Loading both Cloud Data Lake and Warehouse

Let's map out how data can flow from the originating Operational Store to both the Data Lake and Data Warehouse. We have to decide if the Data Warehouse and Data Lake are peers or if one is the gold source and the other is a copy. 

Internet web and connected applications have created a data explosion.  Cheap storage and unlimited computing power are empowering new use cases like ML and revolutionizing old ones like CRM and CDP. Hive and Hadoop ushered in the age of big data. Data used to exist in two locations, operational and reporting databases. Now all data of all types can be collected into a single multi-petabyte Data Lake without expensive custom hardware.

Business requirements and regulatory needs should drive your design.
  1. The top diagram shows the originating systems loading the Data Lake and the Data Warehouse in parallel as peers. 
  2. The second diagram shows the Data Lake as the location for all data with some of that data replicated into the Warehouse.
  3. The third diagram is the inverse, showing the Warehouse as the primary data repository with the Data Lake as a copy/subset.
There are hybrid versions where the primary/copy paradigm varies based on the data type.  A columnar data warehouse might be the primary for binary data while the lake might be the primary for binary and unstructured data.

Video

Parallel Load


Source systems provide data in a stream or via drop location.  Lake and Warehouse-specific sinks pick up the data and store it on independent paths. The saved data is tracked in an external catalog and reconciled by an external process.

Advantages

  1. Data appears in each target store without having to route through an additional layer.
  2. The data stored in the raw zone is as close to the source system format as possible.  It is not mutated by transiting through an extra storage location.
  3. Binary and Unstructured data does not need special processing to land in the lake.
  4. Binary and Unstructured data does not have to transit through two systems.

Disadvantages

  1. Every transformation must be implemented twice, once for each channel.
  2. The specific data available in each repository overlaps with a Venn diagram due to the eventually consistent nature of the process. 
  3. Data loss in the two systems may be different.
  4. The data may be partitioned or ordered differently.

Lake First



All data is first pushed into the data lake.  Rectangular or relational friendly data is then passed on through to the data warehouse. 

Advantages

  1. Conformed and Curated zones are built in their own repository with repository-friendly tools.
  2. All data lands in the lake prior to being copied to another location.  This means the lake holds all of the data possible in other systems.
  3. Binary data can be peeled off early in the pipeline.
  4. Conversion from streaming or other formats into raw zone format only happens once.
  5. ML and other Big Data tools can see all the data, early.
  6. Data is safe at rest once it is written to the lake

Disadvantages

  1. The Data Lake is always the master copy and a superset of the Warehouse
  2. Data appears in the Warehouse after two delays, the lake load and the extract and warehouse load.

Warehouse First

All data is pushed into the Warehouse and then the Data Lake.  Binary data should probably be stored in the Lake as part of the Warehouse load. Raw data is then passed on through to the data warehouse. 

Advantages

  1. Conformed and Curated zones are built in their own repository with repository-friendly tools.
  2. All data lands in the warehouse prior to being copied to another location.  
  3. The SQL store is the analytical store where it can be used by humans.
  4. Conversion from streaming or other formats into raw zone format only happens once.
  5. Reporting and compliance see the data early.
  6. Data is safe as soon as it is written to the Warehouse

Disadvantages

  1. The content of the Data Lake is a disjoint overlap of the data in the Warehouse
  2. Data appears in the Lake after two delays, the warehouse load, and the extract and lakeload.
  3. Binary data should be peeled off early resulting in two different ways data appears in the Data Lake.

Wrap

These are possibilities. There are other patterns with their own strengths and weaknesses.  You may wish to pick a single pattern or a couple patterns that are then used based on the data type and consumer needs.

Created 2021 06 15

Comments

Post a Comment

Popular posts from this blog

Installing the RNDIS driver on Windows 11 to use USB Raspberry Pi as network attached

Image
I do a lot of my development and configuration via ssh into my Raspberry Pi Zero over the RNDIS connection. Some models of the Raspberry PIs can be configured with  gadget drivers  that let the Raspberry pi emulate different devices when plugged into computers via USB. My favorite gadget  is the network profile that makes a Raspberry Pi look like an RNDIS-attached network device.  All types of network services travel over an RNDIS device without knowing it is a USB hardware connection. A Raspberry Pi shows up as a Remote NDIS (RNDIS) device when you plug the Pi into a PC or Mac via a USB cable. The gadget  in the Windows Device Manager picture shows this RNDIS Gadget connectivity between a Windows machine and a Raspberry Pi. The Problem Windows 11 and Windows 10 no longer auto-installs the RNDIS driver that makes magic happen. Windows recognizes that the Raspberry Pi is some type of generic USB COM device. Manually running W indows Update  or  Upd...
Read more

Understanding your WSL2 RAM and swap - Changing the default 50%-25%

Image
The Windows Subsystem for Linux operates as a virtual machine that can dynamically grow the amount of RAM to a maximum set at startup time.  Microsoft sets a default maximum RAM available to 50% of the physical memory and a swap-space that is 1/4 of the maximum WSL RAM.  You can scale those numbers up or down to allocate more or less RAM to the Linux instance. The first drawing shows the default WSL memory and swap space sizing. The images below show a developer machine that is running a dev environment in WSL2 and Docker Desktop.  Docker Desktop has two of its own WSL modules that need to be accounted for.  You can see that the memory would actually be oversubscribed, 3 x 50% if every VM used its maximum memory.  The actual amount of memory used is significantly smaller allowing every piece to fit. Click to Enlarge The second drawing shows the memory allocation on my 64GB laptop. WSL Linux defaul...
Read more

Almost PaaS Document Parsing with Tika and AWS Elastic Beanstalk

Image
The Apache Tika project provides a  library capable of parsing and extracting data and meta data from over 1000 file types.  Tika is available as a single jar file that can be included inside applications or as a deployable jar file that runs Tika as a standalone service. This blog describes deploying the Tika jar as an auto-scale service in Amazon AWS Elastic Beanstalk.  I selected Elastic Beanstalk because it supports jar based deployments without any real Infrastructure configuration. Elastic Beanstalk auto-scale should take care of scaling up and down for for the number of requests you get. Tika parses documents and extracts their text completely in memory. Tika was deployed for this blog using EC2 t2.micro instances available in the AWS free tier. t2.micro VMs are 1GB which means that you are restricted in document complexity and size. You would size your instances appropriately for your largest documents.   Preconditions An AWS account. AWS ac...
Read more