The Management Systems Revolution
Or
The Birth of the Enterprise Management Bus

 

  IT consumers are uniformly embracing the need to more efficiently manage hardware and software resources in a distributed enterprise while simultaneously improving resource utilization and application availability. In many cases the management of computing resources in an enterprise has become the largest component of the cost of ownership. IT consumers further require that new and sophisticated management solutions be comprised of multi-vendor components.

As such, the computer industries ability to deliver the vision of enterprise management is dependent on the existence of a new wave of open and interoperable systems management interface standards through which products from multi-vendors may be seamlessly integrated. To accelerate this emergence, the industry requires a set of standard management interfaces that allows the control systems in a data center to reliably and seamlessly interoperate.

A critical customer requirement associated with the evolution of enterprise management solutions is their ability to fundamentally reduce the human overhead. Today, the management of a datacenter requires the use of a suite of uncoordinated applications from multiple vendors. Additionally, these applications lack the required functionality, distribution, security, and reliability to provide increasing business efficiency. Ultimately, the high cost of management deters end-users from aggressively embracing new technology. The incompatible management APIs that thread through today’s multi-vendor datacenters also inhibits end-users from selecting between multiple suppliers when constructing and automating system solutions

1.1 Technology Requirements

To reduce the cost of systems management, products integrated into a solution must have highly functional interface technology (vastly more functional than the interfaces that exist today). To provide this functionality inter-product management interfaces must provide more advanced features like security, extensibility, auto-discovery, internationalization, and asynchronous event processing (see appendix A for initial technical requirements that would render distributed utility computing solutions manageable). New standards must be targeted at delivering the highly functional and seamlessly interoperable management interfaces needed to realize this vision.

With one set of object models and one protocol stack, management applications will become simpler and more effective. Ultimately, faced with reduced costs for management, IT Consumers will be able to adopt technology faster and build larger more powerful computing solutions. Additionally, interoperable interface standards allow UC vendors and integrators to decrease the time required to bring new more functional technology, products, and solutions to market.

Figure 1 shows the results of a requirements survey of Fortune 1000 companies in 2003 by TheInfoPro. The survey clearly shows a strong desire among users for improved interoperability and manageability.


Figure 1: Survey of End User Management Requirements

Figure 1 illustrates that interoperability among components supplied by multiple vendors is a key concern for end-users and integrators. End-users want to treat data center resources as multi-vendor integration backbones that insulate them from uncompetitive products, supplier failure, and unreasonable pricing practices.

1.2 ISV Management Application Development
Today, to deliver a comprehensive management application, a developer is burdened with the tedious task of integrating the incompatible interfaces of multiple vendors. These disparate interfaces utilize a variety of transport technologies, protocols, underlying object models, event notification systems, security features, durable naming systems, and discovery systems. The result of this integration task is a large and expensive custom integration infrastructure that yields:

  • Final customer visible management functionality that is the “least common denominator” of the functionality integrated across multi-vendor implementations of a given device.
  • Slow time-to market integrating new devices and new features in devices.
  • Hi overhead installing and upgrading the management application in customer environments.
  • Inconsistent and unreliable operation when mapping large and dynamic data center configurations.

These factors lead to high support cost and general customer dissatisfaction, both of which affect profitability negatively. Moreover, as time goes on, more devices with more interfaces are introduced, even as users demand more functionality. Management application complexity increases, with corresponding worsening of time to market and robustness. Figure 2 illustrates this situation.


Figure 2: Management Application Complexity Today

As Figure 2 suggests, a comprehensive management application requires support for multiple object models for each type of device supported, as well as for the management and transport protocols required to communicate with each other. In addition, each component to be supported must be qualified with its object model and protocol infrastructure.

The cost of developing and maintaining this complex management infrastructure is not recoverable as increased revenue. Users don’t pay for infrastructure; they pay for value to their information technology operations, in the form of functionality, coverage, and simplicity. Thus, today, management vendors who strive to integrate a large number of multi-vendor devices/subsystems and seek a related market advantage are forced into building less competitive products.

When broadly adopted, management interface standards will solve this problem by replacing today’s multiple disparate managed object models, protocols, and transports with common models for each object class and a common protocol for management interactions. Management application developers will be able to support devices from multiple vendors quickly and reliably because they “look alike”. Development and testing costs will drop; time to market will be shorter; customer satisfaction will increase, and ultimately, profitability will rise.


Figure 3: Management Applications with Standards

Figure 3 illustrates the management application environment enabled by a single highly functional standard interface using Common Information Model (CIM) technology. The figure suggests, a single object-oriented model for each type of component in a product stack. These models define common attributes and behavior for standard features. Management actions are communicated using a single management protocol and transport that can be carried on any of several interconnects. Included are a uniform set of discovery, security, and durable naming services. To enable multiple management applications to coexist in a network a lock manager that can be used to prevent applications’ actions from interfering with each other.

CIM object models are extensible, enabling easy addition of new devices and functionality to the model, and allowing vendor-unique extensions for added-value functionality.

1.3 OEM Product Development
Figure 4 illustrates the burden of supplying APIs to ISVs from the subsystem developer’s point of view.


Figure 4: Complexity for Subsystem Developers

Device and subsystem developers must implement and support multiple management protocols (and often multiple object models) for their products to be supported by management application vendors. In some cases, proprietary distributed transports must be used and distributed in software development kits (SDKs) that are unique for each operating system, platform, programming language. Component developers must effectively double as software companies (without revenue) to enable management application across the industry to support their products.

