Real-time Data Infrastructure at Uber

Abstract

Uber's business is highly real-time in nature. PBs of data is continuously being collected from the end users such as Uber drivers, riders, restaurants, eaters and so on everyday. There is a lot of valuable information to be processed and many decisions must be made in seconds for a variety of use cases such as customer incentives, fraud detection, machine learning model prediction. In addition, there is an increasing need to expose this ability to different user categories, including engineers, data scientists, executives and operations personnel which adds to the complexity.In this paper, we present the overall architecture of the real-time data infrastructure and identify three scaling challenges that we need to continuously address for each component in the architecture. At Uber, we heavily rely on open source technologies for the key areas of the infrastructure. On top of those open-source software, we add significant improvements and customizations to make the open-source solutions fit in Uber's environment and bridge the gaps to meet Uber's unique scale and requirements.We then highlight several important use cases and show their real-time solutions and tradeoffs. Finally, we reflect on the lessons we learned as we built, operated and scaled these systems. CCS CONCEPTS• Information systems → Stream management; • Computer systems organization → Real-time system architecture.

Recent Advances in Data Engineering for Networking

“…Stream processing engines/platforms (such as Spark Streaming, Apache Flink) enable strictly ordered and exclusive message passing while allowing computational logic to be applied to message streams [52]. In a streaming-based communication pattern, the ordering of events is important so that behaviour can be analyzed based on the sequence of ordered events.…”

“…As a data application framework, Streamlit 51 can be used to easily create ML and data science web applications. Metabase 52 is another open source tool for business intelligence purposes.…”

“…based on the configured alert rules. 47 https://grafana.com/, accessed December-2021 48 https://gephi.org/, accessed December-2021 49 https://dash.plotly.com/, accessed December-2021 50 https://superset.incubator.apache.org/, accessed December-2021 51 https://www.streamlit.io/, accessed December-2021 52 https://github.com/metabase/metabase, accessed December-2021 53 https://www.datadoghq.com/, accessed December-2021 54 https://sentry.io/welcome/, accessed December-2021…”

Assimilating sense into disaster recovery databases and judgement framing proceedings for the fastest recovery

“…By adding more intelligence and automation, the standby begins making decisions based on the rate of concurrency of primary transactions and the out-of-sync status at standby. The relational database management systems (RDBMS) technologies have evolved in recent years to offer the maximum level of support in DR environments [12]- [14]. As a result, the retrieval process and application of the changes on the DR site have reached a new level.…”

Decentralized Data Flows for the Functional Scalability of Service-Oriented IoT Systems

“…14(b-c) only if input data is needed; otherwise, steps 1-2 are omitted. When control triggers an 6 A pipeline of different technologies can be used to efficiently implement data exchanges between services [51].…”