Understanding Nortel DMS-100 Capacity Administration |
Purpose
The purpose of this guide is to provide the Nortel DMS-100 family capacity administrator with descriptions of the capacities of traffic sensitive switch components and to suggest methods and procedures to observe how these capacities are being used by an inservice switch. Traffic sensitive components are those major parts of the switch that are susceptible to service degradation as the offered load is increased and approaches the engineered capacity level.
Monitoring capacity is an essential administrative function because it determines if the switch is operating under the conditions projected for the engineered period. Deviations from the projections may alter the end of design date (forecast date when additional resources will be needed) for the switch.
Note: Remote modules are not addressed in this article. Information on remote equipment may be found in the Operational Measurements Reference Manual.
Capacity Definitions
Typically, the capacities of the DMS-100 family switches are addressed in accordance with terms used by the design engineers. These terms, which reflect the different capacity concepts that are employed in the provisioning process, have been adopted for use in the day-to-day monitoring activities. These terms include:
Physical Capacity
Physical capacity is the total number of terminations that can be accommodated by a switch component or group of components, for example, the total number of terminations for lines in a Line Concentrating Module (LCM) or group of LCMs.
Traffic Capacity
Traffic capacity is the maximum number of terminations or requests for service that can be accommodated by a component or group of components while still meeting established delay and blocking service standards.
Real-Time Capacity
Real-time capacity, as applied to the DMS-100 family central processing unit (NT40) or DMS-Core (SuperNode), is the maximum number of call attempts that the Central Processing Unit (CPU) or DMS-Core can process while meeting the High-Day Busy Hour (HDBH) service objective of not exceeding 20 percent dial tone delay (delay greater than 3 seconds).
Memory Capacity
Memory can be a call limiting factor for a DMS-100 family switch and should be monitored to assure that there is sufficient memory at all times to meet the engineered call capacity of the CPU. Memory capacity administration is not addressed in this article, but information on the subject can be found in Memory Administration Guide and the Office Parameters Reference Manual.
Administration Functions
Administration of capacity includes monitoring capacity use and the effects that it may have on the DMS-100 family switch. To monitor the use, data is gathered through performance indicators such as Operational Measurements (OMs), logs, and capacity tools such as MEMCALC, and various Maintenance and Administration Position (MAP) status reports and counts.
Definitions of Administration Terms
Traffic Sensitive Switch Components
Traffic sensitive switch components are the specific components or resources that are susceptible to performance degradation. Performance degradation may occur when the traffic load on a component or resource approaches or exceeds its engineered limits, or when a component or resource failure occurs.
The traffic sensitive components of the DMS-100 family of switches discussed in this article are listed below (excluding those in the remote applications):
Busy Hours
The DMS-100 family switches are engineered based on empirical data or forecast data for the busiest hour for individual components or for the entire office (switch). These hours are referred to as busy hours. Listed below are the most commonly used busy hours and their definitions.
Call Busy Hour
The call busy hour is the time-consistent 60 minute period having the most call originating plus incoming (O + I) attempts per main station or Network Access Line (NAL). This hour is used primarily for the development of processor real-time capacities.
Usage Busy Hour (Office Busy Hour)
The usage busy hour is the time-consistent 60 minute period producing the most originating plus terminating (O + T) use per main station or NAL. This hour is used primarily for gathering data for load balancing and provisioning of switching hardware and software.
Service Busy Hour (Dial Tone Busy Hour)
The service busy hour is the time-consistent 60 minute period when the highest percentage of customers originating a call must wait more than 3 seconds for dial tone.
Component Busy Hour
The component busy hour is the time-consistent 60 minute period when call attempts or use are the highest for a particular switch component, for example, Digitone receivers, tones, and announcements. These hours may coincide, or each may be in a separate time period and administered separately.
Busy Hour Determination
Busy hours are derived from studies that are taken just prior to the office busy season. The busy season is defined as the three months (not necessarily consecutive) that have the highest average business day traffic during the office busy hour.
