TEMPEST, a.s. is one of leading system integrators in Slovakia and surrounding countries.
TEMPEST has long-term experience in the area of IT service management (ITSM). ITSM is aimed at fulfilling strategic business objectives in order to allow flexible reaction to market demand, to connect the IT capabilities with the business needs, and to enable planning, management and measurement of the quality of IT services.
Many large organizations have different technological systems on which their enterprises run or depend. If there are more than a few devices, they need to be monitored. Usually, each group of technology and/or vendor-related devices comes with its own monitoring solution.
Purpose of the Project
In this project, a major telco company had eight proprietary monitoring systems, each for a group of devices. These included radio data links, signal coders and decoders, radio transmitters and other support systems.
The goal of this project was to collect events from these legacy monitoring systems into an umbrella management console, which could display their data and also some inter-dependencies among systems. The primary purpose was to relieve operators from watching many different systems and, to a certain extent, provide a correlation between these systems.
Analysis and Solution
The solution's central console was HP OpenView Operations (OVO). OVO works as an agent/manager solution, where agents intercept events from the environment and the manager processes, stores, and provides the events to operator consoles.
The OVO manager (or OVO server) was deployed on an HP-UX system. OVO agents were deployed to each of the legacy platforms, most of which were Windows systems, one of which was an older Digital UNIX, and one of which was an even older MS-DOS based solution.
Each of the legacy systems consisted of monitoring software ranging from one to three layers, with one of them being a database or flat file of events. Each of these systems used a very different event logic, event format, and event storage technology.
Python Comes to the Scene
The challenge we faced was to extract the events from each legacy system, transform them to a common format and send them to the OVO manager. For this, we felt we needed a powerful scripting language for the OVO agents on the legacy platforms, one that would facilitate easy deployment and debugging of the scripts there.
Another requirement was that event processing needed to be easily configurable and extensible. Neither the customer nor the implementor could manually configure handling of each specific event type. A configuration tool for filtering and altering messages was needed, and had to be simple enough for the customer to understand and use.
It was clear that a unified approach to integration was needed. Python was chosen for it's flexibility, clarity, and native object-oriented (OO) programming style. Perl was also considered but we felt it would lead to code readability and understandability issues at later times. While easy to write, Perl code often appears very cryptic when read.
A common Python module was developed, with a few common classes that are used on all the OVO agent platforms. Most of the platform-specific customization was accomplished in the class initialization. In some cases, some methods needed to be overridden as well. On most platforms only two Python files were needed.
With IDLE and CVS, development was easy and convenient.
Functionality of Integration Modules
The integration module is first initialized according to the underlying data source. This is an ODBC data source in most cases, although one platform uses flat files. The ODBC data sources include many database formats. One special case with flat files uses a Perl parser as client and the common Python module with a network extension as server.
Next, the module periodically reads new events from the data source. Each event needs to be filtered to determine whether it should be sent at all and, if so, translated into human-readable form. Finally, the event is sent through a common command line interface, OVO agent's opcmsg command. OVO agents have the ability to further modify an event's properties (e.g. provide automatic alarm closure).
Event filtering and translation is controlled through configuration files, which are read in automatically both at start and whenever they are modified. Their purpose is to transform event input fields to output fields or apply rules to transform the event (e.g. drop the event, enrich the event, and so forth). These configuration files have table-like structure and simple syntax, so they are easily readable and comprehensible. Each platform has a different number and format of these files. The files are maintained by the customer.
A powerful yet clear multi-platform scripting language was needed to achieve this project's goal. Python fulfilled this requirement.
With good code reuse -- some 40KB of code in the common module, 5-10 KB of initialization code -- a lot of programming time was saved. Taking into account that each platform is very different, this was a very pleasant result. The majority of the work that had to be done was analysis and testing, not programming.
The modules described here have been running for almost 2 years now without serious issues. The code is still easy to understand and easy to tailor for a new platform or to add a new feature.
Developing integration code with Python is fast and easy. Python's native OO approach -- something that was not available in our older integration languages -- made it easy to create reusable code with minimal effort. This project has shown the great power of Python and open source tools in general.