As the Army places greater priority on collaborative and joint simulation capabilities, the technical community must enable their simulation software and network infrastructure to handle the network load that comes with large, geographically-dispersed federations. The Army Capabilities Integration Center (ARCIC) is the technical lead for the Omni Fusion 2008 and 2009 experiments. Omni Fusion 2009 includes the technical baseline for the Joint Experiment Earth Wind and Fire (EWF) 2009, and is also the technical path to the Omni Fusion 2010/TALON STRIKE, a distributed coalition event used by the UK to train a brigade headquarters for deployment to Afghanistan. Each ARCIC-lead Omni Fusion experiment employed OneSAF as the primary entity driver to successfully model and record approximately 20,000 engaging entities across a Wide Area Network (WAN) using the High Level Architecture (HLA). The first experiment attempted to handle the high network load by improving network scalability and process robustness. The second experiment relied on a comprehensive Data Distribution Management (DDM) strategy, which lightened the load on the network but increased software complexity. The similarity of these experiments' simulation environments offers a unique opportunity to compare technical approaches for handling network load in complex federations. This paper aims to serve as a guide for others in the community who are supporting similar high-entity count federations, both for experimentation and for training applications. It gives detailed lessons learned from both approaches, such as helpful RTI and OneSAF configuration parameters, operating system tweaks, and common DDM pitfalls. As the DoD moves toward greater reliance on distributed training of joint forces prior to deployment, the lessons learned by the Omni Fusion experiments will have great value to the community as we train to fight and fight to win.