We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Distributed, Parallel, and Cluster Computing

Title: Zeus: Locality-aware Distributed Transactions

Authors: Antonios Katsarakis (1), Yijun Ma (2), Zhaowei Tan (3), Andrew Bainbridge (4), Matthew Balkwill (4), Aleksandar Dragojevic (4), Boris Grot (1), Bozidar Radunovic (4), Yongguang Zhang (4) ((1) University of Edinburgh, (2) Fudan University, (3) UCLA, (4) Microsoft Research)
Abstract: State-of-the-art distributed in-memory datastores (FaRM, FaSST, DrTM) provide strongly-consistent distributed transactions with high performance and availability. Transactions in those systems are fully general; they can atomically manipulate any set of objects in the store, regardless of their location. To achieve this, these systems use complex distributed transactional protocols. Meanwhile, many workloads have a high degree of locality. For such workloads, distributed transactions are an overkill as most operations only access objects located on the same server -- if sharded appropriately.
In this paper, we show that for these workloads, a single-node transactional protocol combined with dynamic object re-sharding and asynchronously pipelined replication can provide the same level of generality with better performance, simpler protocols, and lower developer effort. We present Zeus, an in-memory distributed datastore that provides general transactions by acquiring all objects involved in the transaction to the same server and executing a single-node transaction on them. Zeus is fault-tolerant and strongly-consistent. At the heart of Zeus is a reliable dynamic object sharding protocol that can move 250K objects per second per server, allowing Zeus to process millions of transactions per second and outperform more traditional distributed transactions on a wide range of workloads that exhibit locality.
Comments: Sixteenth European Conference on Computer Systems
Subjects: Distributed, Parallel, and Cluster Computing (cs.DC)
DOI: 10.1145/3447786.3456234
Cite as: arXiv:2104.01327 [cs.DC]
  (or arXiv:2104.01327v2 [cs.DC] for this version)

Submission history

From: Antonios Katsarakis [view email]
[v1] Sat, 3 Apr 2021 07:10:57 GMT (5210kb,D)
[v2] Tue, 6 Apr 2021 13:28:53 GMT (5210kb,D)

Link back to: arXiv, form interface, contact.