benchmarks

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