SA Forum enhances platform management and integration

During the last decade standardization of the various computing architecture layers has done wonders for the industry. It gave the enterprise computing industry the ability to put together systems using COTS building blocks from different hardware, operating system, and application providers.

In the past, proprietary, vertically integrated systems threatened to lock system designers into architectures from individual suppliers. Now system designers can choose from an ecosystem of COTS suppliers to build systems using best-in-class building blocks and quickly and cost-effectively deploy highly available systems for a variety of applications. This paradigm shift - the successful evolution from the vertical to horizontal industry model - is appearing in other markets as well, especially the telecommunications industry.

The telecom industry is seeing increasing adoption of standards and requirements at the hardware, operating system, and middleware layers. From the PICMG and blade.org organizations come xTCA and bladed systems. The Linux Foundation has given us Carrier Grade Linux (CGL), and the Application Interface Specification (AIS) and Hardware Platform Interface (HPI) come from the Service Availability Forum (SA Forum). Aiding the work of these organizations are efforts by the SCOPE Alliance and Communications Platforms Trade Association. Among other things, the SCOPE Alliance, which represents the interests of telecom service providers, publishes gap analyses and profiles that help guide the work of these groups and assist them in prioritizing and identifying functional areas that need attention. The CP-TA ensures there are ways to conduct interoperability testing of various commercial building blocks, a crucial function.

As a result, a viable and vibrant ecosystem of COTS component suppliers is emerging. This ecosystem includes multiple suppliers for each of the building blocks. Telecom Equipment Manufacturers (TEMs), who have traditionally developed vertically integrated systems in house using proprietary and often expensive technologies, now have a compelling alternative that also creates an interesting challenge.

The challenge

Traditionally having developed most, if not all, the layers of the systems in house - the hardware, operating system, middleware, and application - TEMs were in the best position to integrate all these layers into a deployable system. Now that they can buy many of these building blocks off the shelf, the question arises as to who is in the best position to integrate these components. In the COTS model one can effectively argue it is no longer cost-effective for TEMs to assume the system integration responsibility. Doing so results in significant effort on their part to integrate components they have little or no involvement in developing. Their R&D costs as well as their time-to-revenue increase.

The industry is recognizing this challenge and is responding in two kinds of efforts. One, key alliances are forming within the ecosystem to ensure someone within the community - often the hardware platform supplier - takes the responsibility to ensure delivery of fully integrated, tested, application-ready systems. Two, these industry groups are creating specifications and requirements that aid in such integration.

The SA Forum is doing its part to address the platform integration piece. The organization's standard interface specifications facilitate integration of service availability middleware capabilities with the underlying hardware platform's management capabilities, and the SA Forum does this in such a way that this integration is industry standard. Already a commercial success, the HPI has standardized the discovery and management hardware platform capabilities as well as integration of such platforms with AIS-compliant middleware services and other user applications.

Later this year the SA Forum is scheduled to release the next major set of its specifications, Release 6. Fully backwards compatible with Release 5, this release adds major enhancements to the existing Information Model Management (IMM) service and the HPI. These enhancements further improve the services enabled by Release 5, which supports Java integration as well as Java mapping services.

Release 6 also introduces an important new service that, when implemented, enhances functionality to enable sophisticated platform management and integration capabilities. Specifically, this service enhances and arguably completes the information model so that it represents hardware, operating systems, and virtualization entities. The new service, called Platform Management (PLM), coordinates with HPI, IMM, Availability Management Framework (AMF), and other AIS services to provided a functional block to assist system developers and integrators. It allows developers and integrators to tightly integrate the AIS-compliant middleware capabilities to the HPI-compliant hardware platform (Figure1). The result is a highly available, application-ready platform that provides a rich set of high availability services as well as platform management capabilities not previously possible - all using COTS building blocks. The PLM, in conjunction with HPI, greatly facilitates the hardware platform integration and addresses the integration challenge mentioned earlier. Further details of PLM follow.

Figure 1 (click graphic to zoom by 1.4x)

Platform management

The new PLM service enhances the information model by introducing two new entities: Physical Resource (PR) and Execution Environment (EE). PR entities represent the hardware resources discovered by HPI for applications and middleware services. This logical representation not only creates a new abstraction layer appropriate for software usage, but also allows developers to match the configured system to hardware that is actually present and makes relevant hardware states easily accessible for software by a convenient state model. EE entities allow for modeling the software resources, such as operating systems running on physical hardware in the system and virtual machines that may be provided through a virtual machine monitor or hypervisor.

