Actual or real workload
The actual or real workload is the work actually performed by the computer system. Ideally, the actual workload is identical to the business workload.
Adapter
An adapter is a hardware interface between a device controller and a system backplane or bus.
Analysis report
An analysis report is a report that identifies the effects of system parameter settings, hardware configurations, and workload mixes on the performance of any cluster node or the entire cluster system. This report provides conclusions with supporting evidence and makes recommendations.
Analysis summary
The analysis summary is a short summary appearing in the analysis report after each node analysis. The summary contains the following information:
- Number of Performance Manager records analyzed for the specific reporting period
- Number of Performance Manager records that satisfied any rules
- Number of Performance Manager records that did not satisfy any rules
- Number of conclusions generated for the node being analyzed
Archive
Archive is the act of compressing Performance Manager daily data records into history files, which can be used in place of daily data to create reports, models, and graphs.
Auxiliary knowledge base
Auxiliary knowledge base is A collection of user-defined site-specific rules and thresholds that have been compiled with the rules compiler (name.KB) and used to augment the factory supplied rules.
Auxiliary rules
Auxiliary rules are the source files containing user-defined site-specific rules and thresholds that collectively comprises the auxiliary knowledge base.
Baseline load
A Baseline load is the Performance Manager's measurements of your existing system or cluster System. This data is stored in a model input file (.MDL).
Baseline model
A baseline model is a model generated from historic or daily Performance Manager data. The model output provides a workload characterization report. An unmodified .MDL file, the baseline load, represents Performance Manager measurements of your existing system. The model generated from this file is known as the baseline model.
Buffered I/O operation
A buffered I/O operation occurs each time an intermediate system buffer is used in place of the process context buffer.
Business workload
A business workload is the work the business expects its computer to perform.
Computer system
The computer system is all of the computer hardware on which business work is performed.
Conclusions
Conclusions are Text displayed in an analysis report.
CPU branch explicit
In a modeling context, the CPU branch explicit is the probable distribution of load across CPUs for a workload or transaction class. Probabilities must sum to one. This data is included in the model input file.
Custom graph
A custom graph is a graph type which allows you to specify which Performance Manager data items to graph.
Daily data files
Daily data files are created by the Performance Manager CPD data collector, one for each node in the cluster, each day. The filename has the following format:
PSDC$DATABASE:PSDC$nodename_yyyymmmdd.CPD
Data cell
A data cell is the basic unit of data used to create analysis reports. This data is either retrieved directly from a field in a subrecord of a daily data record or derived from it. The data cell is typically used as a variable in a rule expression.
Data collection error log
The data collection error log is an ASCII file common to the entire cluster system called PSDC$DATABASE:PSDC$DC.LOG. Errors that occur during data collection are recorded in this file.
Data collection schedule
The data collection schedule is a user-defined schedule by which the Performance Agent determines when to record data and what data to record for each node in the cluster.
Data collection synchronization
Data collection synchronization is a method used to correlate intervals measured on different nodes within a cluster system; those intervals must represent the same real time to make analysis effective.
Data files
Data files are the files containing performance data from which reports and graphs are constructed.
Data record
The data record contains performance data written by the data collector. The CPD data collector writes one data record for each two-minute interval.
Database directory
The database directory is a directory located on a permanently mounted disk, accessible to every node in the cluster. The data collection process writes the daily data files to this area. Performance Manager software references this area via the system wide logical name PSDC$DATABASE.
Dates file
The dates file contains a list of dates used to select data.
Device
A device is a piece of hardware in the computer system. It performs measurable units of work.
Direct I/O
The number of direct I/O operations performed per second. This illustration is tallied at the $QIO application interface layer.
Disk branch by source
In a modeling context, disk branch by source is the probable distribution of load across disks by originating CPU for a workload or a transaction class. Probabilities must sum to one. This data is included in the model input file.
Disk I/O
Disk I/O is the number of I/O operations per second for the device. This illustration is tallied at the physical device driver layer.
Dump report
A dump report contains formatted output of data fields for each record of a Performance Manager daily data file or history file.
Evidence
Evidence supports lines of performance data displayed in an analysis report.
Factory rules
Factory rules are the performance rules supplied with the Performance Manager in the PSPA$EXAMPLES area in the file named PSPA$KB.VPR.
Family name
Family name is an identifier for a group of workload definitions, also known as a workload (transaction or usergroup) family.
Family type
There are two family types
- Usergroup families contain workload definitions based on user criteria.
- Transaction families contain workload definitions based on image and process data.
Specify either family with the /CLASSIFY_BY qualifier to control use of the family for the reporting facility.
File type
A file type is specified by one of the following extensions in its name:
- .COM Various command files
- .CPD Cluster Performance Data file
- .DAT Parameters and schedule file
- .EXE Various image files
- .HLB Parameter Editor help file
- .KB Compiled rules Knowledge Base file
- .LIB Model library file
- .LIS A report file
- .LOG Data collection process error log file
- .MAR Sample macro application file
- .MDL Model file
- .name History file, name is history file descriptor; also, alternate data file name
- .name_JOU History journal file
- .REG A ReGIS graph written to a disk file
- .TXT Holidays and message files
- .VPR Performance Rules source file
Granularity
Granularity is a Performance Manager parameter file element that specifies for each history file descriptor how often a history file is created.