When adopted by subsystem developers CIM based management interface standards promulgated through a single platform independent and distributed management transport will reduce the time and cost of delivering management APIs for subsystems to application developers, and will thereby reduce the time required for new devices and new functionality to be supported.

The figure 4 illustrates device vendor advantages associated with the broad adoption and use management interface standards in the industry.


Figure 5: Management with standard interfaces

1.4 Integration Technology
As defined by the Global Grid Forum’s (GGF) Open Grid Services Architecture (OGSA), a extensive array of software components will ultimately comprise a complete solution in a large enterprise. Some components include program execution management services, resource management services, security services, discovery services, metering services, fault and health management services, rating services, billing services, and resource provisioning/virtualization. To ensure these services integrate effectively into powerful solutions a common object model across services is required. An example of the power of integrating solutions via a common model is illustrated through the desire to perform root cause failure analysis across the products in a utility computing platform. As illustrated below, without rigorously architected and reliable associations between the management models in each product health and fault analysis is nearly impossible (particularly in a multi-vendor environment).

Today, the DMTF Common Information Model (CIM) is the only object model representing a breadth of management capabilities for hardware and software resource spanning an enterprise. To this end, this specification and is a coordinating entity across the vendor community such that a Common Information Model for is achieved.

To achieve seamless interoperability between multi-vendor services within an enterprise needs a precise method for expressing the capabilities of a given services management interface. This interface expression within the DMTF Common Information Model (CIM) is described using a technique called profiles. Profiles define the minimum functionality (objects, properties, methods, and associations) within the larger CIM schema that a given service provides to a management client for the purpose of achieving interoperability. The description of a profile is programming language and system platform independent. That is, a profile within CIM can be built into a product using Web services, CORBA, WBEM, or a modern programming language like C++ on any modern O.S. platform. Associations between profiles (i.e., UC services) in CIM enables powerful management solutions to be integrated from multiple products in support of the vision for utility computing. Cooperation between vendors to use a single distributed communication protocol to connect clients and servers for a given profile ultimately provides consumers of IT technology desired multi-vendor interoperability.

1.5 Welcome the Enterprise Management Bus
Akin to the Enterprise Service Bus (ESB), an Enterprise Management Bus (EMB) is an open standards based software communications backbone that interconnects computing resources, infrastructure, and applications in an enterprise or complex system for the purpose of integrating management. An EMB enables the connection of systems, storage, networks, software, and applications independent of vendor, operating system platform, and programming language for the purpose of maximizing system availability, reliability, and performance while simultaneously reducing resource utilization and the cost of management. An EMB is a software messaging architecture that streamlines the management of enterprise computing resources or complex systems through interoperable integration. An EMB is a key component supporting the ultimate success of Service Oriented Architectures. The first EMB was documented in the architecture of the CIM/WBEM based SMI-S industry standard for interoperable storage and storage network management. The SMI-S standard (originally called Bluefin) was launched and led by Roger Reich of Olocity.

Shown below is the Reference Model for an EMB that illustrates key constituents of the environment.


Figure: EMB Reference Model

Roles for Constituents

A Client is the consumer of the management information in the environment like a graphical presentation framework. The xIM architecture or its specific product instantiations like StorageIM are examples of clients. Olocity provides tools and services to accelerate the construction of clients in this environment.

An Agent implements a subset of the object manager and as such controls only one device or subsystem and is typically incapable of providing support for complex intrinsic methods like schema traversal. An agent may be embedded in a device (like a Fibre Channel Switch) or provide a proxy to a device over a legacy or proprietary interconnect (like a SCSI based array controller). A variety of tools exists to facilitate the rapid construction of Agents – Olocity is available to building or consult in the building of Agents in an EMB system.

Embedding an agent directly in a device or subsystem reduces the management overhead of a customer and eliminates the requirement for a stand-alone host (running the proxy agent) to support the device. Embedding an agent is advantageous in reducing management and setup of Agents in product environments. The construction embedded agents generally requires more effort to reduce footprint – Olocity is experienced in developing Agents and available to build or consult in the building of embedded agents in an EMB system.

An Object Manager serves management information from multiple devices or underlying subsystems through providers. As such an Object Manager is an aggregator that enables proxy access to devices/subsystems and can perform more complex operations like schema traversals. An object manager typically includes a standard provider interface to which device vendors adapt legacy or proprietary product implementations.

A Provider expresses management information for a given resource such as a storage device or subsystem exclusively to an Object Manager. The resource can be local to the host that runs the Object Manager on or can be remotely accessed through a distributed systems interconnect.

A Lock Manager provides a common service for use by agents and object managers to coordinate resources between multiple non-cooperating clients such that Isolation and Consistency for the information in the schema is maintained.

A Directory Server provides a common service for use by clients and agents for locating services in the management environment.

1.6 Olocity and the Future
The management of multi-vendor computing solutions and complex systems is increasingly becoming the largest component of cost of ownership. Olocity Corporation is dedicated to delivering leadership the technology and services that fundamentally streamlines systems management. In support, Olocity Corporation is the center of expertise within the computer industry and is committed to evolving the open industry standards that are critical to accelerating systems management technology. We are making the dream of efficient management systems a reality. Contact Olocity for the services and products to deliver the dream.

Source: Roger Reich, President and CEO, Olocity Corporation - U.S.: 719-884-0133, Info@olocity.com

Back to Article List