Reactive Actor framework for flexible/distributed/scalable and monitorable applications

Introduction

ReActed is an actor oriented framework for creating reactive applications. It naturally embraces the microcomponents paradigm in its design, allowing the realization of distributed, flexible, reliable, monitorable and automatically scalable applications out of the box. ReActed amplifies the productivity of developers providing a ready to use solution that can be easily integrated with almost any other technology and fine-grained tuned only if required.

ReActed can manage for you:

• Parallelization: leveraging the actor paradigm, reactor can manage for you parallelization, saving the programmer the effort to write and maintain concurrent code. Synchronization or memory visibility issues problems are structurally removed from your code

• Distribution: a ReActed instance can naturally and automatically create a cluster with other instances. Resources can be shared, used, added or created through a location agnostic approach, creating an automatically scalable and location agnostic way of designing a system

• Monitoring: creating an event source or a choreography oriented microservices system can be tricky, but monitoring and debugging it can be even more difficult. ReActed can do this for you through its Re-playable Actors. Run the application and if something goes wrong you can just analyze the log or replay the logs to reset the system to its state in any given instant in time.

• Flexibility: in a real case scenario it's hard being able to choose from scratch all the technologies of your stack or even being able to use a technology for all the parts of your infrastructure. Thanks to ReActed driver system, different services or different part of the infrastructure can talk with each other using different technologies. Do you need the speed gRPC? Done. Somewhere else do you need Kafka? Done. Do you need both at the same time in different part of the system? Done as well and thanks to the location agnostic paradigm, you do not need to care about which technology should be used to communicate with a given service, ReActed can do that for you.

ReActed provides you a flexible and dependable environment that allows the developer to focus on producing business logic.

Getting Started

If you want to jump into the code, you can give a look to the examples directory, otherwise give a look to the quickstart guide in the official documentation

Benchmarks

Test: Reaction to a message latency time. Sender and receiver are executed in two different threads

CPU: Apple M1 Pro

Test scenario 1

• Message Number: 10_000_000
• Message Send Rate: 1 every 10us
• Message business payload: 8 bytes

Percentile	Direct Communication Local Driver Latency	Chronicle Queue Local Driver Latency
70	417ns	2.625us
75	417ns	2.875us
80	458ns	3us
85	458ns	3.167us
90	541ns	3.5us
95	1.458us	3.792us
99	1.917us	6.084us
99.9	8us	45.125us
99.99	32us	8.099416ms
99.9999	1.0580425ms	9.511375ms
100	1.6896254ms	9.711917ms

---

Test scenario 2

• Message Number: 10_000_000
• Message Send Rate: 1 every 5us
• Message business payload: 8 bytes

Percentile	Unbounded Mailbox Latency	Fast Unbounded Mailbox Latency
70	417ns	417ns
75	417ns	458ns
80	500ns	458ns
85	1.416us	459ns
90	1.541us	1.333us
95	1.625us	1.542us
99	2us	2.458us
99.9	11.750us	6.167us
99.99	442.958us	219.875us
99.9999	3.826041ms	1.681542ms
100	3.874833ms	1.72425ms

---

Test scenario 3

• Message Number: 10_000_000
• Message Send Rate: 1 every 1us
• Message business payload: 8 bytes

Percentile	Unbounded Mailbox Latency	Fast Unbounded Mailbox Latency
70	458ns	791ns
75	458ns	875ns
80	458ns	1us
85	459ns	1.416us
90	625ns	1.541us
95	1.292us	1.584us
99	2.333us	1.958us
99.9	34.041us	34.084us
99.99	2.966584ms	1.394875ms
99.9999	3.566709ms	1.90175ms
100	4.36955ms	1.909792ms

---