Practices of Building an Elasticsearch Service Platform on Kubernetes at ByteDance

On November 16, the Elastic Chinese Community Shenzhen Meetup featured a talk by senior engineer Huang Yangfeng from ByteDance on the practice of an Elasticsearch (ES) service platform built on Kubernetes.

Huang currently leads ES service platform development at ByteDance Shenzhen and has previously worked at Huawei, YY, and Tencent, where he used ES for WeChat Pay data retrieval.

Q1: What pain points were encountered when building the ES service platform from 0 to 1, and how were they solved?

A: (1) Isolation from large clusters was required; this was addressed by deploying the platform on Kubernetes.

(2) Performance analysis capabilities were relatively weak; the plan is to add metric and trace logs to gain detailed latency insights.

Q2: What future plans are there for the ES service platform?

A: The speaker referred to a roadmap diagram (shown in the original presentation) outlining upcoming features.

Q3: What are your views on the current ES ecosystem, its applications, and its future?

A: ES is evolving rapidly with an increasingly rich surrounding ecosystem. The focus should be on optimizing query performance and improving cluster stability as the technology scales.

Q4: What expectations do you have for this technical salon’s theme sharing?

A: Anticipating updates on performance optimization and the latest advances in machine learning related to ES.

Q5: Do you have any suggestions for the development of the Elastic Chinese Community?

A: (1) Increase ES’s influence among Chinese developers to a level comparable to Google/Amazon. (2) Publish more core‑related books or articles. (3) Reduce pricing.

The meetup also announced that registration was open for the event, with the theme "Practices of Building an Elasticsearch Service Platform on Kubernetes at ByteDance" and an abstract stating that the platform has been adopted by products such as Toutiao and Douyin, covering challenges in Kubernetes, ES feature enhancements, cross‑region disaster recovery, full‑stack monitoring and alerts, and automated deployment.