Prototyping Bubba, a highly parallel database system

Boral, Haran; Alexander, William; Clay, L.; Copeland, George P.; Danforth, Scott; Franklin, Michael J.; Hart, Brian E.; Smith, Marc; Valduriez, Patrick

doi:10.1109/69.50903

Cited by 260 publications

(54 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Most distributed database systems (e.g., [9,6,19]) use some form of two-phase locking for processing concurrent queries, which our results show is best for low contention workloads. Other schemes, such as timestamp ordering [5], can avoid deadlocks but still allow multiple transactions to execute concurrently, and so they require read/write sets, latching, and must support rollback.…”

Section: Related Workmentioning

confidence: 67%

Low overhead concurrency control for partitioned main memory databases

Jones

Abadi

Madden

2010

Proceedings of the 2010 ACM SIGMOD International Conference on Management of Data

108

View full text Add to dashboard Cite

Database partitioning is a technique for improving the performance of distributed OLTP databases, since "single partition" transactions that access data on one partition do not need coordination with other partitions. For workloads that are amenable to partitioning, some argue that transactions should be executed serially on each partition without any concurrency at all. This strategy makes sense for a main memory database where there are no disk or user stalls, since the CPU can be fully utilized and the overhead of traditional concurrency control, such as two-phase locking, can be avoided. Unfortunately, many OLTP applications have some transactions which access multiple partitions. This introduces network stalls in order to coordinate distributed transactions, which will limit the performance of a database that does not allow concurrency.In this paper, we compare two low overhead concurrency control schemes that allow partitions to work on other transactions during network stalls, yet have little cost in the common case when concurrency is not needed. The first is a light-weight locking scheme, and the second is an even lighter-weight type of speculative concurrency control that avoids the overhead of tracking reads and writes, but sometimes performs work that eventually must be undone. We quantify the range of workloads over which each technique is beneficial, showing that speculative concurrency control generally outperforms locking as long as there are few aborts or few distributed transactions that involve multiple rounds of communication. On a modified TPC-C benchmark, speculative concurrency control can improve throughput relative to the other schemes by up to a factor of two.

show abstract

Section: Related Workmentioning

confidence: 67%

Low overhead concurrency control for partitioned main memory databases

Jones

Abadi

Madden

2010

Proceedings of the 2010 ACM SIGMOD International Conference on Management of Data

108

View full text Add to dashboard Cite

show abstract

“…The Bubba prototype was implemented using a 40 node FLEX/32 multi-processor with 40 disks [BORA90]. Although this is a shared-memory multi-processor, Bubba was designed as a shared-nothing system and the shared-memory is only used for message passing.…”

Section: Bubbamentioning

confidence: 99%

Parallel database systems

1992

View full text Add to dashboard Cite

show abstract

“…Several prototype systems have been developed within research projects. These include both shared nothing [12], [13] and shared everything systems [14]. Most prototypical parallel database systems are shared nothing.…”

Section: Related Workmentioning

confidence: 99%

OLAP Query Evaluation in a Database Cluster: A Performance Study on Intra-Query Parallelism

Akal

Böhm

Schek

2002

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. While cluster computing is well established, it is not clear how to coordinate clusters consisting of many database components in order to process high workloads. In this paper, we focus on Online Analytical Processing (OLAP) queries, i.e., relatively complex queries whose evaluation tends to be time-consuming, and we report on some observations and preliminary results of our PowerDB project in this context. We investigate how many cluster nodes should be used to evaluate an OLAP query in parallel. Moreover, we provide a classification of OLAP queries, which is used to decide, whether and how a query should be parallelized. We run extensive experiments to evaluate these query classes in quantitative terms. Our results are an important step towards a two-phase query optimizer. In the first phase, the coordination infrastructure decomposes a query into subqueries and ships them to appropriate cluster nodes. In the second phase, each cluster node optimizes and evaluates its subquery locally.

show abstract

Prototyping Bubba, a highly parallel database system

Cited by 260 publications

References 27 publications

Low overhead concurrency control for partitioned main memory databases

Low overhead concurrency control for partitioned main memory databases

Parallel database systems

OLAP Query Evaluation in a Database Cluster: A Performance Study on Intra-Query Parallelism

Contact Info

Product

Resources

About