Hard page fault
A hard page fault occurs each time a process references a virtual page that is not in its working set and requires a read operation from disk, a hard page fault is generated.
Histogram
A histogram is a (ASCII) graphic chronological chart showing resource use.
History database
The history database is reduced data from the daily data files, which resides in the history files.
History file
The history file contains data archived from the daily data files. The number of history files created and maintained depends upon the number of history file descriptors and the associated granularity. The filename has the form PSDC$DATABASE:PSDC$nodename_dd-mmm-yyyy_dd-mmm-yyyy.name.
History file descriptor
The history file descriptor contains the description that Performance Manager uses to determine how to archive data to the file.
Holidays file
The holidays file contains a list of holiday dates. The Performance Agent uses the holiday schedule on these dates. The file name is PSDC$HOLIDAYS.TXT.
Hot files
Hot files are the most frequently accessed files on each disk. The Performance Agent collects hot file data when the length of a disk queue exceeds the HOTFILE_QUEUE setting.
Interaction
An interaction with a device is a two-step process. First, a unit of work to be done (job, user, process, and so forth) enters the queue of the device. Then, the unit of work is serviced by the device (in a manner dependent on the queueing discipline) and departs the device. The concept may be generalized to the system as a whole. The set of device interactions required to process each unit of work is called a transaction.
Interval
In the context of a history file, interval is an ADVISE EDIT ADD/HISTORY qualifier and a Performance Manager parameter file element.
In the context of a daily CPD data file, the interval for writing records is fixed at two minutes.
In the context of real-time displays, the interval is user-defined (default interval=10 seconds, minimum=1 second).
Journal file
The journal file is when Performance Agent software creates one history journal file for each history file. These are used by the update process in conjunction with the daily data files to recreate corrupt or deleted history files. The filename has the form PSDC$DATABASE:PSDC$nodename_dd-mmm-yyyy_dd-mmm-yyyy.name_JOU. Do not delete these files. If they exist they are needed.
Knowledge base
The knowledge base is a file consisting of Performance Manager performance rules used to analyze daily or historic Performance Manager data. It may be augmented using an auxiliary knowledge base.
Measured workload
The measured workload is the workload that can be observed on the system. Ideally, the measured workload is identical to the actual workload, but specified in different terms.
The measured workload is specified by the resource demands it places on the actual system. These demands, or loadings, are given as the service times of the users at each device along with the number of expected transactions at each device.
Model
A model is an abstraction of a system focusing on high-level performance characteristics.
MODEL_TRANSACTIONS
Model_transactions are a default workload family defined in the parameters file. It can be used to characterize workloads in the transaction class for modeling.
MODEL_USERGROUPS
Model_usergroups is a default workload family defined in the parameters file. It can be used to characterize workloads in the user group class for modeling.
Modeling
Modeling is the process of gathering, organizing, and evaluating principal components of a system and the ways in which they interact for the purposes of understanding and predicting system behavior.
MSCP
MSCP-Mass Storage Control Protocol. A software protocol used to communicate between a VAX or Alpha processor and a disk controller such as an HSC.
OpenVMS Cluster
An OpenVMS Cluster is a highly integrated organization of AlphaServer and HP Integrity server system–or VAX and HP AlphaServer system–applications, operating systems, and storage devices.
OTHER
All workload families have the catch-all workload OTHER to absorb process data that does not match the selection criteria of any defined workload.
Parameters file
PSDC$PARAMS.DAT is the parameters file resides in the PSDC$DATABASE area and serves as a repository for workload characterizations, history file descriptors and other Performance Manager parameters.
Performance Agent
The Performance Agent is a detached process that collects and records performance data for specified nodes in the cluster system according to a weekly schedule.
Performance evaluation report
The performance evaluation report is a statistical report that helps you determine whether changes that you implemented (based on recommendations in the Analysis Report) improved or degraded system performance.
Periodicity
Periodicity is a parameter file element for a history file descriptor that specifies how often the averaging cycle is restarted.
Predefined graph
A predefined graph is a graph in which the Performance Manager defines the metrics plotted.
Queue length
Queue length is the average number of outstanding requests, either waiting for or receiving service.
Queueing network model
Queueing network model is a mathematical abstraction of a system where the computer system is represented as a network of queues. Each queue in the network is evaluated analytically.
Raw data file
Recommendation
A recommendation is the text presented in an analysis report that offers system tuning advice based on rules firing. See Rule conclusion.
Residence time
Residence time is the time, in seconds, between image activation and image termination. Average time that a request spends while waiting for and receiving service.
Response time
Response time is the elapsed time between the arrival of a request and the moment of completion. In the context of modeling, the interval between the moment a request arrives at a device and request completion at a device.
Rule
A rule is one or more rule conditions that are evaluated when Performance Manager Analysis reports are generated. Rules are applied to daily or historic data. If all conditions for a rule are true then there is a rule occurrence. Rules are defined to expose areas of potential system problems. See Rule firing.
