The real-world CDN trace used in the paper is published here, hoping to stimulate more studies focusing on efficient content distribution in futuristic space-terrestrial integrated networks.
The trace was collected from an nanonymous commercial cloud CDN operator on February 22-28, 2015, containing 3.9 million flow records in total. Some details of the trace include:
Parameter | Value |
---|---|
# of Total Requests | 3,998,185 |
# of Total Bytes Requested | 4137 TB |
50th/90th Obj Size | 256KB / 1.3 GB |
Duration | 7 days (February 22-28, 2015) |
The requests are given in time order and all the privacy-related information was ommitted.
Each request in the dataset is given in a comma-separated format with 5 columns:
- Seq : The sequence number of the request in this dataset, starting from 0.
- Timestamp: The time when the request was issued, given in UTC timestamp format.
- ObjectID: The unique object ID of the content requested.
- Size: The size of the content requested, given in B (Byte).
- TimeToComplete: The time spent to complete the content request, given in s (second).
The first five requests are given as example (no column name information was included in the dataset):
Seq | Timestamp | ObjectID | Size | TimeToComplete |
---|---|---|---|---|
0 | 1424534404 | 0x6cc00e85eacfc56c | 31929800 | 111 |
1 | 1424534409 | 0xf2e79b2ae42ea5b | 833462272 | 658 |
2 | 1424534411 | 0x8b4c8304587a4ad5 | 1048576000 | 10070 |
3 | 1424534412 | 0x536cf07acc3eeb9d | 9217283 | 98 |
4 | 1424534413 | 0xc48b2b9b7b00e2f3 | 649075360 | 1892 |
The dataset is available here: download link.
Note that a few illegal records were erased, for the convenience of the dataset users, and thus some details of the public download version may not perfectly match with what stated above.