Category: DuckDB

Develop Python Notebook in your laptop and use Fabric only for Production

TL;DR: wrote my first fabric notebook at work ( download some BOM data) and for some reason, I did find that working first in my laptop using VS code then deploying later in Fabric seemed a more natural workflow to me, maybe working with PowerBI desktop for years has something to do with it, you can download the notebook here

You can  just use pure Python in Fabric Notebook.

Fabric Notebook support Pyspark and R but vanilla Python works just fine, the only big limitation writing to Delta is currently in developpement , but there is a workaround, I created a function that try first to write using Spark, if not available it will write using Delta Python which works fine in my laptop, in that case the code will works in Fabric Notebook without any modification

Why Not PySpark ?

Obviously if you have Pyspark installed in your laptop and you are familiar with the syntax then go for it,  you can even make the argument it is a long term career advantage to be skilful in it, the catch is, I don’t see how I can install it in my laptop and to be honest, I feel DuckDB or pandas for that matter is substantially friendlier to use.

Python is very good to download public Data

I use a Python function to download data from BOM website to  csv files in a  local folder, I copied the same Folder Path as Fabric Lakehouse

I hope the code keeps working, I justed copied some code from the internet and mess with it till it worked, I have no idea how regex is working, and I don’t even want to know

DuckDI API has a vibe of PowerQuery

I used DuckDB to clean the data, you can use Pandas, Polars, Ibis or any library that you like, personally I never liked CTE in SQL, I much prefer step by step logic and seeing the results in every step, and this is exactly what I am doing here

Read everything as a text

Normalize_names will replace empty space with an underscore

All_varchar , will read everything as a text, you can cast the correct type later

DuckDB uses lazy evaluation by default, you need to use show() to see the result

Unpivot Other Columns

Unpivot other columns is one of the coolest feature of PoweQuery, DuckDB has a very nice SQL syntax to do the same

Write the final results as a Delta Table

That’s pretty much the core idea of fabric, if you write your Data into Delta Table then it will be usable by all the Engines, PowerBI, SQL DWH etc

Final Thoughts

The workflow did make sense to me, the only annoying thing is when you import a new notebook in a Fabric workspace it will not overwrite the existing one but create a copy, having the option will be nice.

I can see some advantages for this approach, like if you have a very small SKU and you want to optimize the resource usage, using your laptop for developpement can save you some precious compute, another case maybe if you were not given access to fabric from the admin, showing you have already a working code and you know what you are doing can make the conversation easier.

One thing for sure, users will use fabric in some unexpected ways and that’s just fine.

Fabric as a OS for analytics 

Introduction

Was listening to this short Video (Fabric Espresso: Will Fabric replace Azure Synapse?) and the VP of Synapse described Fabric as the OS for analytics, and I think it is not simply marketing talk but they are into something, This short blog will show that using DuckDB in Fabric maybe a useful scenario.

OneLake Throughput is the unsung hero of Fabric 

I tried to run DuckDB before in the cloud and all the systems I used Google Colab, Cloud functions, AWS Sagemaker etc have the same limitation, The throughput from the remote storage is just too slow, Fabric Onelake which is based on Azure ADLS Gen2 has an amazing throughput see this example

The file size is 26 GB , that’s nearly 433 MB/s, this is simply amazing, as a reference last week I bought one of those cheap USB flash drives and the read speed was 132 MB/s.

DuckDB reading Directly from Remote Storage

DuckDB is very fast when reading from a VM SSD especially when using the proprietary file format, but realistic speaking users in fabric would probably be more interested in reading directly from the Delta table, so I avoided creating another copy with DuckDB storage file.

I  test  two approaches

  • Import the data into Memory and than run the queries
  • Run the queries Directly from OneLake

The script is available here, the main Table is rather small 60 million rows, I am using just 4 cores, and the results are very interesting

32 second to import to Memory this includes decompressing and rewriting the data using DuckDB format (in RAM), but it is done only once.

24 seconds to run the Queries using 4 cores, just to give you an idea , another famous lakehouse vendor when using the same data, The Engine requires 16 cores and finishes the queries in around 40 seconds.

Running the queries directly from OneLake took 58 seconds, I notice though that Query 9 is particularly slow, which usually means a bad query Plan.

I run a profile on that Query and indeed when running directly from Parquet, DuckDB got the join order wrong, as DuckDB ignore the stats when reading from Parquet ( according to the dev, most stats in parquet are wrong anyway)

Note : I use Delta lake package to get the correct list of Parquet files to read, you can read directly from Delta using arrow dataset but it is slower.