Busy hour determination studies are usually conducted for 5-15 days between the hours of 8 a.m. and 11 p.m. The busy hour that is determined is then used during the following busy season.
Busy hour studies may be conducted on a manual basis or through the use of a mechanized system. By whatever means, the studies should select the time periods that provide call data that can be used to engineer the switch most effectively and measure the level of service being rendered to the subscribers.
The characteristics of the office determine the periods of the day to be studied. In some offices, the calling patterns do not change significantly from one busy season to another. For those offices, a five day study is sufficient to verify that the hour has not changed. The hours chosen should be the known busy hour and the hours on either side of that hour. In other offices, several hours may carry loads of approximately the same level, so a longer study period (10 to 15 days) and the full range of hours (8 a.m. to 11 p.m.) should be considered. All data should be collected at least on a half-hourly basis.
The criteria for changing the designated busy hour from one time period to another is determined by the operating company.
Grade of Service
The basic design philosophy of the DMS-100 family is based on delay criteria from the peripheral originator (line or trunk) up to the network. The network and all terminating paths are designed based upon blockage criteria. Blockage is defined as the failure to find an idle channel and is referred to as matching loss. The rates of delay and blockage are referred to as the Grade of Service (GOS). The higher these rates become, the lower the GOS that is experienced by the subscriber.
Delay in the DMS-100 family occurs in the form of Dial Tone Delay (DTD) for originating calls and Incoming Start to Dial Delay (ISDD) for incoming calls. The percentage of delays greater than 3 seconds is used to assess the GOS for the overall switch design.
The criteria are 1.5 percent DTD and ISDD greater than 3 seconds for the Average Busy Season Busy Hour (ABSBH) and 20 percent for the High-Day Busy Hour (HDBH).
Service Criteria
Service criteria are those objective levels of call blocking and delay that are set for the measured busy hour. Effective capacity administration will ensure that these service objectives are met. Service criteria have been developed on the basis of judgment and experience. The overall objective is to provide the best possible service at a reasonable cost.
To establish a service standard, it is necessary to have a measurement that quantifies the inconvenience a customer experienced because of call blocking or call delay. When a call is blocked, a tone or message is delivered to the customer who then must hang up to try the call again. When a call is delayed, the customer is only considered to be inconvenienced if the delay exceeds some maximum tolerable value. The DMS-100 family design applies a mixture of loss and delay criteria.
Loss Criteria
All the line modules are engineered to meet objective service levels during the worst case of Incoming Matching Loss (IML) during either the ABSBH or HDBH. Incoming matching loss is defined as that condition when a call cannot be completed because an idle path cannot be found between an incoming trunk and an idle line. Nortel engineering tables are based on IML objectives.
The existing published matching loss criteria are stated for the entire office. They have two sources, peripheral matching loss and network matching loss. The peripheral portion is the predominant part of the HDBH criteria. The recommended incoming matching loss criteria for a DMS-100 are shown below:
---------------------------------------------------------------------------------------------- Nortel Recommended Matching Loss Criteria Busy Hour Overall Peripheral Network ---------------------------------------------------------------------------------------------- Average Busy Season Busy Hour 2.0% 1.9% 0.1% (ABSBH) High-Day Busy Hour 5.0% 4.0% 1.0% (HDBH) ----------------------------------------------------------------------------------------------
Delay Criteria
When subscribers and calls are served on a delay basis, the concern is usually more with the duration of the delay than the probability of delay. At the present time, delays of less than 3 seconds are considered acceptable to the subscriber, or at least they do not annoy the subscriber if they do not happen too frequently. The delay criteria that are used for engineering purposes are as follows:
---------------------------------------------------------------------------------------------- Delay Criteria Delay Criteria Description ---------------------------------------------------------------------------------------------- Dial Tone Delay The probability that a customer will experience a dial tone delay of (DTD) more than 3 seconds Incoming The probability that an incoming trunk to a multifrequency receiver Start to Dial Delay will experience a delay of more than 3 seconds before the receiver (ISDD) becomes available. ----------------------------------------------------------------------------------------------
The current recommended delay criteria are shown below:
----------------------------------------------------------------------------------------------
Recommended Engineering Delay Criteria
DMS-100 DMS-250
Delay Measurement ABSBH HDBH 10HDBH HDBH
----------------------------------------------------------------------------------------------
DTD 1.5% 20.0% See Note See Note
ISDD 1.5% 20.0% 8.0% 20.0%
----------------------------------------------------------------------------------------------
Note: Not applicable to this office type.