Rule conclusion
A rule conclusion is a rule element. The conclusions are Performance Manager recommendations based on the conclusion text element of a rules file rule construct.
Rule condition
A rule element. A rule condition is made up of one or more rule expressions. Describes the circumstances that must be true to cause a rule occurrence.
Rule elements
The seven rule elements that can exist in a rule construct are as follows:
Brief Conclusion element
Conclusion element
Domain element
Evidence element
Occurrence element
Rule condition element
Rule ID element
Rule evidence
A rule element. Data satisfying a rule occurrence in a Performance Manager analysis report. The evidence consists of data cell names and values. Typically these data cells are some of those contained in the rule expressions.
Rule expressions
Components of rule conditions which may include the following:
Decimal values
Literal symbols
Tally data cells
Numeric data cells
Boolean data cells
Scan routine data cells
String operators
Numeric binary operators
Parentheses for precedence
Rule firing
After all the data has been processed, when creating a report, the Performance Manager examines the number of rule occurrences for each rule. If the rule occurrence threshold is met for a particular rule, the rule is said to fire. For each rule that fires, an entry is made in the Analysis report. The entry may include evidence and conclusions.
Rule identifier
A rule element. A five-character alphanumeric code enclosed in braces, for example, {M0010} which uniquely identifies a rule. (A zero for the second character is reserved for Digital use only.)
Rule occurrence
Each time all the rule conditions for a given rule are true, there is one rule occurrence. See Rule firing.
Rules compiler
The rules compiler generates a binary knowledge base file (name.KB)from an ASCII rules source file (name.VPR).
Rules file
The rules file is a compiled knowledge base file (name.KB). A Performance Manager rules source file (name.VPR).
Rules file constructs
The following five constructs can exist in a Rules File:
- Comment construct
- Disable construct
- Literal construct
- Rule construct
- Threshold construct
Saturation
Saturation is the point at which response time at a device becomes substantially higher than the service time.
Schedule file
The schedule file is a file, PSDC$SCHEDULE.DAT, that resides in the PSDC$DATABASE area and controls when Performance Manager daily data is recorded.
Shadow set
A shadow set is one or more compatible physical disk volumes connected together for volume shadowing and represented by a virtual unit. Thus, the term shadow set refers to the physical unit and the virtual unit.
Soft page fault
Each time a process references a virtual page that is not in its working set but is in memory, a soft page fault is generated.
Split I/O
Split I/O is the number of additional physical disk I/O operations required to complete a single user's I/O request, which could not be satisfied in a single I/O to a device.
Transaction
A transaction is a quantifiable unit of work that typically delineates a single processing step in computer systems.
Transaction class
A transaction class is a group of related transactions. They may be related by the function they perform, by the users who initiate them, or by other quantities you define. Transactions may also be determined by the system resource demands. The Performance Manager software generally refers to a transaction class as a workload.
Transaction class workload
A transaction class workload is a workload that contains data bucketed by a workload definition defined in terms of images.
Transaction workload family
A transaction workload family is a set of image based workload definitions.
User defined graph
Usergroup workload family
A usergroup workload family is a set of user based workload definitions.
Utilization
The percentage of a resource's capacity being used.
VAXcluster system
A multipurpose system configured by interconnecting or clustering VAX processors and storage controllers to provide increased capabilities for sharing data, distributing workloads, and providing greater system and data availability.
VUP
VAX unit of processing speed. The VUP rating measures the CPU power of a system compared to a VAX 11/780. A VAX 11/780 has a VUP rating of 1.0.
Wait time
Wait time, or queueing time, is the average time each request spends waiting in a queue for service. During this time, the request accomplishes no useful work. Wait time may be derived specifically for each device in the system or for the system as a whole.
Workload
A representation of the actual system's resource demands. Performance Manager software reports graphs and capacity plans based on workloads. Workload definitions specify to Performance Manager software how to organize the system load into workloads.
Workload definition
A workload definition can be one or more users, image names, or processes that represent units of work on the system. Workload definitions are identified and stored in the Performance Manager parameters file. The goal is to express the system's total workload in manageable and meaningful units against which Performance Manager can report. Use ADVISE EDIT to create, modify and delete workload definitions.
A Transaction workload definition contains at least one image name. Typically this workload will contain images with similar resource demands.
A Usergroup workload definition contains a user specifier (such as accountname, username, processname, UIC code). Typically this workload will contain a group of users who belong to the same business unit, such as a department.
Transaction and Usergroup are terms applied to a workload by the /CLASSIFY_BY qualifier.
Workload family
A workload family is a collection of workload definitions collectively representing the entire work on a computer system.
Workload family name
A workload family name is a name that identifies a group of workloads that collectively constitute a unit called a family.
Workload name
A workload name is a name that identifies a workload or workload definition.
Workload selection criteria
A workload selection criteria is the criteria by which process data is assigned to a workload. A workload is selected when process data matches on either or both of a user specifier or image name.
|
Copyright © 2008 CA.
All rights reserved.
|
|