MBrace Core Libraries & Runtime Foundations
This repository contains core libraries and runtime foundation components for the MBrace cloud programming model and frameworks.
For a first introduction to MBrace please refer to the main website at mbrace.io. If you have any questions regarding MBrace don't hesitate to create an issue or ask one of the maintainers. You can also follow the official MBrace twitter account @mbracethecloud.
MBrace.Core is a standalone class library that contains the core MBrace programming model, used to author general-purpose, runtime-agnostic distributed computation. It is centered on the concept of cloud workflows, a composable, language-integrated API based on F# computation expressions. It can be used to author specialized cloud libraries like MBrace.Flow.
MBrace.Core.Tests defines a comprehensive suite of abstracted NUnit/FsCheck based tests for use by MBrace runtime implementations. Useful for verifying that an MBrace implementation is up to spec regarding distribution semantics, serialization and fault tolerance. See MBrace.Thespian.Tests and MBrace.Azure.Tests for samples that make use of the test suites.
MBrace.Flow is a distributed streaming library written on top of MBrace.Core. It enables distributed computation using functional pipeline declarations like the following:
fsharp CloudFlow.OfHttpFileByLine "http://my.server.local/large.txt" |> CloudFlow.collect (fun line -> line.Split [|' '; ',' ; '.'|]) |> CloudFlow.filter (fun w -> w.Length > 3) |> CloudFlow.map (fun w -> w.ToLower()) |> CloudFlow.countBy id |> CloudFlow.sortBy (fun (_,c) -> -c) 10 |> CloudFlow.toArrayIt is written on top of the Nessos.Streams library, a fast streaming library inspired by Java 8 Streams.
MBrace.CSharp contains C# friendly wrapper APIs for the MBrace core programming model.
MBrace.Runtime provides an extensive set of foundations and common components for quickly developing MBrace runtimes on top of FsPickler/Vagabond. It removes the hassle of working with MBrace semantics and lets the runtime developer focus on providing cloud infrastructure implementations such as:
We've done work that effectively decouples the runtime implementation of MBrace from any particular cloud vendor. Details such as use of Vagabond, cloud workflow execution semantics, serialization and caching are handled by MBrace.Runtime, which can be used to target any infrastructure that supplies the following resources:
As a result, we have been able to quickly provide MBrace implementations for 3 different settings: standalone, Azure and AWS, all with relatively small codebases: 2900, 6800 and 6200 LoC respectively.
Conceivably it is now possible to provide MBrace bindings for any cloud infrastructure that provides aforementioned resources, private or public. We believe there's space for collaboration here.
MBrace.Thespian is a simple MBrace cluster implementation on top of MBrace.Runtime and the Thespian actor library. Not intended for production deployments, it is used for testing MBrace core development.
MBrace.Azure is an MBrace framework implementation on top of Azure PaaS components. Enables easy deployment of scalable MBrace clusters using worker roles. It also supports on-site cluster deployments using Azure storage/service bus components for communication.
MBrace.AWS is an MBrace framework implementation on top of AWS PaaS components. Enables easy deployment of scalable MBrace clusters using Elastic Beanstalk & Docker. It also supports on-site cluster deployments using AWS S3/DynamoDB/SQS components for communication.
Building MBrace Core requires Visual Studio 2015/F# 4.0. You can build the project either from Visual Studio or by running
build.cmd Buildif using cmd/powershell or
./build.sh Buildif using bash/sh.
Unit tests can be run by calling
./build.sh RunTests. You can skip the time consuming Thespian by adding the
-ef IgnoreClusterTestsflag in the above commands. Alternatively you can run individual tests by opening test assemblies found in the repository's
/binfolder using NUnit-GUI. Building the solution produces the following test assemblies: *
MBrace.Core.Tests.dllfor testing the core MBrace library. *
MBrace.Runtime.Tests.dllfor testing MBrace.Runtime components. *
MBrace.Thespian.Tests.dllfor testing the MBrace Thespian implementation.
When performing changes to MBrace.Core, MBrace.Flow or MBrace.Runtime it is essential to verify that the MBrace.Thespian tests are still passing.
The MBrace project is happy to accept quality contributions from the .NET community. If you would like to get involved, here are a few places you could have a look at:
MBrace Libraries: we are looking for data scientists and domain experts who can help us develop specialized libraries that run on top of MBrace.Core. Examples include Machine Learning and Graph analytics libraries.
MBrace Runtimes: help MBrace grow by extending support to your favorite cloud service. It could be AWS, YARN/HDFS or your private infrastructure.
C# Support: As of MBrace 1.0, development of MBrace.CSharp has been suspended. We are looking for working C# developers interested in extending MBrace support to C# and testing deployments using the upcoming C# Interactive.
Documentation & Code Samples: help improve MBrace documentation and coding samples. See the MBrace.StarterKit for current coding samples and the mbrace-docs repo for documentation and the mbrace.io website.
This project is subject to the Apache Licence, Version 2.0.