Data: NoSQL Databases
Location: B118-119
Presentations: PNUTS Presentation [PPTX],
PNUTS Presentation 1 [PPT]
Average rating: ***..
(3.50, 2 ratings)

PNUTS is a large-scale, geographically-repicated data store in widespread use at Yahoo! PNUTS targets serving workloads, supporting a huge number of concurrent small operations with extremely low latency (reads and writes of single records, reads of small ranges of records). PNUTS provides the key features of a cloud/nosql database: massive scalability, replication, elasticity. In addition, PNUTS supports geographic replication; tables are replicated across data centers all over the world.

I will overview the main components of PNUTS and our key design decisions. I will also discuss some of our extensions, both within the PNUTS core and layered on top. I will also briefly compare it to some of the well-known open-source cloud databases.

Photo of Adam Silberstein

Adam Silberstein


Adam Silberstein is a Research Scientist in the Systems group at Yahoo! Research. His main area of interest is large-scale data management, including work in online serving systems, batch processing systems, and management of social data. His work has appeared in top database conferences such as SIGMOD, VLDB and ICDE. Prior to joining Yahoo!, Adam completed his Ph.D. at Duke University in 2007.

Comments on this page are now closed.


Picture of Sheeri K. Cabral
Sheeri K. Cabral
09/14/2011 7:22pm PDT

A video of this presentation is at

Picture of Adam Silberstein
Adam Silberstein
06/27/2011 2:09am PDT

PNUTS has been running in production at Yahoo for 2+ years and growing constantly. It is one of the few large-scale serving systems that provides geographic replication. I’ll talk about our operational experience and some of the exciting (and unpublished) work we’ve done lately.