----------------------------------------------------------------------------------------------
With configurations that require a high penetration of Meridian Digital Centrex (MDC) or Multiple Appearance Directory Number (MADN) features, a line peripheral can become limited by high-day busy hour attempts. The load service relationship for an attempt limited line peripheral is dial tone delay. The attempt capacity can be obtained by using the Nortel PRTCALC tool. Staying within this attempt limit maximizes throughput, minimizes any delay caused by the peripheral, and supports an overall DTD of 20 percent during the high-day busy hour.
The traffic capacity tables, associated with line peripherals and the PRTCALC program, assume an even (balanced) flow of traffic across all line modules. This PRTCALC function is usually performed by the traffic engineer. The administrator may get the required information from the engineer that is responsible for the office in question.
Measurement Methods
The following section describes methods for measuring the capacity in a DMS-100 family switch. These methods are based on the measurements that are currently available in the data collection system.
Performance Indicators
Performance indicators are measurements or records of events that occur during a given period of time or in a time sequence. For the DMS-100 family switch, performance indicators take the form of Operational Measurements (OMs) and log reports. In addition, a method to measure and control the balance of traffic load offered to like components of the product or system is employed. This measurement method is developed by the operating company and uses the standard operational measurements provided by the DMS-100 family switch.
Operational Measurements
The administration of capacities in the DMS-100 switch makes use of the switch's data collection system. This system collects groups of data designated OMs. Operational measurements are derived by monitoring certain events in the switch and entering the results into registers in the data store. Each register has a unique name. The registers are scored individually each time an event occurs, or when the state of an item is scanned (sampled) at regular intervals regardless of the time of the occurrence of an event. Scan rates are either 100 seconds or 10 seconds.
Single events, measured individually, are referred to as peg counts. Sampled measurements are used to determine the degree of use of DMS-100 hardware and software resources and are referred to as usage counts.
Because each register can record either a single event or a group of similar events, the registers are provided on an office basis, or a unit basis. For example:
The peg counts and usage counts are stored in active registers that are updated whenever new data are entered. The OM data in the active registers are useful only if related to the specific period of time of collection. Therefore, OM data cannot be copied directly from the active registers because of the probability that additional counts may occur during the copying process that would result in an inaccurate data output.
To prevent inaccurate data, two complete sets of registers are provided. During any collection period, one set is used to collect current data and is known as the active class. The other set, known as the holding class, contains the data collected in the previous collection period and is used to provide data to reports or to the various accumulating classes.
At the end of the collection period, data in the active registers are transferred to the holding registers and the active registers are zeroed. This transfer of data from active to holding occurs at the same time for all counts. Operating company defined accumulating registers are used to accumulate data over longer periods of time than the basic period (a day or week). The data accumulation process adds the contents of the holding class registers to the accumulating class registers just prior to the next data collection period. The accumulated data are available to the end of the accumulating period. At the end of the accumulating period, the registers are unloaded to a printer or other recording device and the registers are zeroed.
The control of the length of the basic time periods is in the table designated as OFCENG. The office parameter OMXFR in OFCENG defines the timing value OMXFERPERIOD. This value is set at either 15 or 30 minutes.
