + i for i in list_files]
if len(files_to_upload_full_Path) >0 :
  for i in range(0, len(files_to_upload_full_Path), chunk_len):
    chunk = files_to_upload_full_Path[i:i + chunk_len]
    df=get_scada(chunk)
    write_deltalake("/lakehouse/default/Tables/scada_duckdb",df,mode="append",engine='rust',partition_by=['YEAR'],storage_options={"allow_unsafe_rename":"true"})
    del df

By experimentation, I notice 100 files works fine with 16 GB, 200 files with 32 GB etc

When exporting to Parquet, DuckDB managed the memory natively and it is faster too.

Native Lakehouse Is the future of Data Engineering

The combination of Open table format like Delta and Iceberg with ultra efficient Open Source Engine like DuckDB, Polars, Velox, datafusion all written in C++/Rust will give data engineers an extremely powerful tools to build more flexible and way cheaper data solutions.

if I have to give an advice for young Data engineers/Analysts, Learn Python/SQL.

Would like to thanks Pedro Holanda for fixing some very hard to reproduce bugs in the DuckDB csv reader.

And Ion Koutsouris for answering my silly questions about Delta lake writer.