Large-scale modeling of critical telecommunications facilities and data centers

Abstract

The management of power and air-conditioning infrastructure in large, complex facilities presents many challenges. These range from advanced strategic planning to the daily task of keeping an installation stable. As an installation grows in size and complexity, the sheer number of installation, maintenance and operational activities add risk to the installation as a whole. The lack of a means to monitor overall infrastructure coherence can lead to serious deficiencies in reliability or system capacity. Silcar has pioneered a method to model entire facilities permitting comprehensive virtual stress- testing. This provides a unique and hitherto unavailable global view of the overall state of an installation. Modeling can faithfully represent many important areas of an installation showing actual loading of critical AC and DC subsystems. A model can predict potential areas of risk arising from changes in electrical load, power factor and thermal load. The model can be used to carry out what-if analysis showing the effect on different parts of the installation. The same model can also be used to estimate the reliability of the installation using statistical Monte Carlo simulation. Telecommunications installations with generators undergo regular testing of the generator system to maintain high reliability. This testing is important but can provide a false sense of security since there are many sub-systems in a complex installation that are not operated and proven during such tests. A faithful model of the installation can be used to subject untested subsystems to further scrutiny. A model of an actual telecommunications installation is presented showing the problems that were detected from putting the model through a virtual stress-test. Some issues were already known and the model served to confirm these. The paper will discuss how the model was validated and processes needed to keep it up to date. The model consists of more than 3,000 components representing the DC, UPS and air- conditioning systems, AC & DC distribution elements of the facility. Following the initial success of the pilot model in locating weaknesses, the technique is now being rolled out to 68 mission critical facilities across Australia.