Endemic infectious livestock diseases impact animal health and welfare, and food safety. Pathogens spread between farms mainly due to animal movements (purchases/sales) and neighboring relationships. The risk of spreading depends on the within-farm proportions of infected animals, which varies within and between farms over time. A modelling approach is relevant to represent such a complex biological system, permitting the ex-ante evaluation of control strategies under various scenarios. Developing epidemiological models at a regional scale requires to couple within-farm epidemiological models, leading to complex models and to the need for large computational resources, especially when stochastic processes are involved. The objective is to find the best generic framework in terms of computational performance to represent pathogen spread in a cattle metapopulation. Three requirements should be fulfil: (1) a common interface should be used to run population dynamics, and within- and between-herd infection dynamics; (2) a common data structure should be used for animal movements; (3) the shared interface and structure should be easy to understand, to be usable by persons with various skill levels in modelling. Two implementations are available in this framework: synchronized or desynchronized methods. In the first case, herd dynamics evolve simultaneously. At each time step, dynamics are simulated for all the herds. The second case was conceived to be used with distributed computing. Herd dynamics evolve independently from each other as long as no purchase occurs. As a purchase corresponds to a sell, the destination herd has to wait until an animal is available and its infection status known. For the latter case, neighboring contacts cannot be modelled as herd dynamics are not synchronized. These implementations have been tested on a single processor, then will be tested on a computing grid. We have investigated the computing load according to herd size, herd number, number of years, and distribution of animal movements. When on average one movement occurred per year and per herd, the desynchronized method was slower than the synchronized one. On the contrary, if only a few herds (1/5) exchange animals, for the same total number of movements, the desynchronized method was faster. We successfully applied the synchronized framework to Mycobacterium avium subsp. paratuberculosis, which is spread by animal movements. The next step is to evaluate it on a grid, before using it for other pathogens with other spread characteristics