Some of the key capabilities with which the PLM service enhances platform management and facilitates deeper platform integration are summarized here.

Information model

Currently, there is a key difference between the HPI view of the system and that of the AIS. HPI performs the hardware resource discovery and serves it up to the user application. AIS works on a preconfigured system model. Before the AIS services are started, some entity within the system must ensure that the discovered hardware indeed matches what the AIS services expect as defined by its information model. Enter PLM!

The PLM service represents PR as well as EE objects in the information model. For PR entities, PLM objects in the information model represent the configured hardware entities. PLM provides its user applications - the AIS services, for example, that have subscribed to the notifications from PLM - callbacks to inform them about the hardware events on entities of interest to the subscribing application. For EE entities, PLM introduces new objects within the information model to represent virtualized architectures where virtualization facilities are able to run multiple operating systems on virtual machines.

State model

AIS implements a sophisticated state model that is used to manage redundancy. Changes in states of various objects in the system trigger specific notifications and subsequent actions to maintain high availability. In addition to introducing the PR and EE objects, PLM defines the various states for these objects to be incorporated in the state model. For example, for a configured physical resource, the states in the model reflect whether the PR is physically present, configured, and active. Similarly, for the EE entities the PLM defines and introduces states that can be used to indicate whether a particular software entity has been loaded or not.

OA&M

PLM defines objects in the information model to affect operation, administration, and maintenance of both object types. A set of administrative commands are provided to locate errors and inform user applications when events occur that require OA&M operations. One example is a hot swap event where a Field Replaceable Unit (FRU) is inserted or extracted.

Virtualization

In an SA Forum view of the world, the virtualization layer resides between the AIS services and the HPI. Since PLM coordinates AIS and HPI, it makes sense for it to manage this layer. As such, PLM defines specific EE entities to support virtualization. For example, a virtualized operating system would be modeled as an EE entity of the physical resource it is running on.

Fault analysis and correlation

Effective fault management is the key to achieving high availability. This requires quick detection, analysis, correlation, and isolation of faults so recovery actions can be put into effect. In a clustered system, it is crucial to centralize analysis of the health states of various entities, given that usually a particular fault affects multiple user applications. PLM offers fault analysis of both PR and EE entities. Furthermore, in conjunction with the AIS Notification Service (NTF), PLM correlates faults to isolate failures and speed up the recovery actions.

Upgrade support

The SA Forum Software Management Framework (SMF) defines interfaces for software upgrade functionality. PLM provides specific administrative operations on PR as well as EE entities that, in combination with the SA Forum SMF, can affect a hardware upgrade, or a rolling upgrade of an operating system instance.

The last word

In the past few years, the SA Forum has made significant strides in helping to accelerate COTS adoption in telecommunications by proliferating standard interface specifications for high availability services that aid in creating deployment-ready systems. Integration of these services with various hardware platforms is the next industry challenge, and the SA Forum is doing its part in helping the industry address it.

Dr. Asif Naseem is Chief Technology Officer and Chief Operating Officer, GoAhead Software, Inc. in Seattle, WA. He has more than 19 years of experience in the computer and communications industry. Dr. Naseem is a veteran speaker having presented at national and international events such as ITU Telecom Geneva, GSM World Congress, CTIA, and numerous other events. He has also presented papers at conferences organized by ACM, IEEE, and others, and is a frequent author of papers published in several technical journals and magazines. He has an MS in Electrical Engineering and a PhD in Computer Engineering from Michigan State University.

Links

www.blade.org
www.cp-ta.org
www.linuxfoundation.org
www.picmg.org
www.saforum.org
www.scope-alliance.org

References

[1] Service Availability Forum documentation

[2] Asif Naseem. "An Update on the Service Availability Forum" ATCA Newsletter. April 2008.

[3] Asif Naseem. "Integration is the key to Open Systems Development!" ATCA Newsletter. June/July 2008.

[4] Uli Kleber, Frederic Herrmann, and Ulrich Horstmann." "SA Platform Management with SA Forum and Its Role to Achieve High Availability." ISAS 2008