These traces are published by Tencent Cloud to help researchers understand workloads in the cloud block storage systems, which includes subscription information and I/O traces for approximately 16,000 cloud virtual disks (CVD).
These traces were collected from a production cluster of Tencent Cloud’s block storage service. Specifically, 16,000 CVDs ware sampled from 10 warehouses in the Shanghai data center.
This repository contains all the compressed trace files in Cloud_Disk_dataset.zip.
Notes : The total size of data is about 1.3 GB compressed and 7.3 GB uncompressed.
The directory structure of the dataset is as follows:
/dataset
├── README.md
├── disk_subscription_info
└── disk_load_data
├── 0a0a7b84-2279-405e-9cbc-b395c725e01a
├── 0a0c1dea-723a-4484-a89c-53c02fc4b107
└── ...
disk_subscript_info
Each row of disk_subscript_info describes the subscription features of a CVD:
| Feature | Example | Description |
|---|---|---|
| disk_uid | 0a0a7b84-2279-405e-9cbc-b395c725e01a | A unique identifier for the CVD. |
| disk_attr | 0 | Product attributes corresponding to different service levels. The dataset includes two categories (0 and 1). |
| disk_type | 1 | Indicates whether the CVD is a data disk (0) or a system disk (1). |
| user_type | 0 | Indicates whether a user is an occasional (0) or regular (1) purchaser of cloud services. |
| vm_cpu | 4 | Number of CPU cores in the virtual machine attached to the CVD. |
| vm_memory | 32 | Memory size of the virtual machine attached to the CVD (GB). |
| disk_capacity | 50 | Subscribed capacity of the CVD (GB). |
disk_load_data
Each file in disk_load_data records the load of a CVD over a period of up to two months. The file name corresponds to the disk_uid. Each row in the file provides load information collected at 300-second intervals:
| Column | Example | Description |
|---|---|---|
| timestamp | 1593748800 | The collected timestamp, in microseconds. |
| read_IOPS | 11 | Read IOPS of the CVD at the time. |
| read_bandwidth | 50 | Read throughput of the CVD at the time (KB/s). |
| write_IOPS | 7 | Write IOPS of the CVD at the time. |
| write_bandwidth | 12 | Write throughput of the CVD at the time (KB/s). |
| disk_usage | 3360 | Disk usage of the CVD at the time (MB). |
We thank Difan Tan and Jiawei Li of Huazhong University of Science and Technology for their analysis and anonymization of the data.
The trace is a subset of the workload described in "TELA: A Temporal Load-Aware Cloud Virtual Disk Placement Scheme" in ASPLOS’25. When using this trace in your research, please make sure to cite the paper appropriately.
The trace data and document are licensed under CC-4.0.