mehdimo / data-generator-based-on-google-trace

A script to generate dataset based on Google cluster traces for Aution-based Internet of Things (IoT) applications

Data Generator based on Google Trace Dataset

The dataset generated by this application is an adaptation of google cluster trace dataset. We used 'task events' table as the source of requests for Auction-based IoT applications. In the processed dataset, we have the following fields separated by comma.

• Job id
• Task id; job id and task id taken directly from the original dataset. They together represent the unique requests in the system.
• Arrival time
• Finish time
• Delay sensitivity (ds), a number between 0 and 3 (3 means that a request is more delay sensitive).
• Service rate (mu), assigned base on the delay sensitivity of the request and a random uniform number between -0.005 ~ 0.005.
• Number of clouds, which is picked proportional to the delay sensitivity (ds). It is a random number between 2^(ds) and 2^(ds+1).
• Payment, calculated based on the requested resource usage (cpu, memory and disk) and an approximate standard google cloud pricing plan.

    computePrice = 0.0006  // dollor per minute
    diskPrice    = 0.00015 // dollor per minute
    unitPay = computePrice * (cpuUsage + memoryUsage) + (diskPrice * diskUsage)
    payment = (1+unitPay) * (duration* mu * N) 

The order of the fields is (Job id, Task id, arrival time, finish time, delay sensitivity, payment, service rate, number of clouds).