dynamocacher

Description

A package which allows you to use dynamodb to cache output of a slow or expensive function, to reduce cost and increase the efficiency

Dax is supported using a Pynamodb fork, and its highly recommended for performance-critical application 50-100x performance difference is expected with dax

ToDos

• Dax support : Done
• Auto create table : Done
• policy template : in progress
• example readme : Done

Examples

colabNotebook

SampleUsage

cache this slow function

def power(input:dict):
  ''' 
  This is a very bad and slow function
  accept a dict containing these keys
    base: float
    power: int
  and return the power embarrasingly slowly
  response
    result: float
  '''
  base = input['base']
  power = input['power']
  result = base
  for i in range (power):
    result *= base
  return {
      'result':result
  }

def powerWithCaching(input:dict):
  cache = Cache.getCache(input, timeout = 3600)
  if cache: return cache
  result = power(input)
  Cache.addCache(input=input, output= result)
  return result

performance

Example code for trying it yourself

sampleInput = {
    'base':  1 + 1e-10,
    'power': 86289369
}

%timeit power(sampleInput) # 3.72 s ± 35 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
%timeit powerWithCaching(sampleInput) # 238 ms ± 482 µs without dax
%timeit powerWithCaching(sampleInput) # 5.43 ms with dax

SetUp

from dynamocacher.cacher import Cacher

class Cache(Cacher):
  class Meta:
      table_name = 'dynamoCache'
      region = 'us-east-1'
      aws_access_key_id = USER
      aws_secret_access_key = PW
      billing_mode= 'PAY_PER_REQUEST'
      dax_read_endpoints = ['....:8111'] # optional dax, note that dax will speed up dynamodb significantly
      dax_write_endpoints = ['....:8111']# optional dax, note that dax will speed up dynamodb significantly

createTable

Cache.create_table()

basic

addCache

sampleInput = {
    'query': 'testQuery',
    'fruit': 'strawberry'
}
Cache.addCache(
    input = sampleInput,
    output = sampleReturn
)

getCache

output = Cache.getCache(sampleInput, timeout = 3600) # ignore cache older than 3600 seconds

creating a table with sam

  Properties:
  TableName: dynamoCache
  PrimaryKey:
    Name: cacheKey
    Type: String
  Tags:
    Department: Engineering
    AppType: Serverless