Equation solving at scale
using Apache Spark
- Praveen Rachabattuni, Sigmoid
@praveenr019
What are we solving?
• Calculate specialised rate of return of investment
data of different customers of a investment
company.
• Compute the returns of 1000s of customers on a
near real-time.
Money weighted rate of
return (MWRR)
MWRR Example
• V1 = 298,082
• V0 = 250,000
• 25,000 investment made on 258th day
• r = 0.00897 or 8.97%
4
Newton-Raphson Solver
• xn+1 = Better approximate
• xn = Initial approximate
• f(xn) = Function value at a value of x
• f’(xn) = Function differential value at a value of x
Demo
Spark
• Computation should all happen in executors
• Custom map operations should be well optimised
Apache Common Solvers
• BisectionSolver
• MullerSolver
• MullerSolver2
• NewtonRaphsonSolver
• SecantSolver
Queries ?
Thank you

Equation solving-at-scale-using-apache-spark

  • 1.
    Equation solving atscale using Apache Spark - Praveen Rachabattuni, Sigmoid @praveenr019
  • 2.
    What are wesolving? • Calculate specialised rate of return of investment data of different customers of a investment company. • Compute the returns of 1000s of customers on a near real-time.
  • 3.
    Money weighted rateof return (MWRR)
  • 4.
    MWRR Example • V1= 298,082 • V0 = 250,000 • 25,000 investment made on 258th day • r = 0.00897 or 8.97% 4
  • 5.
    Newton-Raphson Solver • xn+1= Better approximate • xn = Initial approximate • f(xn) = Function value at a value of x • f’(xn) = Function differential value at a value of x
  • 6.
  • 7.
    Spark • Computation shouldall happen in executors • Custom map operations should be well optimised
  • 8.
    Apache Common Solvers •BisectionSolver • MullerSolver • MullerSolver2 • NewtonRaphsonSolver • SecantSolver
  • 9.
  • 10.