Alibaba Cloud Retail China Translated from Chinese

CDF Sunrise (中免日上)

Marketing Personalization

CDF Sunrise relied on a self-built open-source Milvus cluster to power recommendations across its app and mini-program, but recommendation traffic and vector volumes were growing faster than the team could manage. The self-hosted cluster became harder to keep stable, harder to scale, and more expensive to operate as the ecommerce platform expanded.

monitoring covers 100+ indicators including CPU, memory, QPS, and RT
performance improved more than 10% versus the open-source deployment

Reported impact

Business outcome

CDF Sunrise gained a more stable recommendation backbone and reduced the engineering burden of operating vector infrastructure. The migration improved performance, gave the team deeper observability into recommendation traffic, and created a smoother path to scale as new products and recommendation demand continue to grow.

  • Customer Experience
  • Productivity Improvement
  • Scalability

Solution

CDF Sunrise migrated the vector-search core of its recommendation system from a self-built Milvus cluster to Alibaba Cloud Vector Search Service Milvus Edition. The managed service preserved compatibility with open-source Milvus so the team could move existing recommendation workloads without redesigning the application logic, but it added the operational tooling the self-hosted cluster lacked. Recommendation vectors for app and mini-program homepage modules now run on a platform with built-in observability, more flexible vertical and horizontal scaling, and managed version operations for testing or upgrading cluster behavior. Alibaba Cloud also exposes more than one hundred monitoring signals such as CPU, memory, QPS, and response time, which lets the team detect issues earlier and tune resources before customer experience suffers. The result is a more reliable similarity-search layer for ecommerce recommendations and a lower-ops path for continued growth.

Data flow

Product and recommendation vectors from the CDF Sunrise commerce platform are stored in managed Milvus, where similarity searches power homepage recommendation modules in the app and mini-program. Monitoring, alerting, and scaling signals from the managed service feed operations workflows that keep the recommendation engine healthy.

Solution architecture

1 components · 1 layer
  1. Storage
    • Vector Search Service Milvus Edition Runs the vector-similarity engine for recommendation scenarios across the app and mini-program.

Architecture clues

  • Managed version upgrades are part of the experimentation flow for future recommendation tuning.
  • Recommendation workloads span both the main app and a mini-program, so scaling and observability must cover multiple channels.
  • The platform uses managed Milvus as the vector-search core while preserving compatibility with the open-source API surface.

Evidence from the source

After testing, the team found Alibaba Cloud Milvus delivered more than 10% better performance than the open-source version and decided to migrate all recommendation clusters. (original: 与开源版本相比,阿里云 Milvus 在多项性能指标上提升超过 10%... 决定,将 App 及小程序内推荐系统涉及的自建 Milvus 集群,全部搬迁至阿里云向量检索服务 Milvus 版上。)
Alibaba Cloud also exposed more than 100 monitoring metrics and flexible vertical or horizontal scaling, improving stability and cluster control. (original: 涉及超过 100 项监控指标... 且支持自定义报警规则... 提供更灵活的资源调整方案,包含垂直升降配与横向扩缩容。)
CDF Sunrise said app and mini-program recommendation modules had become a major traffic entrance and generated massive vector data, which exposed the limits of its self-built Milvus cluster. (original: App 及小程序首页推荐栏目成为重要的流量入口... 因此产生了海量的向量数据... 随着业务规模快速增长,自建 Milvus 的限制也逐渐出现。)
Transparency Evidence & methodology 90/100 reference value · 88% confidence · 1 source

Review how this reference was scored and trace each normalized claim back to its documented source.

01

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value90/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth25/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity20/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact15/25
    1 metric describe measurable impact.
  • Deployment scale10/10
    Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence88%
  • Source authority20/25
    The main evidence is an official vendor-published customer story.
  • Claim traceability25/25
    12 of 12 key fields link to a named source.
  • Direct evidence15/20
    3 direct source quotes retained.
  • Date provenance10/10
    The publication date comes from the source page or publisher.
  • Source transparency8/10
    1 source reference document publisher, URL, authority, retrieval, and supported fields.
  • Implementation specificity10/10
    Rewards source-backed components, data flow, quantified impact, and a specific outcome.
02

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

  1. Alibaba Cloud

    阿里云向量检索服务 Milvus 版助力中免日上搭建在线推荐系统

    First-party
    https://www.aliyun.com/customer-stories/retail-2025-cdfsunrise
    Source type
    Vendor-published customer story
    Supports
    • Customer
    • Region
    • Company size
    • Business problem
    • Solution
    • Products
    • Solution components
    • Data flow
    • Integrations
    • technical area
    • use case category
    • Business outcome
    • outcome category
    • Metrics
    • Architecture
    • evidence quotes

Date evidence

  • Published on source page Feb 25, 2025 · Alibaba Cloud · First-party
    Open date source
Not separately sourced Industry. Treat these normalized attributes as lower-confidence until a dedicated source is added.