EdgeWorkloadsTraces

Paper: [IMC'21] From Cloud to Edge: A First Look at Public Edge Platforms

Dataset list

• NEP-small: The workloads traces of 14 edge sites of China's largest public edge platform during June 2020 (5-min per record).
• NEP-large: The data of 139 edge sites of China's largest public edge platform during June 2020 (5-min per record). (In this data set, we additionally provide three-month continuous records for bandwidth trace, but the granularity is 1 hour.)

Dataset description

Each dataset contains the workloads traces (CPU, memory, storage, bandwidth) at virtual machine (VM) granularity from a large-scale edge cloud platform in China.

Specifically, each dataset consists of five CSV files.

• VM_CPU.csv: the CPU usage data of each VM;
• VM_BANDWIDTH.csv: the bandwidth (both uplink and downlink) usage data of each VM;
• SITE_RTT.csv: the RTT among edge sites;
• VM.csv: the VM affiliation table, including customer id, mirror id, specifications, etc;
• PM.csv: physical machine(PM) ID, the CPU, memory, and storage size of each physical machine.

Data sample

VM_CPU.csv

(1) vm_id is the VM ID; (2) site_id is the edge site ID; (3)cpu_rate is the CPU usage. For example, 0.6 indicates 60%; (4) report_ts is the Unix timestamp of data collection.

VM_BANDWIDTH.csv

pub is the public network, pri is the internal network, up is the uplink, down indicates the downlink, bw is the bandwidth usage (bps), and flow is the number of flows

SITE_RTT.csv

VM.csv

uid is the customer ID; pm_name is the physical machine (PM) ID where the VMs are hosted; status is the status of instance; image_id is the system image ID used by the instance;os_type is the OS type; os_name is the OS name.

PM.csv

cores, memory, and storage are the number of CPU cores, memory size (MB), and disk size (MB) of the physical machine, respectively.

Statement

1. The data can be only used for research purpose.
2. The data can not be shared offline.

To request the data, please submit a form here: https://forms.gle/j3QDp9qtCVyrcTwm9, and we will response AFAP. Contact Mengwei Xu ([email protected]) for any inquiry.