Parallel Traces
Parallel traces, generally taken from supercomputers or in large datacenters, record activity from multiple computers running in parallel.
Similar traces can also be found in the Key-Value Traces section.
The following traces are free to download under the terms of the
SNIA Trace Data Files Download License.
Please note that cookies must be enabled within your browser in order to
download traces.
For questions about downloading using a shell script,
see Using Shell Scripts,
and for more information about downloading using a Windows
batch script, see
Using Batch Scripts.
| Trace Name | Details | Details | Related Tools | Year Recorded | Timespan | Record Count | File Size | Actions |
|---|---|---|---|---|---|---|---|---|
| Thesios Synthetic | Synthetic traces from Google's storage servers, generated by downsampling raw traces and published in the paper Thesios: Synthesizing Accurate Counterfactual I/O Traces from I/O Samples by Phitchaya Mangpo Phothilimthana, Saurabh Kadekodi, Soroush Ghodrati, Selene Moon, and Martin Maas, ASPLOS '24 | Synthetic traces from Google's Thesios project, ASPLOS '24 | 2024 | 2 months | 2.53 Billion | 65.7 GB |
|
|
| HPC I/O System Pattern Traces | Traces related to On server-side file access pattern matching from HAL-Inria. These traces are hosted on Zenodo due to access restrictions. | Traces related to On server-side file access pattern matching from HAL-Inria. These traces are hosted on Zenodo. | 2018 | about 8 hours | 39.6 Million | 671 MB |
|
|
| Tencent Block Storage | Traces used for the paper OSCA: An Online-Model Based Cache Allocation Scheme in Cloud Block Storage Systems by Yu Zhang, Ping Huang, Ke Zhou, and Hua Wang. In Usenix ATC '20. | Traces from OSCA: An Online-Model Based Cache Allocation Scheme in Cloud Block Storage Systems by Yu Zhang et al. | 2018 | 9 days | 33.6 Billion | 198 GB |
|
|
| K5cloud Traces | Traces collected for the paper Analysis of commercial cloud workload and study on how to apply cache methods by Kazuichi Oe(1), Kazutaka Ogihara(1), and Takeo Honda(2), 1: Fujitsu Laboratories Ltd. 2: Fujitsu Software Technologies Ltd. In IEICE CPSY (SWoPP 2018), Kumamoto, Japan (July 2018). | Traces from Analysis of commercial cloud workload and study on how to apply cache methods by Kazuichi Oe et al. | 2017 - 2018 | 6 months | 17.1 Billion | 139 GB |
|
|
| Tencent Photo Cache | Traces used for the paper Demystifying Cache Policies for Photo Stores at Scale: A Tencent Case Study by Ke Zhou(1), Si Sun(1), Hua Wang(1), Ping Huang(1,2), Xubin He(2), Rui Lan(3), Wenyan Li(3), Wenjie Liu(2), and Tianming Yang(4). 1: Wuhan National Laboratory for Optoelectronics. 2: Temple University. 3: Tencent, Inc. 4: Huanghuai University. In ICS-2018: The 32nd ACM International Conference on Supercomputing, Beijing China, June 2018. | Traces from Demystifying Cache Policies for Photo Stores at Scale: A Tencent Case Study by Ke Zhou et al. | 2016 | 8 days | 5.68 Billion | 138 GB |
|
WARNING: These traces are over 10 years old! They should not be used for modern research!
The following traces are free to download under the terms of the
SNIA Trace Data Files Download License.
Please note that cookies must be enabled within your browser in order to
download traces.
For questions about downloading using a shell script,
see Using Shell Scripts,
and for more information about downloading using a Windows
batch script, see
Using Batch Scripts.
| Trace Name | Details | Details | Actions | |||||
|---|---|---|---|---|---|---|---|---|
| Mambo Traces | Traces for seven parallel I/O-intensive application which were run on eight nodes of an IBM SP-2. Used the AIX trace utility to trace I/O-related system calls (open, close, read, write and seek). Some characteristics of these traces have been described in this University of Maryland Technical Report. Due to access restrictions, these traces are externally hosted. | Traces for seven parallel I/O-intensive application which were run on eight nodes of an IBM SP-2. Some characteristics of these traces have been described in this University of Maryland Technical Report. | 1996 | about 1 month | 0 Bytes |
|
||
| Sprite Traces | The software for opening these traces no longer functions, so they are currently unreadable. Traces were collected from approximately forty workstations connected to four servers. Further information about the traces is available at the Berkeley site, including documentation, software, and sample code. *Please contact us if you successfully read these traces. | May 1991 traces from multiple servers at UC Berkeley. *Please contact us if you successfully read these traces. | These traces are stored in a binary format, code for processing them can be found here and the OS is here. | 1991 | 5 months | 218 MB |
|