CSC470 Fall, 2000 Computer System Modelling And Analysis Laboratory #4 Oct. 31st/Nov. 2nd and Nov. 7th/9th, 2000 Verification of Authorship: By the signatures below, we certify that we are thoroughly familiar with the contents of this laboratory report: it is our own work, we have referenced all our sources of information, and we are the sole authors. Team Members: Name Signature Date (1). ________________________ _________ (2). ________________________ _________ (3). ________________________ _________ Assignment Starting with the baseline model you validated in the prepara- tion, use analytical modelling with MAP to answer the questions below. VALIDATED BASELINE MODEL: ***** insert "list" of baseline model ***** ***** insert performance report of baseline ***** 1. Current Workload (a). Assume that the service level objectives (SLO's) with the current workload are: Xdsn > .007 Racc < 30.0 Rmnt < 50.0 Evaluate and compare the three different ways suggested below for modifying the system so the service level objectives are satisfied. (i). Adjust the avg_mpl of the DSN class. ***** insert "list" of model with adjusted MPL ***** ***** insert performance report of model ***** (ii). Modify the priorities of the various classes at the cpu, in addition to adjusting the avgmpl. ***** insert "list" of model with prior. and mpl adjusted ***** ***** insert performance report of model ***** (iii). Upgrade the SPARK cpu to a different model. ***** insert "list" of model with upgraded cpu ***** ***** insert performance report of model ***** Given the results of the experiments indicated above, make a recommendation for the most cost-effective change to make to the system in order to make sure that the SLO's of all three classes are satisfied. ***** insert brief answer referring to models above ***** (b). STARTING AGAIN FROM THE VALIDATED MODEL of the base current system, assume now a different set of SLO's: Xdsn > .004 Racc < 30.0 Rmnt < 30.0 Evaluate and compare the three different ways suggested below for modifying the system so the service level objectives are satisfied. (i). Add memory to the system (32MB per domain slot), and adjust the capacity of the memory domain (limited_mpl). ***** insert "list" of model with larger domain ***** ***** insert performance report for the model ***** (ii). Starting again with the baseline system, upgrade one or more of the BRONZE disks to different model(s). ***** insert "list" of model with upgraded disk(s) ***** ***** insert performance report for the model ***** (iii). Starting again with the baseline system, adjust the class priority levels at the cpu. ***** insert "list" of model with adjusted priorities ***** ***** insert performance report for the model ***** Given the results of the experiments indicated above, and using your knowledge and intuition about how expensive each change would be, make a recommendation for the most cost-effective way of modifying the system so that the SLO's are all satisfied. ***** insert brief answer referring to models above ***** 2. Capacity Planning Problem for a Future Workload Start again with the baseline model that you validated in the preparation. Given assumptions about how the workload is expected to evolve in the future, make recommendations for specific cost- effective system upgrades to support the anticipated workload satisfying the performance level objectives specified for the sys- tem. The cost of upgrades are measured in Computer Engineering Monetary Units (or CEMU's). Future Workload + The transaction rate for accounting will increases to .18 + The active users in the maintenance class will become 15 + The throughput required for design class will become .008 System Service Level Objectives class desirable acceptable design Xdsn > .010 Xdsn > .008 accounting Racc < 12.0 Racc < 15.0 maintenance Rmnt < 25.0 Rmnt < 30.0 Alternative System Upgrades Here are a number of possible upgrades that can be applied to the system, along with their associated costs in CEMU's: 1. Change the scheduling priorities of the classes at the processor [0 CEMU's]. 2. Change the admission priorities of the classes to the memory domain [0 CEMU's]. 3. Change the batch activation level (i.e., avgmpl) for class design [0 CEMU's]. 4. Rearrange the files of one class among the available disks [2 CEMU's per class]. 5. Upgrade the processor from a SPARK to a FLAME [10 CEMU's], BLAZE [30 CEMU's], TWIN [20 CEMU's], or QUAD [40 CEMU's]. 6. Upgrade the disks from BRONZE to SILVER [4 CEMU's per disk] or GOLD [8 CEMU's per disk]. 7. Add additional disks (of some model) and shift some frac- tion of the files to them [where CEMU's per disk is 4, 8 and 12 for BRONZE, SILVER, and GOLD models, respec- tively]. 8. Add additional memory in order to allow the domain capa- city (limited_mpl) to be increased. Assume that each additional 32 MB allows an increase of one in the limited_mpl, and costs 5 CEMU's. RECOMMENDATIONS FOR UPGRADES: (i) Specify a cost-effective system in which all SLO's are met at at least the "acceptable" level. ***** summary of upgrades applied ***** ***** insert "list" of model including the upgrades ***** ***** insert performance report of the model ***** (i1) Specify a cost-effective system in which all SLO's are met, or at least very nearly met, at the "desirable" level. ***** summary of upgrades applied ***** ***** insert "list" of model including the upgrades ***** ***** insert performance report of the model ***** +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ For two or three of the MAP Execs you used in solving this Lab, indicate the pupose of the MAP Exec, and provide a listing of it. ***** purpose of MAP Exec #1 ***** ***** list of MAP Exec #1 ***** ***** purpose of MAP Exec #2 ***** ***** list of MAP Exec #2 ***** ***** purpose of MAP Exec #3 ***** ***** list of MAP Exec #3 ***** ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++