# Angel-ML/angel

A Flexible and Powerful Parameter Server for large-scale machine learning

Repository: https://github.com/Angel-ML/angel
Canonical: https://ross.abutalabs.com/products/angel
Language: Java
License: NOASSERTION
License Family: other
Topics: machine-learning, parameter-server, spark, scala, model, high-dimensional, online-learning, spark-streaming
Last push: 2026-07-26T06:10:31+00:00

## Health v2 (maintenance only)
Score: 69/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 94, release rhythm 18, longevity 100
- inputs: {"age_days": 3417, "days_push": 38, "days_rel": 338, "gap_med": null, "n_releases_24m": 1}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 6787, forks 1589 (observed 2026-08-28T04:09:48.604606+00:00)

## What it is
Angel is a high-performance distributed parameter server for large-scale machine learning and graph computing, developed by Tencent and Peking University. It partitions model parameters across parameter-server nodes and integrates with Spark via Spark on Angel, running on Yarn clusters.

## Use cases
- train large-scale machine learning models on a Spark cluster
- run distributed parameter server for high-dimensional models
- perform online learning with streaming data on Spark
- run graph computing algorithms at scale
- accelerate Spark ML workloads with PS Service
- train models on Yarn with fault-tolerant parameter synchronization

## When to choose
- you need a parameter server for very high-dimensional models on Hadoop/Yarn
- you want to boost Spark MLlib with distributed model storage and updates
- you run large-scale graph algorithms alongside machine learning pipelines
- you need a battle-tested system tuned for industrial big data workloads

## When to avoid
- you need deep learning training on GPUs rather than CPU-based parameter server workloads
- your project uses Python-native ecosystems like PyTorch or TensorFlow distributed training
- you want a lightweight single-node ML library
- you cannot deploy on Yarn or a JVM-based cluster environment

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, llm-training, streaming
- domain: machine-learning, big-data, microservices, graph-processing
- platform: jvm
- tags: parameter-server, spark-on-angel, yarn, high-dimensional-models, online-learning, graph-computing, tencent, linux, docker

## Member repositories
- Angel-ML/angel (main) score 69

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:48.604606+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:42:20.822957+00:00, confidence not recorded.
  - readme: https://github.com/Angel-ML/angel (fetched 2026-08-28T04:09:48.604606+00:00, sha ab927febc184)
- Data as of 2026-08-30T08:39:29.467469+00:00.