An Empirical Study of Task Switching Locality in MVS
IEEE Transactions on Computers
Cache performance of operating system and multiprogramming workloads
ACM Transactions on Computer Systems (TOCS)
ACM Transactions on Computer Systems (TOCS)
The Stack Growth Function: Cache Line Reference Models
IEEE Transactions on Computers
A bibliography for multiprocessor cache memories
ACM SIGARCH Computer Architecture News
Second bibliography on Cache memories
ACM SIGARCH Computer Architecture News
ACM Computing Surveys (CSUR)
Cold-start vs. warm-start miss ratios
Communications of the ACM
Use of the LRU stack depth distribution for simulation of paging behavior
Communications of the ACM
Properties of the working-set model
Communications of the ACM
The working set model for program behavior
Communications of the ACM
Bibliography and reading on CPU cache memories and related topics
ACM SIGARCH Computer Architecture News
Program Behavior: Models and Measurements
Program Behavior: Models and Measurements
| Hi-index | 0.00 |
A hierarchical program behavior model in a multitasking environment was proposed and applied to a cache multitasking model for performance evaluation. The hierarchical program behavior model consists of the task switching model, execution interval model, and the line (block) reference behavior model for each individual task. An execution interval is a continuous execution of a task between task switches. As a task executes in an execution interval, it brings its lines into a cache according to the line reference behavior model. The Stack Growth Function (SGF) model was used for this purpose. The state of a cache is defined by the numbers of lines of the individual tasks. The state of a cache at task switches then constitutes an imbedded Markov chain. Although a set of simultaneous linear equations in steady state cannot exactly be solved practically because of its excessively large state space, it can be solved very efficiently by a Monte-Carlo simulation. The model was validated against the miss rate measured by a hardware monitor in a controlled environment on a mainframe running IBM MVS operating system.