Future Internet services require access to large volumes of dynamically changing data records that are spread across different locations. With thousands or millions of distributed nodes storing the data, node crashes or temporary network failures are normal rather than exceptions and it is therefore important to hide failures from the application.
We suggest to use peer-to-peer (P2P) protocols to provide self-management among peers. However, today's P2P protocols are mostly limited to write-once/read-many data sharing. To extend them beyond the typical file sharing, the support of consistent replication and fast transactions is an important yet missing feature.
We present Scalaris, a scalable, distributed key-value store. Scalaris is built on a structured overlay network and uses a distributed transaction protocol. As a proof of concept, we implemented a simple Wikipedia clone with Scalaris which outperforms the public Wikipedia with just a few servers.
2009. 148-159 p.