AWS Big Data Blog
How Getir unleashed data democratization using a data mesh architecture with Amazon Redshift
This blog post is co-written with Pinar Yasar from Getir.
Amazon Redshift is a fully managed cloud data warehouse that’s used by tens of thousands of customers for price-performance, scale, and advanced data analytics. Amazon Redshift enables data warehousing by seamlessly integrating with other data stores and services in the modern data organization through features such as Zero-ETL, data sharing, streaming ingestion, data lake integration, and Redshift ML.
In this post, we explain how ultrafast delivery pioneer, Getir, unleashed the power of data democratization on a large scale through their data mesh architecture using Amazon Redshift.
We start by introducing Getir and their vision—to seamlessly, securely, and efficiently share business data across different teams within the organization for BI, extract, transform, and load (ETL), and other use cases. We’ll then explore how Amazon Redshift data sharing powered the data mesh architecture that allowed Getir to achieve this transformative vision. We will also explain how Getir’s data mesh architecture enabled data democratization, shorter time-to-market, and cost-efficiencies. Next, we’ll provide a broader overview of modern data trends reinforced by Getir’s vision. In conclusion, we’ll offer some thoughts on how you can apply a similar approach to eliminate costly and barrier-inducing data silos using Amazon Redshift.
Who is Getir?
Getir is an ultrafast delivery pioneer that revolutionized last-mile delivery in 2015 with its 10-minute grocery delivery proposition.Getir’s story started in Istanbul, and they have launched multiple products since inception: GetirFood, GetirMore, GetirWater, GetirLocals, GetirBitaksi (taxi service), GetirDrive (car rental service), and GetirJobs (recruitment).
Getir serves dozens of cities throughout the world with more than 30,000 employees. The following figure shows the Getir app.
Overview of Getir’s main use case
Getir’s business is characterized by a tremendous volume of data generation and growth, in addition to ample opportunities to gain valuable insights. However, siloing this data and creating friction for teams trying to access the information they needed wasn’t a viable option. Allowing teams to duplicate data wherever required can be an anti-pattern, leading to operational complexity, cost overruns, and fragile data storage bloat.
Similarly, relying on dedicated teams to create data extracts or insights for downstream consumers introduces bottlenecks, stifles innovation, and increases the time-to-market. This approach isn’t optimal for a data-driven organization like Getir, which needs to empower its teams with seamless access to the information they require to drive the business forward. The various business lines within the organization made it abundantly clear that they wanted unfettered access to the company’s entire data ecosystem in a secure, cost-efficient, near real-time, and well-governed manner.
Furthermore, the organization was anticipating the emergence of data-as-a-service and generative AI use cases in the near future. This would necessitate the ability to securely share and potentially monetize the company’s data with external partners, such as franchises.
Overview of Getir’s use of Amazon Redshift and modern data architecture
To strike a balance that addresses these concerns and enables Getir teams to effectively use the wealth of data to generate meaningful insights and drive strategic decision-making across the organization, we chose a data mesh architecture.
Getir’s data analytics environment encompasses hundreds of terabytes of data, thousands of tables, and billions upon billions of data rows. Additionally, it processes millions of messaging events daily, all of which must be ingested, refined, and made available to analysts querying multiple Amazon Redshift warehouses. The end-to-end service level agreements (SLAs) for this data ecosystem can be extremely aggressive, with requirements that can be as stringent as single-digit minutes to single-digit seconds. This underscores the scale and complexity of Getir’s data analytics capabilities, which must operate with the utmost efficiency and responsiveness to meet the demands of the business. We were able to easily implement the envisioned data mesh architecture using Amazon Redshift’s native data sharing capabilities.
As the preceding diagram shows, at the heart of Getir’s architecture, was an ETL Redshift data warehouse that was used for various data sets from all over the organization, creating a refined 360-degree view of critical assets. It also was a producer for downstream Redshift data warehouses.
The demand was quite heavy on this main ETL cluster, so we relied on data sharing to isolate noisy workloads on a different Redshift data warehouse without having to duplicate the data on the main ETL cluster.
Using Redshift data sharing, individual business line teams could now rely solely on their dedicated Redshift cluster to provide them with their own data and analytics capabilities, but also the refined 360-degree views of data generated from all over the organization—without any data duplication or overstepping compute boundaries. BI analysts gained access to all of the data they needed to power their most complex dashboards with consistent performance free of noisy jobs. Additional warehouses were integrated into the data mesh for visualization, reporting, and machine learning.
