Manageability refers to the collective set of properties of an entity (that is, a device or a network) that allow it to be managed. Those properties can largely be attributed to the entity's management instrumentation—the management interfaces that it exposes and the capabilities that those interfaces provide. However, it involves other properties as well that make the entity easier to manage, such as capabilities that reduce the need of a device to be managed in the first place. Manageability is not binary in nature (either you have it, or you don't); instead, it represents a continuum that consists of a multitude of different factors that can make management easier or harder. No single factor determines manageability.
Therefore, we take a look at some of the factors that influence manageability. There are other factors besides those presented here. However, the discussion should provide a good notion of what types of aspects to look for in networks and managed devices that make them easy to manage.
We begin with manageability factors that affect cost of ownership. Two aspects are associated with cost of ownership: the cost of integrating the managed devices into an operations support environment, and the cost of operating and maintaining the device.
The cost that is involved with getting managed devices supported by management applications is often also referred to as the "integration tax." Higher manageability results in decreased development time and faster time to market of management applications, leading in turn to devices that are more widely supported by management application vendors. (Clearly, other factors influence this cost, including the ease of maintaining the management applications themselves.)
Here are examples of how various aspects of manageability can impact the integration tax:
■ Consistency of (machine) management interfaces facilitates design and code reuse for application development and device integration, and is hence crucial to containing the integration tax. Consistency can be an issue at multiple levels—the consistency of management interfaces across different devices, across different revisions of the same device type, and even across different but related functions of the same device type. In each case, consistency means that the same management feature, instrumented in different places, works and behaves in the same manner. For example, if a device supports two kinds of network interfaces, it would be considered consistent if the aspects that are common to both can be managed using the same interface MIB. On the other hand, it would be considered inconsistent if both types of interfaces were represented by entirely different MIBs, or if one network interface would need to be managed using SNMP and the other through Netconf. Likewise, if two different devices both support the same routing protocol but they are managed in different ways with regards to that routing protocol, that would be inconsistent.
■ Adherence to management standards, such as support for standardized MIBs, is one technique for ensuring consistency of management interfaces. This is not surprising: The very purpose of management standards is to promote consistency of management interfaces.
■ Reliable events facilitate event-driven design of management applications. They reduce the amount of exception handling that management applications must implement to deal with the possibility that important events are missed, and they obviate the need for additional polling mechanisms.
■ Self management of certain functions reduces their need to be managed (and to be supported by management applications) in the first place. An example is a capability for devices to exchange certain configuration parameters about each other through a mutual handshake instead of needing to have the configuration parameters provisioned by a management application.
■ Introspection capabilities allow applications to obtain information about the management interface and capabilities that are supported by a device, without requiring this information to be coded into the application. Such information consists of metainformation (that is, information about information) concerning data revisions, management parameters, and commands supported. Introspection capabilities can thus facilitate data-driven design of management applications that makes maintaining the applications easier.
■ Threshold-crossing alerts help management applications avoid the need to implement elaborate polling schemes and make it easier for them to scale.
The second aspect of cost of ownership concerns lowering operational cost. Loosely speaking, this refers to the cost that is required for operations personnel to support networking equipment. Of course, in many cases, operations personnel does not interact with the managed devices directly; instead, it interacts with management systems in the operations support infrastructure that hide many of the device intricacies. Nevertheless, there are examples of instances in which improved manageability at the device level directly impacts operational cost. It turns out that many are related to the same factors that influence the integration tax:
■ Consistency of (human) management interfaces (for example, the CLI) facilitates the task of network administrators when interacting with the device directly, not through a management application. Consistency reduces training cost and increases the pool of personnel that can manage different pieces of networking equipment.
■ Support for management policies allows network managers to configure only a few management policies instead of requiring tuning of many low-level parameters. An example is the capability to configure parameter defaults that will be applied to all ports on a given line card (for example, an echo-cancellation setting for voice ports) instead of requiring each port (possibly hundreds of them) to be configured individually. It might still be possible to override such policies on an individual basis, but chances are that most settings that need to be applied are the same. Admittedly, this is an extremely simple example of a policy, yet one that will clearly make life easier for network managers.
