eli-lim / springboot-portfolio-app

A tiny springboot app simulating a pricing engine publisher and subscriber

Portfolio

Components

flowchart LR
    PP[("Positions (CSV)")] -- positions --> Printer
    PES[PriceEngine] -- " schedule(.5 ~ 2s) " --> PES
    PES <-- securities --> SEC[(Securities DB)]
    PES -- updates --> PStore[(Price Store)]
    PES -- publishes --> E((PriceTick\nEvent))
    E -- notifies --> Printer[PortfolioPrinter]
    PStore -- prices --> Printer
    Printer -- prints --> C[Console]

Loading

1. All positions are read from a CSV file and are loaded into the PortfolioPrinter when the application starts.
   1. In reality, this would be from some dynamic source.
2. The PriceEngine schedules price ticks for all securities in the system, using Spring's out-of-the-box scheduler.
3. Each tick updates the PriceStore with the new price information in an atomic fashion (all options for a stock are updated in a single transaction).
4. A PriceTickEvent is published to all subscribers (in this case, only PortfolioPrinter)
5. On receipt of a PriceTickEvent, the PortfolioPrinter reads from the PriceStore, and prints the updated portfolio and its value to the console.
6. The components are wired together using Spring's IOC container, each behind interfaces to allow swapping out of implementations.

I designed this app largely with the following patterns in mind:

• CQRS; for separation of concerns between reading (PortfolioPrinter) and writing (PriceEngine)
  • Each component has a single responsibility
  • The readers and writers to the PriceStore are decoupled and can be scaled independently
• Event-driven
  • Akin to a pub-sub model, the PriceTickEvent is published to all subscribers, to simulate a real-world price publishing system where multiple consumers can subscribe to and act on changes in state.

Price Engine

• The most complex component in the system it is responsible for:
  • Simulating stock price ticks over a random time interval (between 0.5s to 2s)
  • Persisting price information
  • Implementing mathematical models for generating stock prices
  • Providing appropriate interfaces for consumers to subscribe to price ticks
• As the number of stocks increase, the number of writes to the price store increases linearly in the order of n * m, where n is the number of stocks and m is the number of options per stock.
• When a stock price ticks, all of its options should be updated in a single atomic transaction to ensure that consumers of the price store don't see partial updates and read inconsistent data.
• For the sake of simplicity, the price engine implemented for this assignment uses a single thread to simulate stock price ticks, leveraging Spring's async capabilities to schedule the writes in a non-blocking manner.
• In a real application, the writes to the PriceStore can be scaled horizontally by spinning up more threads / workers.

PriceTickEvent

• A simple event object that encapsulates information about a stock price tick.
• This is written in an event-driven manner to:
  • Decouple the PriceEngine from the PortfolioPrinter
  • Allow for multiple subscribers to the event in the future
  • Leverage Spring's async capabilities and abstractions instead of hand-rolling inter-thread communication

PortfolioPrinter

• A simple component that listens for PriceTickEvents, reads the latest prices from PriceStore and prints the updated portfolio to the console.

Testing Strategy

• Some tests have been written demonstrating the typical unit testing strategy.
• Bulk of the testing is focused on the PriceEngine given that it contains the bulk of the business logic.
• To test GeometricBrownianMotion, we initialize a static RNG (in the test only) with a fixed seed and compare the output with expected values.

Follow-ups / Improvements

• To avoid IEEE-754 floating point precision issues, we can use BigDecimal for storing prices.
• The PriceEngine can be scaled horizontally by spinning up more threads / workers to perform price ticks for multiple stocks in parallel.
• The PriceStore could be replaced with a durable, performant data store like Redis to provide quick look-ups and atomic writes.
  • Different architectures and data fetching / eager loading strategies can be explored if more extreme performance demands is required.

---

Running the application

(Recommended) via IDE

1. Import the project into your favorite IDE (IntelliJ IDEA, Eclipse, etc.)
2. Run the main method in PortfolioApplication

via command line

1. Run ./gradlew bootRun from the root of the project