Whenever an active register count exceeds its 65,536 limit, an extension register needs to be assigned or the data will be understated. The extension register will peg once each time the limit is exceeded. The count on the regular register is added to the product of the extension register count multiplied by 65,536, for example:
---------------------------------------------------------------------------------------------- Regular Register Count = 236 Extension Register Count = 2 236 + (2 x 65,536) = 131,308 (true total for this register) ----------------------------------------------------------------------------------------------
The OMDUMP command (input at the MAP) may be used to determine which registers have been assigned extension registers. The command is as follows:
>OMDUMP CLASS (class name) FORMAT
The following figure shows an example of a portion of a printout containing a register value and its extension register value:
----------------------------------------------------------------------------------------------
Example of a Register and its Extension Register
----------------------------------------------------------------------------------------------
INOUT2 INTONE NIN
OUTMFL OUTRMFL OUTOSF
ORIGANN ORIGKT ORIGOUT
ORIGTONE NORIG NORIG2
TRMNWAT2 TRMMFL TRMBLK
0 111 31642
0 0 101
1993 10 32146
1480 11205[1] 1[2]
0 1 0
[1] NORIG Register
[2] NORIG Extension Register
----------------------------------------------------------------------------------------------
If an accumulating register is expected to exceed the register limit, then it should be assigned to double-precision. This assignment raises the limit to 4,294,967,296 counts (65,536 x 65,536) with a printout limit of 8 characters. Double-precision uses two registers as previously described. When changing a class precision from single precision to double-precision, all OM groups must first be deleted from the class. Refer to the Basic Administration Procedures, under command OMACCGRP, for detailed procedures.
The output from the OMs may be sent to a local printer or collected on a mechanized system, for example, the Engineering and Administrative Data Acquisition System (EADAS).
----------------------------------------------------------------------------------------------
Example of a Double-Precision Register
----------------------------------------------------------------------------------------------
INOUT2 INTONE NIN
OUTMFL OUTRMFL OUTOSF
ORIGANN ORIGKT ORIGOUT
ORIGTONE NORIG NORIG2
TRMNWAT2 TRMMFL TRMBLK
0 375 91141
0 0 13
1993 10 150585[2]
1480 325569[1] 0
0 1 0
[1] NORIG Double-Precision Register
[2] ORIGOUT Double-Precision Register
----------------------------------------------------------------------------------------------
Log Reports
A log report is a message from the DMS-100 whenever a significant event has occurred in the switch or one of its peripherals. Log reports include status and activity reports as well as reports on hardware or software faults, test results, and other events or conditions likely to affect the performance of the switch. A log report may be generated in response to a system or manual action. Complete descriptions of all log reports are contained in the Log Report Reference Manual.
Subscriber Trouble Reports
Subscriber trouble reports are another source for monitoring the capacity of switch components. These reports can often point to off-busy hour capacity problems that otherwise may go undetected.
Capacity Factors
Capacity factors are those events that affect the capacity of a hardware or software component of the switch. The status of the capacity of switch components is measured by capacity indicators such as operational measurements and log reports. Capacity factors include such items as:
Automated Tools
Several automated tools are available to the administrator that will aid in the monitoring of capacity. Nortel developed these tools to assist in the initial provisioning of an office and for use in the ongoing surveillance of a working switch.
REAL::TIME
REAL::TIME is a PC program designed to provide an estimate of the DMS-100 family CPU real-time requirements. The DMS-100 switch provides distributed processing over many switching entities. The call attempt capacity of each of these switching entities must be predicted to establish operating guidelines. These guidelines are used to determine the loading levels for specific applications, including residential services. The real-time can be predicted by using the anticipated call mix and timing per call.
Using traffic criteria along with detailed office provisioning data, REAL::TIME generates an estimated occupancy for the central processor. REAL::TIME can be used in the following office configurations:
REAL::QUICK
REAL::QUICK is an abbreviated form of REAL::TIME. Some assumptions and considerations are applicable to each processor. If there is a significant variance from these assumptions and considerations, a more detailed study should be performed using PRTCALC or REAL::TIME.
PRTCALC
PRTCALC is a PC program designed to provide an estimate of DMS-100 family peripheral real-time requirements. PRTCALC can be used to calculate the real-time call attempt capacity for peripheral modules. PRTCALC is composed of three sections:
Input for PRTCALC comes either from projected (forecast) data based on current operational measurement trends, or from inputs to the NT-ACCESS tool.