DB System

Distributed ML and Blockchain Open Source Systems

October 21, 2019

Original posted on DB World

1. Apache SINGA: a distributed machine/deep learning platform

apache_singa

On 16 Oct 2019, Apache SINGA, a distributed machine/deep learning engine mainly driven by database researchers, has graduated from Apache Incubator as a Top Level Project (TLP) of Apache Software Foundation. For the current release, please check:

https://github.com/apache/incubator-singa

https://en.wikipedia.org/wiki/Apache_SINGA

Current release:

  • SINGA-auto (SINGA v2.0, a.k.a. RAFIKI) provides AutoML services such as model construction, hyper-parameter tuning, model training and inference. Users can simply upload their datasets and configure the service to conduct training and then deploy the model for inference. As a cloud service system, SINGA manages the hardware resources and failure recovery. It speeds up hyper-parameter tuning by distributed tuning to achieve almost linear scalability.

Upcoming Releases:

  • SINGA-easy (a.k.a. PANDA) allows people without much ML knowledge to use the technology with ease. SINGA-easy empowers a domain expert such as a doctor/clinician to fuse his domain knowledge into ML systems.
  • SINGA-lite is being developed for edge computing and 5G environment
  • SINGA-db is being developed as a self-driving data system for complex analytics.

Contributions to the project and community building are most welcome and appreciated.

2. FabricSharp: a permissioned blockchain system

FabricSharpLogo

FabricSharp system is a variant of Hyperledger Fabric 1.4, a permissioned blockchain platform from Hyperledger. Compared with the vanilla version, FabricSharp supports fine-grained secure data provenance, sharding, use of trusted hardware, and a blockchain native storage engine called ForkBase, to boost system performance. Please check:

https://github.com/ooibc88/FabricSharp

Contributions to the project and community building are most welcome and appreciated.