PDF of this tutorialThe critical requirement for wide deployment of ATM networking equipment is the ability to dynamically establish switched virtual circuits (SVCs). It is impossible to support the growing number of users with the current method of cross-connecting using permanent virtual circuits (PVCs).
The function of establishing SVCs across an ATM network is called signaling. To fully test the signaling features utilized in an ATM switch, both the signaling equipment and the network over which the equipment will operate should be tested together (see Figure 11).
Figure 11. Signaling Functional Block Overview
(Click on image for full-size version.)
The signaling function of an ATM switch includes establishing, maintaining, and clearing VCs. These switch operations must conform to the industry standards for network signaling. When the signaling equipment is part of a network, the signaling equipment and the network over which the equipment will operate must both be certified to the same operating specifications.
Conformance testing involves applying a series of predefined tests to the equipment based on the matching specification. These tests are generally applied under ideal conditions and are intended to demonstrate conformance (of supported function) to the detailed specification. Conformance tests usually verify only the base (minimum) set of functions for the equipment and do not include any dynamic stress test or error recovery or demonstrate interoperability of the equipment.
Interoperability
Interoperability testing verifies compatibility between two or more products intended to function in a network environment. With many differences in features and supported functions in vendor's products, it is difficult to develop general specifications for interoperability testing.
Interoperability testing procedures are designed to exercise a specific set of signaling functions across a network of two or more switches. Two key areas to be tested are QoS–based routing and the scalability (i.e., reachability information). Interoperability testing should be performed for all signaling protocols to be used with the equipment.
The private network-network interface (PNNI) signaling protocol provides a particular challenge for interoperability testing. The routing components, adopted from existing packet routing protocols (i.e., used on the Internet) are the most complex routing protocols in current use. Specifically, portions of the PNNI signaling protocol using the ATM routing and signaling combinations should be tested.
Signaling Performance Testing
When assessing ATM switch performance, or the ability of a network to transfer cells, most switch vendors and network designers focus on the cell relay component (i.e., switch fabric and cell transmission rate). Most ATM interworking protocols rely implicitly on the use of signaling to establish SVCs that will be used to transfer data between users. Signaling performance is a measurement of the efficiency of this function.
The ability to transfer data through an ATM network is restricted by the following:
- the rate at which SVC connections can be set up through an ATM switch
- the aggregate latency for an SVC connection setup request to be propagated through a network
In many ATM transactions, the amount of time to establish the SVC connection may be many times greater than the time to transfer the data. It is important that signaling performance be measured to establish operational benchmarks.
The methods for measuring signaling performance are influenced by the architectural design of the ATM switch. Switch design issues that effect signaling protocol performance include the following:
- processor type
- depth of memory to buffer signaling messages
- sharing processor(s) between multiple ports
- structured design of the signaling software
Performance Testing is Necessary
The operating characteristics and the function of an ATM switch during signaling operations often varies widely for different manufacturers and models. These variations in performance can be the result of choices in the following:
- architecture (hardware)—the operating characteristics of the switch change under different loading conditions depending on the switch architecture (cell relay components) and type of microprocessor(s)
- programming (software)—The signaling operations of switches are usually performed under software control.
Testing routines should include point-to-point and point-to-multipoint calls as well as support for all signaling protocols used. Automated signaling performance testing is preferred for maintaining consistent testing methodology and results. This benchmark provides a relative measurement, not a pass and fail result (as with conformance testing).
Signaling Performance Analysis
Performance testing establishes benchmarks by measuring the operational characteristics of the signaling protocol state machine under different loading conditions. Signaling performance testing measurements and analysis should focus on four main areas:
-
latency—Latency measurements establish the processing time required for signaling messages to be processed (e.g., set up) through the switch. Varying the call-loading rate applied to the switch allows measurements to be taken under a variety of conditions. These signaling functions exercised should include processing of the following:
- setup messages
- connect messages
- release messages
-
burst—These tests focus on measuring the time required to establish a number of calls. Network managers must know how long it will take to establish a burst of calls when networks are initialized or during cut-over periods (i.e., after catastrophic failure). This testing focuses on measuring the amount of time needed to establish a number of calls. A specific number of calls are transmitting to the system under test during a limited period of time. The amount of time to release a number of calls can also be analyzed. Short-term burst analysis is not sufficient to confirm full functional operation of a switch signaling performance. Switch performance tends to degrade over time for the following reasons:
- memory fragmentation—As blocks of memory are reserved and then released there is a resulting loss of contiguous free space. This can have a severe impact on performance.
- status message traffic—Although status messages only constitute low bandwidth traffic they affect overall performance for the following reasons: status messages are continually being transmitted and received; they require high priority processing; and processing of messages may require or result in rescheduling or retransmission of data messages.
- protocol complexity—The signaling protocol requires a significant amount of processing to maintain databases for tracking message transaction information.
-
throughput—Signaling performance testing must include long-term conditions to identify the effects of large number of network calls being established and then released.
-
limits—identifies the maximum ranges of settings supported by the switch; as an example, tests could include determining the maximum number of concurrent point-to-point calls that a device under test can establish and actively maintain
Delays in switch operations above may result in increased buffering of new signaling messages. Increased buffering may cause overflow (message loss) of the call queue resulting in message retransmission that will increase the switch loading levels.
