"... In PAGE 42: ...Table4 : Comparison of the 13 Questions and Methods Oriented Evaluation Approaches on Main EVALUATION METHODS Evaluation Approaches (by identification number)* Evaluation Methods 3456789101 12131415 Operational objectives a19a19 Criterion-referenced tests a19a19 a19a19 Performance contracting a19 Program Planning a19a19 amp; Budgeting System Program Evaluation a19 amp; Review Technique Management by objectives a19a19 a19 Staff progress reports a19 Financial reports amp; audits a19 Zero Based Budgeting a19 Cost analysis, cost- a19 effectiveness analysis, amp; benefit-cost analysis Mandated program a19 drivers amp; indicators Input, process, output a19a19 databases Independent goal a19a19 achievement auditors Procedural compliance a19 audits Peer review a19 Merit pay for individuals a19 and/or organizations Collective bargaining a19 agreements Trial proceedings a19 Mandated testing a19a19 Institutional report cards a19 Self-studies a19 Site visits by experts a19 Program audits a19 Standardized testing a19a19 a19 a19 Performance measures a19a19a19 Computerized or other a19a19 a19 database Hierarchical mixed model a19 analysis Policy analysis a19 Experimental amp; quasi- a19a19... ..."

"... In PAGE 4: ... The sources vary as to the extent of formalization. The set of usability-related activities proposed in the HCI field are detailed in Table1 , where sources follow an order of increasing formalization from left to right. We have analyzed the activities proposed by the different authors in order to extract the common ones or, at least, the activities that are at the same abstraction level and are common to several sources.... In PAGE 5: ...The resulting usability activities (the left column in Table1 ) are represented in Fig. 1, grouped according to the generic kind of activity to which they belong: Analy- sis, design or evaluation.... ..."

"... In PAGE 22: ... Fortunately, this number is far more pessimistic than the behavior typical programs exhibit. Table2 shows that for our benchmarks, the total number of lock records allocated is very small, and pales in comparison with the total number of objects allo- cated. Moreover, at 24 bytes per lock record, even the worst program seen, the volano server, con- sumes 77 Kbytes for lock records, a trivial amount compared with the several Mbytes used for objects and thread stacks.... ..."

"... In PAGE 22: ... Fortunately, this number is far more pessimistic than the behavior typical programs exhibit. Table2 shows that for our benchmarks, the total number of lock records allocated is very small, and pales in comparison with the total number of objects allo- cated. Moreover, at 24 bytes per lock record, even the worst program seen, the volano server, con- sumes 77 Kbytes for lock records, a trivial amount compared with the several Mbytes used for objects and thread stacks.... ..."

"... In PAGE 19: ... Table2 : Classification based on synchronization issues. 3.... ..."

"... In PAGE 13: ... In this section, we look at the characteristics of the objects in the various workloads to better understand what is required of the un- derlying filesystem, whether it is provided by the operating system or the DBMS. The total number of objects and the fraction of them that are referenced or modified are presented in Table10 . The total object count was obtained from the catalog dumps that were taken when the systems were traced.... In PAGE 15: ... Interestingly, the distribution of file sizes in PCs running Windows in an office enivironment has also been recently reported to follow a lognormal distri- bution but the files are again much smaller than the objects in the database workloads [10]. Notice from Table10 that, for most of the workloads, less than half of the objects are referenced for the duration of the trace. In general, a common approach to improving computer system performance is to place the items that are likely to be used in faster storage.... ..."

"... In PAGE 13: ... In this section, we look at the characteristics of the objects in the various workloads to better understand what is required of the un- derlying filesystem, whether it is provided by the operating system or the DBMS. The total number of objects and the fraction of them that are referenced or modified are presented in Table10 . The total object count was obtained from the catalog dumps that were taken when the systems were traced.... In PAGE 15: ... Interestingly, the distribution of file sizes in PCs running Windows in an office enivironment has also been recently reported to follow a lognormal distri- bution but the files are again much smaller than the objects in the database workloads [10]. Notice from Table10 that, for most of the workloads, less than half of the objects are referenced for the duration of the trace. In general, a common approach to improving computer system performance is to place the items that are likely to be used in faster storage.... ..."