evadne / ets-playground

Leveraging ETS Effectively

Leveraging ETS Effectively

This repository contains sample code for the talk Leveraging ETS Effectively, presented at ElixirConf EU 2019 on 9 April 2019.

Running the Code

To run the code:

1. Clone the repository
2. Run mix deps.get
3. Run iex -S mix
4. Run one of the following Scenarios: <module>.run.

Scenarios

The Scenarios included in this repository are:

1. Playground.Scenario.Agent

   Use an Agent which holds a Map with integer keys.

   Tasks:

   • Bulk Load of count items
   • Concurrent Load of count items
   • Sequential Read of count items
   • Randomised Read of count items
   • Sequential Update of count items, incrementing the integer

2. Playground.Scenario.ETS

   Use an ETS table.

   Tasks:

   • Bulk Load of count items (in one go)
   • Concurrent Load of count items
   • Sequential Read of count items
   • Randomised Read of count items
   • Sequential Update Counter
   • Sequential Update Element
   • Sequential Lookup Element + Update Element

3. Playground.Scenario.Data

   Read constants from Module.

   Tasks:

   • Build Module with count items, based on List
   • Sequential Read of count items
   • Randomised Read of count items
   • Build Module with count items, based on Map
   • Sequential Read of count items
   • Randomised Read of count items

4. Playground.Scenario.Counters.One

   Comparison of Counters in ETS vs Atomics

   Tasks:

   • Sequentially update an ETS counter count times
   • Sequentially update an 1-arity atomics count times
   • Concurrently update an ETS counter count times
   • Concurrently update an 1-arity atomics count times
   • Concurrently get an ETS counter count times
   • Concurrently get an 1-arity atomics count times

5. Playground.Scenario.Counters.Many

   Comparison of Counters in ETS vs Atomics

   Tasks:

   • Sequentially update count ETS counters
   • Sequentially update a count-arity atomics
   • Concurrently update count ETS counters
   • Concurrently update a count-arity atomics
   • Concurrently get count ETS counters
   • Concurrently get count ETS counters again
   • Concurrently get count-arity atomics
   • Concurrently get count-arity atomics again

6. Playground.Scenario.Counters.Many.Atomics.One

   Deeper Comparison of various Atomics access patterns

   Tasks:

   • Sequentially update a count-arity atomics (sequential, ordered)
   • Sequentially update a count-arity atomics (sequential, randomised)
   • Concurrently update a count-arity atomics (concurrent, sequential, unordered tasks)
   • Concurrently update a count-arity atomics (concurrent, sequential, ordered tasks)
   • Concurrently update a count-arity atomics (concurrent, randomised, unordered tasks)
   • Concurrently update a count-arity atomics (concurrent, randomised, ordered tasks)

7. Playground.Scenario.Counters.Many.Atomics.Many

   Deeper Comparison of various Atomics access patterns

   Tasks:

   • Sequentially update count 1-arity atomics (sequential, ordered)
   • Sequentially update count 1-arity atomics (sequential, randomised)
   • Concurrently update count 1-arity atomics (concurrent, sequential, unordered tasks)
   • Concurrently update count 1-arity atomics (concurrent, sequential, ordered tasks)
   • Concurrently update count 1-arity atomics (concurrent, randomised, unordered tasks)
   • Concurrently update count 1-arity atomics (concurrent, randomised, ordered tasks)

8. Playground.Scenario.Northwind

   ETS backed Ecto interactions with Northwind Database as an example.

   Scenarios:

   • List all Employees
   • Insert / Delete Employee
   • Ingestion from JSON file
   • Select with Bound ID
   • Where
   • Select Where
   • Select / Update
   • Assoc Traversal
   • Promote to Customer
   • Stream Employees
   • Order / Shipper / Orders Preloading

Licensing & Legal Matters

Contents of this repository is under MIT license, except that you are not to publish the ETS adapter under Playground.Repository to Hex. This is because I would like to do so in the near future.

This project contains a copy of data obtained from the Northwind database, which is owned by Microsoft. It is included for demonstration purposes only.

Acknowledgements

Special thanks to:

• Mr Jay Nelson, for the excellent Erlang Patterns of Concurrency which served as inspiration for this talk.

• Mr Francesco Cesarini, for suggesting a deeper look into the performance characteristics of Atomics vs ETS-based counters.

• Mr Wojtek Mach, for the inspiration regarding creation of an Ecto adapter for ETS.

References

Further Reference Materials can be found within the Deck.