Another benefit of Amazon Redshift data sharing and the data mesh architecture, was the relative ease with which we were able to maintain a chargeback model for ensuring costs were spread fairly across different teams.
Finally, the data sharing capability also enabled the seamless propagation of newly created tables within a schema to the subscribed consumers.
Modern data trends reinforced by Getir’s case study
Getir’s case study showcases the strategic uses of a data mesh architecture and Amazon Redshift, but more importantly provides tremendous insights into five key trends across all industries as modern data organizations move away from costly data silos that hinder collaboration, business insights, and time-to-market. As highlighted in the following diagram, those trends are 1/interconnected, purpose-built data stores that enable users to access data regardless of its physical location, 2/data democratization empowering users with self-service analytics capabilities, 3/real-time insights to drive greater value from data, 4/resilient data services ensuring business continuity, 5/leveraging generative AI to extract even deeper insights from data more expeditiously.
As Getir showed, the modern data organization is adopting data architectures that democratize data securely and enable self-service analytics. To realize data’s true potential, the modern data organization has progressed beyond basic dashboarding and reporting on limited, point-in-time data sets, and evolved to use more sophisticated ETL processes that can ingest data from diverse sources. Near real-time analytics in addition to predictive models have become standard fare, significantly reducing the time to actionable insights.
Furthermore, the data landscape has been democratized to empower analysts in numerous ways through the rise of transactional data lakes powered by open table formats such as Apache Iceberg and the assistance of generative AI. This holistic approach has elevated data organizations’ capabilities well beyond traditional reporting, unlocking greater business value from the wealth of data available.
Using generative AI with data mesh architecture
In addition to the five key trends previously mentioned, the present-day data landscape is characterized by three key facts that are leading data organizations like Getir to increasingly harness the power of generative AI to drive the next evolution of data-informed decision-making.
Data is an organization’s most valuable asset and the ability to effectively use data is central to an organization’s success and growth. Data analytics and insights are absolutely crucial to strengthening and expanding the business. Deriving meaningful insights from data is essential for making informed, strategic decisions. Democratizing data and enabling self-service analytics can greatly expand the range of business insights, while reducing the time to market for those insights. Empowering users across the organization to access and analyze data can unlock tremendous value. Generative AI’s ability to respond to natural language prompts, explore and analyze complex data, and summarize lengthy content makes it a valuable tool for translating large amounts of data into valuable insights. However, the true potential of generative AI for organizations lies in Retrieval Augmented Generation (RAG).
Out of the box, generative AI models start with a relatively generic knowledge base, which can lead to unreliable or inaccurate information. RAG addresses this by introducing the model to additional datasets that are specific to the organization or context. This allows generative AI models to produce far more accurate, attributable, and highly contextualized outputs to support decision-making.
Data mesh architecture can play a crucial role in enabling and facilitating RAG. By facilitating access to multiple data sources within the organization, the data mesh provides the necessary fuel for the generative AI model to draw from, resulting in more reliable and insightful information. This, in turn, empowers data-driven decision-making and helps organizations harness the full potential of their data assets.
Conclusion
In this post, we examined how Getir implemented a data mesh architecture and Amazon Redshift data sharing to meet their evolving data requirements. This entailed dedicated data warehouses tailored to different business lines and needs, while maintaining robust data governance and secure data access. Additionally, we highlighted the key industry trends that Getir’s case study reinforces across the broader data landscape. For more information, contact AWS or connect with your AWS Technical Account Manager or Solutions Architect, who will be happy to provide more detailed guidance and support.
About the Authors
Asser Moustafa is a Principal Worldwide Specialist Solutions Architect at AWS, based in Dallas, Texas, USA. He partners with customers worldwide, advising them on all aspects of their data architectures, migrations, and strategic data visions to help organizations adopt cloud-based solutions, maximize the value of their data assets, modernize legacy infrastructures, and implement cutting-edge capabilities like machine learning and advanced analytics. Prior to joining AWS, Asser held various data and analytics leadership roles, completing an MBA from New York University and an MS in Computer Science from Columbia University in New York. He is passionate about empowering organizations to become truly data-driven and unlock the transformative potential of their data.
Pinar Yasar is the Data Engineering Manager at Getir. Her passion is to accelerate self-service analytics for her internal customers and build highly scalable and cost-effective solutions in the cloud.