Take away

In The medium term, we can imagine Fabric supporting More Engines. There is already some talks about SQL Server (although Parquet is not designed for OLTP, but that’s another subject)  The storage layer which is the foundation for any analytical work is very solid, the Throughput is very impressive, let’s just hope we get a lighter VM to run smaller workloads.

DuckDB currently does not support Buffer pool when reading from remote storage, I am afraid they need to add it. Microsoft supporting Delta has changed the market dynamic and DuckDB can’t ignore it. Parquet alone is not enough, we need those Delta stats for better Query plans.

Save Fabric Delta Tables as DuckDB file

The file section of Fabric Lakehouse is very interesting as although it is a blob storage, somehow behaves more or less like a real filesystem, we leverage that to save all Delta Tables in one DuckDB file

Limitation

  • DuckDB storage format is experimental at this stage and doesn’t offer backward compatibility yet, everytime they upgrade a major version, you have to export the file to parquet and import it back.
  • I am using Python Deltatlake package to read the Delta table, currently it supports only Delta reader version 1, which Microsoft uses, but this may change in a future update of Fabric.

How it works

Install DuckDB and Delta lake, and copy this function

The data will be saved for the current version of the table, old data not removed by vacuum will be ignored.

A note about concurrency

You can have multiple readers using the file at the same time (Using read_only=True), but multiple writers is  not a supported scenario nor 1 Writer and multiple reader, use at your own risk 🙂 Having said that, in the case of 1 writer and multiple reader the worst case scenario is reading inconsistent data 🙂

What is the Fastest Engine to sort small Data in a Fabric Notebook?

TL;DR : using Fabric Python Notebook to Sort and Save  Parquet files  up to 100 GB shows that DuckDB is very competitive compared to Spark even when using Only half the resources available in a compute pool.

Introduction : 

In Fabric the minimum Spark compute that you can provision is 2 nodes, 1 Driver and 1 Executor,  my understanding, and I am not an expert by any means is : Driver Plan the Work and Executor do the actual Works, but if you run any no Spark code, it will run in the driver, basically DuckDB use only the driver, the executor is just sitting there and you pay for it.

The experiment is basically : generate the Table Lineitem from the TPCH dataset as a folder of parquet files and sort it on a date field then save it. Pre-sorting the data on a field used for filtering is a very well known technique.

Create a Workspace

When doing POC, it is always better to start in a new workspace, at the end you can delete it and it will remove all the artifacts inside it. Use any name you want.

Create a Lakehouse

Click New then Lakehouse, choose any name

You will get an empty lakehouse (it is a just a storage bucket with two folders, Files and Table)

Load the Python Code

The Notebook is straightforward, Install DuckDB , create the data files if they don’t exist already, sort and save in a delta table using both DuckDB and Spark

Define Spark Pool Size

By default the notebook came with a starter pool that are warm and ready to be used, the startup is in my experience is always less than 10 second, but it is a managed service and I can’t control the number of nodes, instead we will use custom pool where you can choose the size of the compute and the number of nodes in our case 1 driver and 1 executor,  the startup is not bad at all, it is consistently less than 3 minute.

Schedule the Notebook

I don’t not know, how to pass a parameter to change the initial value in the pipeline, so I run it using a random number generator`, I am sure there is a better way, but anyway, it does works, and every insert the results

The Results

The Charts  show the resource usage by data size, CPU(s) =  Duration * Number of cores * 2.

Up to 300 Million rows, DuckDB is more efficient even when it is using only half the resources. 

To make it clearer , I build another chart that show the Engine combination with less resource utilization by Lintem size

From 360 Million rows, Spark became more economical ( with the caveat that DuckDB is just using half the resources) or maybe DuckDB is not using the whole 32 cores ?

Let’s filter only DuckDB

DuckDB using 64 cores is not very efficient for the size of this Data.

Partying Thoughts

  • Adding more resources to a problem does not make it necessarily an optimal solution, you get faster duration but it costs way more.
  • DuckDB Performance even using half the compute is very intriguing !!!
  •  Fabric Custom pools are a very fine solution, waiting around 2 minutes is worth it.
  • I am no Spark expert, but it will be handy to be able to configure at runtime a smaller Executor compute, in that case, DuckDB will be cheaper option for all sizes up to 100 GB and maybe more.