IBM Zurich Research Laboratory
Luca Deri
lde@zurich.ibm.com
Andreas Weder
awe@zurich.ibm.com
The Advanced Network Management group at the IBM Research Laboratory in Rüschlikon, Switzerland, has been working on the management of large telecommunication networks for several years now. Its current research is focused on OSI Systems Management.
OSI Systems Management is a powerful way to manage the large networks many telecommunications companies have built during the last years. The TMN (Telecommunications Management Network) standards will continue to accumulate inportance for service providers and network equipment vendors, as TMNs main points - interoperability, integration and automation - will be major issues in network management for the remainder of this decade.
An issue of great importance when implementing applications is the user interface. Far too
often, programs pass the underlying complexity on to the user or choose the wrong
presentation for data and make things even worse.
We feel that the World Wide Web not only offers a nice, customizable interface, but could also
help in the acceptance of OSI Systems Management by a larger public; in addition, it allows us
to play with different user interface configurations. The WWW world offers easy-to-learn
applications and therefore is highly suitable to make classes of OSI data accessible to
clearly restrictable groups of users with the need to access certain management information
(e.g.on the business level).
The framework used by OSI to define Systems Management on a top level is called the Manager-Agent Framework. It describes the two basic roles assigned to any application willing to participate in managing a network:
The Agent is responsible for mapping the database containing the managed
information onto a network. This database is composed of several so-called Managed
Objects (MO), entities that may represent some underlying resource, which needs to be
managed. An object consists of a set of attributes reflecting certain aspects of the mirrored
resource.
The MOs may be compared with the well-known objects of OOP languages. They allow data to be
organised in a object-oriented way, in the sense that the interface between them and the
network is clearly defined and that objects are merely instances of Managed Object
Classes capable of inheriting properties from superclasses.
The Manager on the other hand is simply an application intending to perform some
operations on managed information. It connects to an agent and issues commands on the MOs in
that agents database.
The list of commands is defined by OSIs Common Management Information Service (CMIS).
The main directives allow a manager to create and delete instances of classes, to set and get
values of attributes of an MO and to spontaneously issue notifications, which are used to
inform an agent of the occurance of an event associated with the managed object.
By combining the strengths of both worlds, OSI Systems Management and the idea of the World Wide Web, we first of all gained the ability to "browse without limits": WWW clients exists for almost all existing machines and tty terminals. And because we are using a lightweight protocol, a manager (i.e. a browser in our case) can also run on a PC rather than on hosts only.
Furthermore, the presentation of OSI data can be easily adapted to match new requirements with little effort. We plan to provide template HTML files, which will allow a customer to design his/her own output files.
Our solution adds new capabilities to building managers and distributes their functionality without introducing security risks; in addition to the standard CMIP security mechanisms we can exploit the security offered by the Internet and HTML.
We designed a small network of server daemons connected to each other and a gateway, which in turn is connected to an HTTP server running on the same system.
The implementation has some properties worth mentioning:
We feel that our idea provides new insights on how to selectively link networks to the
World Wide Web, and our experience with servers and information retrieval in general may
give valuable hints for other parties willing to mirror their resources onto the Web.
We have found it exciting to further expand this system and would like to make a step forward
and discuss our solution with others. And we think that the upcoming WWW 95 conference is a
good way to do all that.