■ Comprehensive diagnostics capabilities allow a network manager to more easily troubleshoot a device or network.
■ Self management of certain functions reduces the need for management intervention, clearly simplifying life not just for application developers, but for network administrators as well.
Everything listed so far relates to lowering the total cost of ownership. However, as previously mentioned, manageability can also impact the revenue side of the equation. For example:
■ Threshold-crossing alerts allow early and proactive reaction to performance degradation. This can result in higher service levels and ultimately higher revenues.
■ Comprehensive data-collection capabilities in the device to provide detailed data about service levels can be used in accounting and dispute resolution.
■ Support for resource partitioning enables network providers to divide a device into multiple logical devices and might facilitate providing new types of service altogether, such as a "virtual router" service.
Finally, manageability can be a factor that contributes to the availability of networks and services. Manageability features can help reduce the possibility of availability being affected by accidental operational errors. They also help accelerate problem diagnostics and repair, to minimize downtime. Increased availability in turn increases customer satisfaction and leads to increased revenues as well. Here are some examples of such features:
■ Concurrency support within a device and across devices facilitates coordination between competing management applications. Concurrency support includes the capability of management applications to lock parts of the configuration for their exclusive use. This helps avoid situations that result in inconsistent network configurations and service outages from multiple management applications that unintentionally conflict with each other.
■ Integrity checksums help avoid outages from corrupted configuration files. An integrity checksum is a number that is computed from the rest of the file contents and then appended to the file. The basic idea is that if the file has been altered, this can be easily recognized because recomputation of the checksum will lead to a different number that does not match the one that was originally appended.
■ Input validation—syntax and semantic checks—before actually attempting command execution helps avoid outages that are caused by invalid commands sent to a device. This is particularly important when there are groups of commands that can be submitted to a device for execution at the same time. For example, if there is an obvious error in the fifth command down the line, it would be nice to know before attempting to execute the commands. This way, the situation can be avoided in which it becomes necessary to back out of the first four commands, leaving the network in an inconsistent state in the meantime.
■ Rollback capabilities allow managed devices to automatically fall back to the last-known working configuration in case of a configuration error. This allows devices to recover from errors that were not caught or avoided by other means.
■ Self management capabilities decrease the dependency on external network managers and management applications, which also avoids operational errors or untimely management actions that could affect availability.
■ And again, threshold-crossing alerts allow early reaction to degradation of performance before availability is impacted.
As can easily be seen from the examples, in many cases the same manageability features contribute simultaneously to several goals. For example, support for threshold-crossing alerts affects not just operational cost, but also the capability to generate revenue as well as network availability.
Management Systems and Operations Support Infrastructure
We now take a look at management applications and identify what types of properties affect management effectiveness at that level.
As we did in the case of manageability, we turn to the cost side of the equation first. It must be recognized that management systems are themselves a cost factor, even if that cost is (hopefully) more than offset by the benefits they provide. The following are some of the aspects that affect the cost associated with management infrastructure—most of them completely unsurprising and in line with that of other business software applications:
■ The capability of a management application to scale. Management applications that scale well might reduce the number of hosts that are required as part of the operational support infrastructure. This lowers the secondary cost of needing to manage that operations support infrastructure itself. Management applications that scale well also facilitate buildout of the network itself because this buildout is less hampered by issues relating to the capability of the operations support infrastructure to keep up.
■ The capability of a management application to be distributed across multiple hosts. This can be one cost-effective way of allowing applications to scale because it is possible to simply add hosts instead of needing to upgrade to a larger (and potentially much more expensive) host.
■ Use of low-cost operating systems, hardware platforms, and databases.
■ Capabilities that facilitate management of management systems themselves. It is important that the task to manage the operations support infrastructure does not unduly distract from the management of the production network.
