In research presented first at the 22nd ACM Symposium of Operating Systems Principles in 2009, a team from Carnegie Mellon University and Intel Labs concluded that future compute servers for large data centres should use simple embedded processors similar to and build them into huge arrays of, as they put it, 'wimpy nodes'. The result was a cluster that achieved 300 queries per joule, which the researchers claimed as being two orders of magnitude better than traditional disk-based clusters. HP Labs achieved similar results with the 'microblade' server developed as a predecessor to its Project Moonshot a data warehouse architecture made of simple, low-power processors.