Skip to main content

Datasets

Standard Dataset

NAND Flash-based Synthetic Block IO traces from FIO benchmark

Citation Author(s):
Seungsu Baik
Submitted by:
Seungsu Baik
Last updated:
DOI:
10.21227/j9rb-5v64
Data Format:
No Ratings Yet

Abstract

NAND Flash-based Synthetic Block IO traces from FIO benchmark is focused on understanding characteristics of flash memoy under various access patterns. The dataset comprises distinct seven wrorkloads.

Workoad 1 and 2 are sequential write workloads. Workload 1 represents the typical seuqential write pattern, while Workload 2 represents partioning sequential write pattern, dividing 8 logical partitions.

Workload 3 and 4 are random write workloads. Workload 3 follows a uniform random distribution, and Worklod 4 follows the Zipfian random distribution with a theta of 0.8.

Workload 5, 6, and 7 include a combination of sequential and random writes, dividing 8 logical partitions.

  • Workload 5: 2 partitions with sequential writes and 6 partitions with random writes
  • Workload 6: 4 partitions with sequential writes and 4 partitons with random writes
  • Workload 7: 6 partitions with sequential writes and 2 partitons with random writes

Instructions:

There are 7 I/O traces, each represented in CSV (comma-separated text) file format.

The fields in the CSV files include timestamp, type, offset, and size.

Timestamp denotes the time when the write request was issued, measured from the timestamp of the first request (where the first request's timestamp is zero).

Type specifies whether the request is a 'Read' or 'Write,' although these traces exclusively consist of write requests.

Offset indicates the starting offset of the I/O in bytes from the beginning of the logical disk, and size represents the size of the I/O request in bytes.