■ Features that facilitate integration into operations support environments. Perhaps most important, this includes support for comprehensive northbound interfaces. Properties that those interfaces should possess to facilitate an application's integration with other systems largely mirror those of management interfaces of the managed devices themselves. For example, integration cost can be reduced if interfaces adhere to standards as it leads to greater consistency among the interfaces of different management systems that might all have to be integrated as part of the same operations support infrastructure. Likewise, support for events to facilitate event-driven management simplifies the job of systems at higher layers.
In addition to features that lower the cost of the operations support infrastructure, other features lower the operational cost that is associated with using that infrastructure. Here are some examples—it should be clear, however, that specifics often depend on the specific function that the management system is trying to fulfill:
■ The capability of a management system to integrate with different systems can simplify workflows and avoid the need to enter and maintain redundant data.
■ Productive user interfaces can help network managers accomplish more. One aspect to consider is that, in many cases, the productivity of an interface depends on the target user. Occasional users require interfaces that are easy to learn and understand, often relying on drop-down menus, GUI wizards, and elaborate graphical displays. Power users, on the other hand, might find that the same interfaces slow them down, preferring interfaces that are more character based and that offer plenty of keyboard shortcuts.
■ Programmability and customization of management functions enables network managers to automate routine management tasks.
■ Features that assist users (and other systems) to sift through large volumes of management information enables management by exception and further increases productivity of network managers. Examples of such features include these:
— Functions to sort, search, and filter information according to a wide variety of criteria
— Graphical representation of large data sets—for example, through diagrams, curves to visualize trends, and the like
— Visual highlighting of information that seems out of the ordinary
Just as in the case of manageability, management applications are not only about reducing operational cost. There are also ways in which management application features can help increase revenue. For example, management applications can play an important role in raising service levels (for example, service level monitoring applications). This allows service providers to charge more for their services than they would otherwise. Another example concerns programmability of applications, which allows users to customize operations sequences. This can accelerate the deployment of services, as can service provisioning applications. One technique to improve provisioning throughput is to support concurrent communication with multiple devices, which allows the provisioning application to direct provisioning requests at multiple devices simultaneously instead of having to serialize requests.
In addition, in some cases, network management can itself be a service feature that a service provider might be able to charge for. For example, a service provider can make statistical data about a service available to customers, allowing them to track service levels and service usage themselves. In addition, a service provider might offer facilities that allow customers to provision certain aspects of their service; for example, to add bandwidth to a link in a Virtual Private Network in a self-service fashion. Such capabilities are sometimes marketed using labels such as "customer network management."
Finally, many management system features can help increase network robustness and availability. In fact, arguably this is one of the very purposes of many network management applications, such as fault management applications. However, other examples apply to a much wider range of management systems. For example, checkpointing, audit logs, and backup and restore functions allow management systems to more quickly recover from failures. In addition, management systems can perform a significant number of syntax and semantic checks before sending requests to managed devices, thereby decreasing the chance of introducing inconsistent network configurations.
Because this is a book on network management technology, we focus on the effectiveness of that technology, not on the management organization itself. Of course, effectiveness of the management organization that uses management technology is an equally important factor in the running of a network. Here lessons of general organizational effectiveness apply. Providing services and operating networks constitutes one particular scenario to which methodologies developed in that space can be applied.
One example of such a methodology is the Balanced Scorecard. This methodology suggests viewing the organization from four distinct perspectives—learning and growth, business process, customer, and financial—and developing metrics, collecting data, and analyzing the organization with respect to each of these perspectives.
A second example is the Performance Prism, a methodology that focuses more on optimizing organizational effectiveness than assessing it. The Performance Prism distinguishes among five perspectives: stakeholder satisfaction (identifying the stakeholders and what they want), strategies (satisfying the stakeholders), processes (executing the strategies), capabilities (supporting the processes), and stakeholder contribution (helping develop and maintain the capabilities).
Other methodologies exist. It is interesting to observe that this is the blind men and the elephant all over again—one aspect that many of those methodologies have in common is that they all identify a set of different dimensions that are used to characterize the problem space. Of course, there is no single way to assess organizational effectiveness, but there are multiple ways to skin the cat. Accordingly, different methodologies come up with different ways to divide and conquer the problem.
Was this article helpful?