Press "Enter" to skip to content

Category: Architecture

Architecting a Data Lake

James Serra provides some guidance:

I have had a lot of conversations with customers to help them understand how to design a data lake. I touched on this in my blog Data lake details, but that was written a long time ago so I wanted to update it. I often find customers do not spend enough time in designing a data lake and many times have to go back and redo their design and data lake build-out because they did not think through all their use cases for data. So make sure you think through all the sources of data you will use now and in the future, understanding the size, type, and speed of the data. Then absorb all the information you can find on data lake architecture and choose the appropriate design for your situation.

The concepts are simple but there are some interesting implications to what James includes as well as additional resources, so check it out.

Leave a Comment

Event-Driven Microservices in Python with Kafka

Dave Klein demonstrates how event-driven microservices work:

Along came microservices. Individual, smaller applications that could be changed, deployed, and scaled independently. After some initial skepticism, this architectural style took off. It truly did solve several significant problems. However, as is often the case, it brought new levels of complexity for us to deal with. We now had distributed systems that needed to communicate and depend on each other to accomplish the tasks at hand.

The most common approach to getting our applications talking to each other was to use what we were already using between our clients and servers: HTTP-based request/response communications, perhaps using REST or gRPC. This works, but it increases the coupling between our independent applications by requiring them to know about APIs, endpoints, request parameters, etc., making them less independent.

Read the whole thing.

Leave a Comment

Tips for using Synapse Database Templates

James Serra provides some guidance:

I had previously blogged about Azure Synapse Analytics database templates, and wanted to follow-up with some notes and tips on that feature as I have been involved on a project that is using it:

– Purview does not yet pull in the metadata for database templates (table/field descriptions and table relationships). Right now it pulls in the metadata as if it was a SQL table or as if it was a file in ADLS. Both just have the basic information supported by those types. The SQL one is probably preferred

– Power BI does not import the table and field descriptions when connecting to a lake database (where the database templates are stored), but it does import the table relationships. You can see the table descriptions by hovering over the table names in the navigator when importing tables using the “Azure Synapse Analytics workspace (Beta)” connector. Note you are not able to see the table descriptions when hovering over the table names using the “Azure Synapse Analytics SQL” connector. Also note the “Select Related Tables” button does not work in the navigator

Click through for more notes from the field.

Leave a Comment

Building a Lakehouse with Azure Synapse Analytics

Arshad Ali does a bit of construction:

Data Lakehouse architecture has become the de facto standard for designing and building data platforms for analytics as it bridges the gap and breaks the silos created by the traditional/modern data warehouse and the data lake. This blog post introduces you to the world of data lakehouse and it goes into details of how to implement it successfully in Azure with Azure Synapse Analytics.

Read the whole thing.

Comments closed

Inverted Indexes for Full-Text Search

Maria Zakourdaev twists some text inside-out:

Sometimes there are properties in the document with unstructured text, like newspaper articles, blog posts, or book abstracts. The inverted index is easy to build and is similar to data structures search engines use.

Such document structures can help in various complex search patterns, like common word detection, full-text searches, or document similarity searches, using humming distance or l2distance algorithms. Inverted indexes are useful when the number of keywords is not too large and when the existing data is either totally immutable or rarely changed, but frequently searched.

This post and Maria’s MSSQLTips post both cover the high-level concept, focusing on tradeoffs between different data models. I like this sort of idea a lot and like telling people that sometimes, the right answer in a relational database involves thinking backwards.

Comments closed

Understanding Write-Ahead Logging

Kevin Sookocheff explains how write-ahead logging protects data in databases:

A central tenet of databases is that any committed data survives a crash or a failure. Write-ahead logging is a fundamental primitive that ensures all changes to data are first written safely to stable storage before being applied. Coupling that with some careful use of sequence numbers and we can guarantee that changes made to a database can survive system crashes.

This is a core feature in pretty much every relational database and Kevin dives into how one of the key algorithms behind it works.

Comments closed

Optimizing Azure Pricing for Storage and VMs

Shane Baldacchino continues a series on cost optimization in the cloud:

Cost. I have been fortunate to work for and help migrate one of Australia’s leading websites ( in to the cloud and have worked for both large public cloud vendors. I have seen the really good, and the not so good when it comes to architecture.

Cloud and cost. It can be quite a polarising topic. Do it right, and you can run super lean, drive down the cost to serve and ride the cloud innovation train. But inversely do it wrong, treat public cloud like a datacentre then your costs could be significantly larger than on-premises.

Click through for some good advice, including an appreciation of spot instances.

Comments closed

Building a Data Serving API in Azure

Justice Zishanhi has some recommendations for serving data in Azure:

Data is an important asset to all organizations big and small. As these organizations mature, building an end-to-end data platform to enable BI and AI at scale has become part of that journey. Some organizations, have the requirement to expose modelled data in a data warehouse or data lake (Azure Data Lake Storage Gen2) to downstream consumer applications (mobile or web apps) where access patterns can be unpredictable in respect to frequency of access and/or type of data that is requested.

Data warehouse engines and data lakes are not designed for singleton transactional (request / response) interactions.  To serve these requests at scale and to meet the different SLAs and access pattern unpredictability, data needs to be offloaded to a suitable database engine (i.e., a caching layer) that is built to serve such queries.  

The “Design Patterns” section of this article highlights a generalized pattern for implementing a data serving API which meets this requirement – consisting of a Data Platform component and an API component. For implementing the API, two patterns are commonly adopted – a synchronous pattern or an asynchronous pattern. Both are explored in the “API Implementation Patterns” section of this article.

The example focuses on Cosmos DB and provides quite a bit of helpful guidance.

Comments closed

Multi-Developer Power BI Development

Reza Rad architects a solution for multiple developers working on a Power BI project:

Before I start explaining the architecture, it is important to understand the challenge and think about how to solve it. The default usage of Power BI involves getting data imported into the Power BI data model and then visualizing it. Although there are other modes and other connection types, however, the import data is the most popular option. However, there are some challenges in a model and a PBIX file with everything in one file. Here are some;

– Multiple developers cannot work on one PBIX file at the same time. Multi-Developer issue.

– Integrating the single PBIX file with another application or dataset would be very hard. High Maintenance issue.

– All data transformations are happening inside the model, and the refresh time would be slower.

– The only way to expand visualization would be by adding pages to the model, and you will end up with hundreds of pages after some time.

– Every change, even a small change in the visualization, means deploying the entire model.

– Creating a separate Power BI file with some parts it referencing from this model would not be possible; as a result, you would need to make a lot of duplicates and high maintenance issues again.

– If you want to re-use some of the tables and calculations of this file in other files in the future, it won’t be easy to maintain when everything is in one file.

– And many other issues.

After laying out all of the challenges, Reza puts together a plan to resolve them.

Comments closed

Data Mesh at Netflix

Bo Lei, et al, describe their Data Mesh architecture:

Realtime processing technologies (A.K.A stream processing) is one of the key factors that enable Netflix to maintain its leading position in the competition of entertaining our users. Our previous generation of streaming pipeline solution Keystone has a proven track record of serving multiple of our key business needs. However, as we expand our offerings and try out new ideas, there’s a growing need to unlock other emerging use cases that were not yet covered by Keystone. After evaluating the options, the team has decided to create Data Mesh as our next generation data pipeline solution.

Click through for a high-level overview of the architecture.

Comments closed