Cloud Stateless System Performance Metrics and Status

Citation Author(s):
Nutt
Chairatana
Department of Robotics and AI Engineering, School of Engineering, King Mongkut's Institute of Technology Ladkrabang, Bangkok, Thailand
Rathachai
Chawuthai
Department of Computer Engineering, School of Engineering, King Mongkut's Institute of Technology Ladkrabang, Bangkok, Thailand
Submitted by:
Nutt Chairatana
Last updated:
Wed, 01/31/2024 - 12:12
DOI:
10.21227/8wf2-2y40
Data Format:
Link to Paper:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

We utilized Digital Ocean's cloud service, setting up three Linux virtual machines, each with 1vCPU, 1GB of memory, and a 10GB disk. The architecture included an API gateway for routing requests to a stateless application service backed by a database for storing application data. The application operates the service under a fluctuating workload generated by a load-testing script to simulate real-world usage scenarios. The target source or the application service is integrated with Prometheus, a monitoring tool for gathering system metrics. To extract data from Prometheus, we devised a custom script capable of tapping into its local storage, thereby collecting resource utilization and performance metrics. The resulting dataset encompasses roughly 8,000 data points gathered at 5-second intervals. These data points span a variety of metrics: CPU and memory usage (in percentages), network traffic (inbound and outbound rates in GB/s or MB/s), transactions per second (TPS), and response times (in seconds or milliseconds). A critical aspect of our dataset was the real-time health status of the system, assessed through HTTP response codes. Using our custom script, we monitored these codes; if predefined error codes (5xx) were detected, the system was marked as "unhealthy." In all other scenarios, the system was deemed "healthy." 

Instructions: 

Data features include 

  • Time
    • labeled as Time in the CSV header
    • expressed as yyyy-mm-dd hh:mm:ss
  • Timestamp
    • labeled as Timestamp in the CSV header
    • expressed as milliseconds
  • CPU Request
    • labeled as cpu_usage in the CSV header
    • expressed as percentage
  • Memory Request
    • labeled as memory_usage in the CSV header
    • expressed as percentage
  • Inbound Bandwidth
    • labeled as bandwidth_inbound in the CSV header
    • expressed as gigabytes per second (GB/s) or megabytes per second (MB/s)
  • Outbound Bandwidth
    • labeled as bandwidth_outbound in the CSV header
    • expressed as gigabytes per second (GB/s) or megabytes per second (MB/s)
  • Transactions Per Second 
    • labeled as tps in the CSV header
    • expressed as requests per second (req/s)
  • Average Response Time
    • labeled as response_time in the CSV header
    • expressed as seconds (s) or milliseconds (ms)
  • System Status
    • labeled as status in the CSV header
    • expressed as a binary label - 0 for healthy and 1 for unhealthy