1. Files 
 
This distribution contains the following files: 

a) README.txt: this file 
b) dataset1/2/3.tar.gz: IO trace logs, described below 


 
2. IO trace logs 


 
These IO trace logs are created by capturing the FUJITSU K5 cloud service, which is operated in over 25 data centers across four continents and includes more than 5,300 cloud customers with over 10,000 expert cloud employees. It is also supported by more than 13,000 servers and 640 business systems and built using OpenStack and VMware platforms. We installed a capture system in the TCP/IP network between servers running on hypervisor and storage systems in a K5 data center in Japan. The capture system copies IO requests and replies without overheads because of using network taps. We retrieved these IO access logs by using the capture system. These IO access logs consisted of 3088 virtual storages.  
 We gathered these IO access logs from Monday to Sunday but not continuously because one day's IO access logs often consumed the storage capacity of the capture system. We divided these seven day of IO access logs by each virtual storage volume id. We created three sets of these seven day's IO access logs (dataset1.tar.gz, dataset2.tar.gz, and dataset3.tar.gz). Each virtual storage volume id is unique in the same dataset, while each virtual storage volume id is not unique between the different datasets.  
 Each dataset has the following directory structure.  
 
           Date and time    IO access log file for a virtual storage volume id 
dataset1 --+-- 20171003 --+-- 0 
          |             +-- 1 
          |             ...
          +-- 20171012 --+-- 0 
                        +-- 1 
                        ... 
 
3. IO trace log format 


 
The fields in the IO access log are:  
 
ID, Timestamp, Type, Offset, Length 
 
ID is the virtual storage volume id.  
Timestamp is difference value from the first IO request of all IO access logs in seconds. 
Type is 'R(Read)' or '(W)Write'. 
Offset is the starting offset of the IO access in bytes from the start of the virtual storage.  
Length is the transfer size of the IO request in bytes.  
4. Attribution 


 
Please cite the following publication as a reference in any published work using these IO access logs. The publication includes the analysis for dataset1.  
 
Analysis of commercial cloud workload and study on how to apply cache methods 
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). 
https://www.ieice.org/ken/paper/20180730o1Fe/ 
or 
https://www.researchgate.net/publication/327834506_Analysis_of_commercial_cloud_workload_and_study_on_how_to_apply_cache_methods 
 
Thank you.