Achieving scalability in computer systems without sacrificing Liability requires a synergistic combination of system architecture, fundamental technologies and implementation. The art of building and running a super-scale compute cluster is complex and challenging. When growing a cluster into the super-scale range, issues that are minor or go unnoticed in a smaller cluster suddenly become glaring issues. What works at small scale offers no promise of working at super-scale levels. Future requirements are beginning to emerge that are expected to require a three orders of magnitude increase in computing speed and associated infrastructure.