THE FRACTAL STRUCTURE OF DATA REFERENCE- P29 ppsx

Index Access density, 113, 120, 123, 125 Arrival rate, 35, 39 Back end I/O, 15 Back end, 12,15 Generation, 75 - 76 Benchmark, 35, 42 Binary tree, 37 Heavy - tailed, 1, 7, 85, 98 Burst, 6, 88 CA - ASTEX, 14 Cache Analysis Aid, 14 Cache visit, 13 Cache 107, 110 Front - end miss, 23 Garbage collection, 71 Generalized LRU, 67-68 Granularity, 11, 86, 91, 96 Henley, M., 5 Hierarchical reuse model, 1, 5, 7, 35, 47, 56, 78, 80, Hierarchical storage management, 97, 99, 103 - 04, 87-88, 106 memory, 16, 22, 43-44, 51-52, 56, 59, 61-63, Hi n t s , 53 66, 68 Hiperpool, 55 plain vanilla, 2 Hiperspace, 55 residency time, 9, 22, 26, 43-45, 47-48 storage control, 1 - 3, 8, 22, 25, 51-53, 55 - 56, 59 visit, 11 History dependent, 76, 81, 83 History independent, 76 Hit ratio, 2, 59 Hit, 2, 52 Homogeneous pattern of updates, 77, 80 Hyperbolic, 7 I/O interface, 51, 53 IMS, 22, 53, 89, 97, 110 Independent reference model, 3, 5, 69 Interarrival lime, 8–9, 78, 85, 87 Level 0 storage, 103 Level 1 storage, 103, 105, 112 Level 2 storage, 103 Linear model, 74, 77, 80, 83 Little’s law, 13, 16, 24, 79, 81 Little, J. D. C., 13 Log structured array, 71 Log-structured disk subsystem, 71 Calibration, 107 CICS, 18, 53, 89, 97, 110 Collection threshold, 76 Compression, 71, 103 - 104, 112 Constrained optimization, 104 - 105, 108 Contour plot, 110 Criterion of time - in - cache, 8, 14, 52 Cylinder image, 86 David - Johnson approximation, 38 DB2, 18, 53, 74, 89, 97, 110 Deferred write, 74 Deployable applications model, 114, 120, 123 - 124 Destage, 72, 74, 82 - 83 Dirty data, 82 Disk array, 71 DL/I, 22 Dormitory, 74 LRU, 2, 55, 61, 63 Driver, 35, 42 LRU - K, 68 Early demotion, 24, 55 ECKD, 2 Fractal, 11, 42 Free space collection, 71, 74, 77, 83 Front end, 12 - 13, 15, 87 Levy, Paul, 8 Mandelbrot, Benoit B., 7, 118 Memory hierarchy, 1, 5, 27 Memoryless, 1, 3, 5 Migration age, 105, 110 - 111 132 Migration, 103, 106 Miss ratio, 2, 16-17, 47, 66 Miss, 2 Skew, 122 Moves per write, 72, 80 MRU, 2 Multiple workload hierarchical reuse model, 33, 47, Operational conventions, 82 OS/390 Stage, 2, 55 THE FRACTAL STRUCTURE OF DATA REFERENCE Single - reference residency time, 8, 14, 36, 43, 52 - 53, 56, 63, 87 SMF, 46, 86, 96, 110 SMS Optimizer, 110 SMS, 97, 103 61 Solid state disks, 4 Stack distance, 42 Steady state, 86 Storage cost, 120 Storage intensity, 1 16, 118, 120 Synthetic application, 37 Tape robotics, 109 - 110, 112 Thiébaut, Dominique, 42 Tivoli Storage Manager, 103 Touch, 15, 88 Toy application, 35, 37 Track image, 2, 86 Transaction volume, 115, 117 Transient, 1, 61, 73, 80, 85 - 87, 89, 91, 99 TSO, 22, 97, 110 Usable capacity, 124 - 125 environment, 2, 9, 14, 23 - 25, 46, 103 workload measurements, 6, 18, 27, 89, 97, 110 Page frame, 11 Page, 2 Storage utilization, 76-77 Pareto, Vilfredo, 8 Partitioned memory, 63 Persistence metric, 87 - 88, 91 Persistent, 85 - 86, 89, 91, 96, 99, 101 Power law, 8, 17, 118 Primary storage, 97, 103,105, 112 Processor file buffer, 1, 9, 25, 33, 5 1, 53, 56, 59 Productivity, 104 Random walk, 37 Recall, 103, 106 Record, 2 Update - in - place, 7 1 Replacement criterion, 2 SAS, 110 VLF/LLA, 55 Secondary storage, 103 VM Segment, 74 Self - similar, 7, 1 18 Sequential, 24, 68, 75 Window, 87, 89, 91 Service time, 116 Simulation, 39 - 40, 52 - 53 environment, 2, 9, 14, 24 workload measurements, 9, 18, 78 Sequential prestage, 24 VSAM, 22, 53 Working hypothesis, 43, 47, 50 Write promotions, 46 About the Author Bruce McNutt is a senior scientist/engineer working in the Storage Subsystems Division of International Business Machines Corporation. He has specialized in disk storage performance since joining IBM in 1983. Among the many papers which he has presented to the annual conference of the Computer Measurement Group, as an active par - ticipant for more than fifteen years, are two that received CMG “best paper” awards. The present book brings to - gether two threads which have run through his work: the hierarchical reuse model of data reference, first introduced in 1991 , and the multiple - workload approach to cache planning, first introduced in 1987. Mr. McNutt received his B.S. degree in mathematics from Stanford University, and his master’s degree in electrical engineering and computer science from the University of California at Berkeley. . paper” awards. The present book brings to - gether two threads which have run through his work: the hierarchical reuse model of data reference, first introduced in 1991 , and the multiple - workload. hierarchical reuse model, 33, 47, Operational conventions, 82 OS/390 Stage, 2, 55 THE FRACTAL STRUCTURE OF DATA REFERENCE Single - reference residency time, 8, 14, 36, 43, 52 - 53, 56, 63,. in disk storage performance since joining IBM in 1983. Among the many papers which he has presented to the annual conference of the Computer Measurement Group, as an active par - ticipant for