References that may be relevant to RAMCloud:

• Lorch, et al., The SMART Way to Migrate Replicated Stateful Services, EuroSys 2006. Describes a new approach for implementing configuration changes in Paxos.
• Nishtala, et al., Scaling Memcache at Facebook, NSDI 2013. Describes several interesting optimizations made to memcache as part of the large-scale Facebook deployment.
• Triplett, et al., Resizable, Scalable, Concurrent Hash Cables via Relativistic Programming, USENIX 2011. Describes an implementation of concurrent hash tables that can be resized without blocking reads (uses "wait for readers", which is similar to what we do in log cleaning). Lots of references to other concurrent hash table implementations.
• Stonebraker et al., The End of an Architectural Era (It's time for a Complete Rewrite),
  VLDB '07. Discusses limitations of current RDBMS, argues for "clean-slate" redesign of database systems.
• Peta-Scale Solid State Storage at UCSD. A project for data-center storage based on Flash. Their ASPLOS'09 paper describes their custom node design and custom FTL FLASH remapping.
• The Apache UIMA frameworkthat could be useful as an application on top of RAMcloud.
• The Yahoo! PNUTS project : shares many goals with RAMCloud, including scalability and multi-tenancy. Not clear how much focus there is on main-memory storage, though.
• The FAWN project at CMU: optimizing queries/joule using small nodes.
• Gordon: Peta-scale Solid-state Storage Systems
• Scalable, Distributed Data Structures for Internet Service Construction, by Steven D. Gribble, Eric A. Brewer, Joseph M. Hellerstein, and David Culler. Proceedings of the Fourth Symposium on Operating Systems Design and Implementation (OSDI 2000).
• Main Memory Database Systems: An Overview , by Hector Garcia-Molina et al.
• Distributed data structure work by Steven Gribble: paper, thesis
• Lin et al., Low-Latency, High-Throughput Access to Static Global Resources within the Hadoop Framework - A report on how a Hadoop system using memcached can enable implementations of algorithms which were previously not possible with MapReduce. Their latency is around 2 or 3 ms, which puts an upper bound on their throughput. WIth a system like RAMCloud, which would have much lower latency, much higher throughput could be achieved.
• The Datacenter as a Computer by Barroso and Holzle.
• DRAM Errors in the Wild: A Large-Scale Field Study by Schroeder, Pinheiro, and Weber
• The Storage Chasm: Implications for the Future of HDD and Solid State Storage by Steven R. Heltzer
• Jeff Dean's Keynote at LADIS 2009
• Slides on Information Retrieval and the video from the presentation by J. Dean at this year's WSDM.
• Measurement and Analysis of Large-Scale Network File SystemWorkloads by Leung et al., from 2008 USENIX ATC.
• Migrating Server Storage to SSDs: Analysis of Tradeoffs, by Narayanan et al.: argues that SSDs don't make sense for today's datacenter workloads.