The test bench generates and measures various file operations. Iozone has been worn on many machines and operates under many operating systems.
Iozone is useful for performing a broad analysis of the file system of a vendor's computer platform. The test bench tests the file I / O performance for the following operations:
- Read
- Write
- Reread
- Rewrite
- Reading back
- Read to stride
- Fread
- Write
- Random reading
- Pread
- Mmap
- Aio_read
- Aio_write
Although computers are usually purchased with an application in mind, it is also likely that over time, the mix of applications will change. Many vendors have improved their operating systems to work well for some frequently used applications. Although this speeds up the I / O for these few applications, it is also likely that the system will not work well for other applications that have not been targeted by the operating system. An example of this type of improvement is: Database. Many operating system vendors have tested and tuned the file system for it to work well with databases. Although database users are satisfied, other users may not be as happy as the entire system can give all system resources to users of the database to the detriment of all other users.
Over time, the system administrator may decide that some additional office tasks might be transferred to this machine. The load can now go from a random read application (database) to a sequential drive. Users can discover that the machine is very slow when running this new application and become dissatisfied with the decision to buy this platform. By using Iozone to gain extensive file system performance coverage, the buyer is more likely to see hot or cold spots and choose a more balanced platform and operating system.
Test Features:
- Source ANSII C
- POSIX Asynchronous I / O
- Mmap () file I / O
- Normal File I / O
- Single flow measurement
- Multiple Flow Measurement
- Distributed File Server (cluster) Measures
- POSIX pthreads
- Multi-process measurement
- Excel importable output for chart generation
- Latency Plots
- 64bit compatible source
- Large compatible file
- Stonewalling in flow tests to eliminate lagging effects
- Configurable Processor Cache Size
- Selectable measurements with fsync, O_SYNC
- Designed for: AIX, BSDI, HP-UX, IRIX, FreeBSD, Linux, OpenBSD, NetBSD, OSFV3, OSFV4, OSFV5, OpenServer SCO, Solaris, MAC OS X, Windows 95/98 / Me / NT / 2K ) / XP)
The NFS3 read latency graph shows the latency of 4k read operations on a NFS version 3 file system mounted on a client running Iozone. Click on the image to enlarge.
The NFS3 replay latency graph shows the read latency of 4k reads on a NFS version 3 file system. The effects of the client-side cache can be clearly seen. Here, the latencies are the same as the latencies of the local buffer cache. Click on the image to enlarge.
[ad_2]
Source link