| 1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768 |
- Theme
- Data movement and the cost thereof
- Latency
- Time per operation
- CPU cycles per operation
- Bandwidth
- MB / sec
- CPU cycles / MB
- Media
- Memory (load, bcopy)
- Disk (randoms, sequentials)
- File system (directory ops, sequential)
- Network (hot potato, transfer)
- Pipes (hot potato, transfer)
- VM system (mmaps/munmaps, bcopy)
- Systems
- All Unix systems
- Windows NT
- VMS (?)
- Mainframes (?)
- Memory
- Small transfers (randoms)
- Load latency
- Large transfers (sequential)
- Bcopy bandwidth
- Processes
- Null process execution time
- Context switching
- Misc
- Null entry into the system
- Networking
- Small transfers (randoms)
- Transfers per second
- CPU cycles per transfer
- socket/bind/close per second
- Large transfers (sequential)
- MB per second
- CPU cycles per MB
- Disks
- Small transfers (randoms)
- Transfers per second
- CPU cycles per transfer
- Large transfers (sequential)
- MB per second
- CPU cycles per MB
- File system
- Small transfers (randoms)
- Creates / second
- Removes / second
- Random I/O's per second in large file
- CPU cycles per transfer
- MB / sec when reading many related small files
- Large files
- MB / second read/write
- CPU cycles per MB
- Hardness
- Measure fsck time?
- Virtual memory system
- Creation
- mmaps per second
- munmaps per second
- Also vary size of mapped region
- Small transfers (randoms)
- Random reads per second of large mmaped file
- CPU cycles per read
- Large transfers (cached sequential)
- MB per second read rate
- CPU cycles per MB
|
