https://github.com/avestura/serverless-paper-impl
https://github.com/avestura/serverless-paper-impl
Last synced: 7 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/avestura/serverless-paper-impl
- Owner: avestura
- Created: 2021-11-14T19:06:38.000Z (almost 4 years ago)
- Default Branch: main
- Last Pushed: 2022-07-06T07:40:11.000Z (over 3 years ago)
- Last Synced: 2025-01-13T17:32:45.490Z (9 months ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 9.1 MB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Serverless Paper Impl
My masters paper impl
## Tasks
### General
- [x] Implement Weilbull Stretched function
- [x] Implement Normalizer function
- [x] Implement "restore time" calculator based on function deps
- [x] Implement Russuan Roullete
- [x] Implement Similarity Function
- [x] Create function data generator
- [x] Fully-random generator
- [x] Uniform Dist
- [x] Normal Dist
- [x] Package-Frequency aware generator
- [x] Uniform Dist
- [x] Normal Dist
- [x] Implement the Synthetic Data Generator
- [x] Random Invocation time generator
- [x] Invocation Frequency aware generator
- [x] Write the event simulator based on synthetic generator
- [ ] Implement schedulers
- [x] No-merge
- [x] Random Wait
- [x] Static Wait
- [x] Dynamic Wait
- [x] Implement Co-op Network
- [x] Implement Digestra
- [x] Policies
- [x] Always Neutral
- [x] Context Based
- [x] Evaluate scheduler metrics
- [x] Using Fully Random Synthetic Data
- [x] No-merge
- [x] Random Wait
- [x] Static Wait
- [x] Dynamic Wait
- [x] Using Realistic Synthetic Data
- [x] No-merge
- [x] Random Wait
- [x] Static Wait
- [x] Dynamic Wait
- Massure approaches for each container:
- utilization: total working / (working + creation + wait + restoration)
- Response time = Restoration Duration
- Turnaround time = creation + total of restoration duration
- Throughput = always same
- Response/Cost Ratio
#### No-Merge approach
Terminate container after function terminates
### Random-Wait Approach
Wait for a random time after function terminates
### Static-Wait Approach
Wait for 5 mins after function terminates
### Dynamic-Wait Approach (this paper)
- [x] Create coop network
- [x] Weight Evaporation
- [x] Co-op probability calculator
- **WarmContainer**: Number of previous function runs to consider = 1
- **Container**::Number of reusing the container = k = 5
- **Contaner**::rMax = 20 min _// can't wait more than 20 min_
#### Policy
Waiting container should consider these data:
1. Similarity of its function to functions inside the running containers
2. Previous merge fail or success status for its function
3. co-op score
4. budget
## Statistical Distributions
- Normalaizer: **Stretched Weilbul** with `a = 5`, `b = 2`
- Number of Function dependencies: **Normal**, u = 50, stddev = 15
- Function Duration: **Chi-Square**, `v = 2.1`
- Function Run Request Every: **Normal**, `u = 30`, `stddev = 10`
- Random Wait Scheduler Wait Time: **Normal**, `u = 240`, `stddev = 50`
- Container Creation Time Cose: **Chi-squared**, `v = 5`
## Simulation Runner Engine
- schedulers
- noMerge
- staticWait
- randomWait
- dynamic :: AlwaysNeutral policy
- dynamic :: ContextBased policy
- configs:
- 8 different configs * numberOfFunctions * 5 different schedulers
1. for each appraoch, meassure configs * number of functions
2. for each scheduler, UFD_AW_N with increasing number of functions