ServerTrack

Performance tracking service, with data residing solely in server memory, employing a threaded model for concurrent request handling.

What inspired this?

It's a response to a coding challenge I addressed during the application process with a major Internet infrastructure firm, moving into the cloud services industry, and looking to automate service infrastructure management.

The assignment required that I complete this within a 24 hour period. In the interest of candid review, I'll keep at least one branch in this project at the state it was for final delivery within that 24 hour window.

There are a few areas where I can imagine this project being improved, namely in data structure and query performance. There are also a few minor errors, but they're largely inconsequential to the task's objective.

Design Considerations

Recording of incoming data needs to be fast, as presumably it'll run at much higher volume. Reporting (i.e. querying) of the dataset should still be fast as possible, but this is a secondary priority.

This provides a simplistic RESTful HTTP interface for both recording (data in-take) and reporting. As such, individual hosts (reporting their performance) are treated as "collections" in the REST sense. Individual samples are not accessible through the REST API though.

Recording

POST /perf/{hostname}/

cpuload={cpuload}&memload={memload}

Parameters

• cpuload a float/double, representing CPU load percentage
• memload a float/double, representing Memory load percentage

Append a record to the performance history of the given hostname.

Reporting

GET /perf/{hostname}/{mode}

Mode is required and can be one of:

• last_hour: list average load values by minute for the preceding 60 minutes
• last_day: list average load values by hour for the preceding 24 hours
• last_minute: as above, values by seconds for preceding 60 seconds (for testing)

The response is given in JSON.

Example Output

[   
    {"memload": 83.47, "cpuload": 4.23, "interval": 1, "samples": 59, "time": "2015-03-14T04:16:11"}, 
    {"memload": 85.21, "cpuload": 4.35, "interval": 1, "samples": 58, "time": "2015-03-14T04:16:12"}, 
    ...
]

The values memload and cpuload represent averages for the given time intervals.