Conference Scheduler

This application generates a conference schedule algorithmically using constraints.

Pre-requisites

To build and execute this, you need the following installed on your system:

• Java 8
• Maven 3

Build and execution

Build

Go to the home folder of this project and build the package.

cd conf-sked
mvn clean package

Execute

After the package has been built, execute the jar.

java -jar target/conf-sked.jar

Sample output

Day 1
09:00 Create a Serverless Searchable Media Library KEYNOTE
09:30 Deep Learning for Edge to Cloud WORKSHOP
 -or- AWS DevOps Essentials: An Introductory Workshop on CI/CD Best Practices WORKSHOP
10:30 Machine Learning for the Enterprise SESSION
11:00 AWS DeepLens Projects: Tips & Tricks LIGHTNING
11:10 Deep Learning for Developers: An Introduction SESSION
11:40 Unlock the Full Potential of Your Media Assets SESSION
12:10 Create the Voices You Want with Amazon Polly LIGHTNING
12:20 Amazon SageMaker and TensorFlow: Tips & Tricks LIGHTNING
12:30 LUNCH
13:30 Build an ETL Pipeline to Analyze Customer Data WORKSHOP
14:30 Make Money with Alexa Skills SESSION
15:00 TEA
15:15 Monitor Your AWS Usage with Amazon ES WORKSHOP
16:15 Technology Trends: Data Lakes and Analytics SESSION
16:45 Scale up a Web Application LIGHTNING
17:00 Create a Custom Celebrity List for Your Media Assets CLOSING

Day 2
09:00 Improve Accessibility Using Machine Learning KEYNOTE
09:30 What's New in Amazon Athena SESSION
10:00 Log Analytics with Amazon ES, Amazon ECS, and Amazon EKS WORKSHOP
11:00 AWS Reference Architectures SESSION
11:30 Serverless Architectural Patterns and Best Practices SESSION
12:00 Chaos Engineering and Scalability at Audible.com SESSION
12:30 LUNCH
13:30 Real-Time Analytics with AWS SESSION
14:00 Amazon SageMaker and Chainer: Tips & Tricks LIGHTNING
14:10 Three Core Pillars of Conversational UX for Alexa SESSION
15:00 TEA
15:15 Serverless Data Prep with AWS Glue WORKSHOP
17:00 Effective Cost Optimization for Business CLOSING

Note: The sample talks are from AWS re:Invent

Customize

You can modify the property file scheduler.properties to change the different conference parameters or use your own property file from a different location by adding it in the command line argument.

java -jar target/conf-sked.jar /path/to/mycustom.properties

Docker image

If you are familiar with docker, a docker image is also available to run this without downloading the code. The DockerHub image is adr1/conf-sked.

Running the image

This will run the application using the default settings and talks.

docker run --rm adr1/conf-sked

Running the image with mounted volumes

This will run the application using your own settings and talks at the current working directory mounted to /conf on the container.

docker run --rm -v $PWD:/conf adr1/conf-sked /conf/scheduler.properties

Specifications

Talk types

There are six types of talks.

Talk type	Duration	Data type
Workshop	60 mins	WORKSHOP
Chalk talk	60 mins	CHALK
Session talk	30 mins	SESSION
Lightning talk	10 mins	LIGHTNING
Keynote	30 mins	KEYNOTE
Closing	30 mins	CLOSING

Breaks

There are two break types.

Type	Duration
LUNCH	60 mins
TEA	15 mins

In the code, breaks are still defined in the the TalkType enum.

Constraints

Here are the specific constraints the scheduler will follow when generating the code:

• Conference runs for two days (29-30 Jun 2018).
• Each day starts with a 30-min keynote speech at 09:00.
• Each day ends with a 30-min closing speech at 17:00.
• Day starts at 09:00 and ends at 17:30.
• There's a 60-min lunch break at 12:30 to 13:30 daily.
• There's a 15-min tea break at 15:00 to 15:15 daily.
• All talks must be scheduled.
• Talks in parallel is allowed to accommodate all talks.

The constraint parameters can be modified by changing the property file scheduler.properties

Design

Analysis

• Talks at 09:00 to 09:30 and 17:00 to 17:30 are fixed for KEYNOTE and CLOSING, respectively.
• Breaks at 12:30 to 13:30 and 15:00 to 15:15 are fixed for LUNCH and TEA, respectively.
• Since the constraints are mostly predefined timings, we will split a day into multiple blocks where each block applies the MustStartWithConstraint implementation.
  • The MustStartWithConstraint means that a block must start with a specific talk type only.
  • The rest of the talks scheduled on that block can be of any type except KEYNOTE and CLOSING.
• In this case, we will split a day into four blocks with the following conditions:
  • Block B1 starts at 09:00 and ends at 12:30 and must start with KEYNOTE.
  • Block B2 starts at 12:30 and ends at 15:00 and must start with LUNCH.
  • Block B3 starts at 15:00 and ends at 17:00 and must start with TEA.
  • Block B4 starts at 17:00 and ends at 17:30 and must start with CLOSING.

Data model

• A Conference contains one or more days.
• A Day contains one or more blocks.
• A Block contains one or more slots.
• A Slot contains one talk with start and end times.
• A Talk contains the talk details and maps to the JSON input file.

Scheduler

The Scheduler is the main processor that schedules the conference given a pool of talks. All the blocks from each day in the conference are iterated and filled with slots until all the talks in the pool is exhausted. If all the blocks are filled and the pool still has talks not scheduled, these talks will be scheduled in parallel with other slots. The parallel talks will be of the same type to sync the timings.

Constraint

The schedule is applied against a Constraint defined for each block. Specific rules or requirements are implemented using constraints. An example constraint MustStartWithConstraint is included here.

Extensibility

New constraint implementation can be added to apply different rules when processing the scheduling.

Implement the interface Constraint to add a new constraint.

Limitations

The current algorithm used in this project uses the MustStartWithConstraint only. The following are some of its limitations:

• The talks are searched from the pool in the same order as they are found in the input file.
• The scheduling is not optimal and some free time slots are not used fully.

For improvements

• Schedule talks based on content using their tags.
• To maximize the time slots of each block:
  • If a block is large, prioritize the long running talks in it.
  • If a block is small, prioritize the short running talks in it.

Testing

Unit tests are included to check if the requirements are met.