New Delhi, Apr 17, (CXO Media): MakeMyTrip needed sub-second latency to power its “last-searched” hotels feature for millions of daily users.
Users tap the search bar and expect a real-time, personalized list of their recent interests, based on their interaction with the system.
At the scale required, even a few hundred milliseconds of delay creates friction in the user journey, directly impacting click-through rates.
Subsequently, MakeMyTrip became the first customer to adopt Databricks’ Apache Structured Streaming introduces RTM.
RTM allowed MakeMyTrip to achieve millisecond level latency primarily through three key technical innovations.
These innovations include continous data flow, pipeline scheduling and streaming shuffle.
MakeMyTrip and Databricks collaborated to build, tune and validate stream union merging B2C and B2B into a single platform.
The collaboration also validated task multiplexing with more partitions and fewer cores. Last but not the least, it validated pipeline hardening including checkpointing, backpressure and fault tolerance.
“As we move into the era of AI agents, steering them effectively requires building real-time context from data streams. We are experimenting with Spark RTM to supply our agents with the richest, most recent context necessary to take the best possible decisions,” claimed Aditya Kumar, Associate Director of Engineering, MakeMyTrip.
