Generic Multiversion STM

Lü, Li; Scott, Michael L.

doi:10.1007/978-3-642-41527-2_10

Cited by 13 publications

(12 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An interesting strategy in STMs has been to reduce spurious aborts only for readonly transactions. This idea has been formally characterized as mv-permissiveness [Perelman et al 2010], and has been used in both single-versioned [Attiya and Hillel 2011] and multiversioned [Diegues and Cachopo 2013;Lu and Scott 2013;Perelman et al 2011] TM algorithms. Here, we seek to reduce spurious aborts even further than mv-permissiveness.…”

Section: Related Workmentioning

confidence: 99%

Time-Warp

Diegues

Romano

2015

ACM Trans. Parallel Comput.

View full text Add to dashboard Cite

The multicore revolution that took place one decade ago has turned parallel programming into a major concern for the mainstream software development industry. In this context, Transactional Memory (TM) has emerged as a simpler and attractive alternative to that of lock-based synchronization, whose complexity and error-proneness are widely recognized.The notion of permissiveness in TM translates to only aborting a transaction when it cannot be accepted in any history that guarantees a target correctness criterion. This theoretically powerful property is often neglected by state-of-the-art TMs because it imposes considerable algorithmic costs. Instead, these TMs opt to maximize their implementation's efficiency by aborting transactions under overly conservative conditions. As a result, they risk rejecting a significant number of safe executions.In this article, we seek to identify a sweet spot between permissiveness and efficiency by introducing the Time-Warp Multiversion (TWM) algorithm. TWM is based on the key idea of allowing an update transaction that has performed stale reads (i.e., missed the writes of concurrently committed transactions) to be serialized by "committing it in the past," which we call a time-warp commit. At its core, TWM uses a novel, lightweight validation mechanism with little computational overhead. TWM also guarantees that read-only transactions can never be aborted. Further, TWM guarantees Virtual World Consistency, a safety property that is deemed as particularly relevant in the context of TM.We demonstrate the practicality of this approach through an extensive experimental study: we compare TWM with five other TMs, representative of typical alternative design choices, and on a wide variety of benchmarks. This study shows an average performance improvement across all considered workloads and TMs of 65% in high concurrency scenarios, with gains extending up to 9× with the most favorable benchmarks. These results are a consequence of TWM's ability to achieve drastic reduction of aborts in scenarios of nonminimal contention, while introducing little overhead (approximately 10%) in worst-case, synthetically designed scenarios (i.e., no contention or contention patterns that cannot be optimized using TWM).

show abstract

Section: Related Workmentioning

confidence: 99%

Time-Warp

Diegues

Romano

2015

ACM Trans. Parallel Comput.

View full text Add to dashboard Cite

show abstract

“…An interesting strategy in STMs has been to reduce spurious aborts only for read-only transactions. This idea has been formally characterized as mv-permissiveness [26], and has been used in both single-versioned [3] and multi-versioned [11,22,27] TM algorithms. Here, we seek to reduce spurious aborts even further than mv-permissiveness.…”

Section: Related Workmentioning

confidence: 99%

“…Indeed, it can rely on standard garbage collection algorithms for MVCC schemes that maintain any version that can possibly be read by an active transaction (as in different implementations in [14,22,27]). The key idea of those algorithms is the following: assume that T is the oldest active transaction, with S(T ) = k; then versions up to (and excluding) k can be garbage collected -note that the newest version is preserved regardless of this condition.…”

Section: Garbage Collection Privatization and Lock-freedommentioning

confidence: 99%

“…However, similarly to the concern of garbage collection, time-warping does not present additional challenges to privatization. Existing approaches to support privatization, in fact, are based on the notion of quiescence, which forces privatizing transactions to wait for concurrent transactions to finish [22] (using, if possible, explicitly identified privatizing operations to minimize waiting). These techniques suffice to ensure that, once a privatizing transaction P has committed, no transaction can time-warp commit and serialize before P .…”

Section: Garbage Collection Privatization and Lock-freedommentioning

confidence: 99%

See 1 more Smart Citation

Time-warp

Diegues

Romano

2014

Proceedings of the 19th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

View full text Add to dashboard Cite

The notion of permissiveness in Transactional Memory (TM) translates to only aborting a transaction when it cannot be accepted in any history that guarantees correctness criterion. This property is neglected by most TMs, which, in order to maximize implementation's efficiency, resort to aborting transactions under overly conservative conditions.In this paper we seek to identify a sweet spot between permissiveness and efficiency by introducing the Time-Warp Multi-version algorithm (TWM). TWM is based on the key idea of allowing an update transaction that has performed stale reads (i.e., missed the writes of concurrently committed transactions) to be serialized by "committing it in the past", which we call a time-warp commit. At its core, TWM uses a novel, lightweight validation mechanism with little computational overheads. TWM also guarantees that readonly transactions can never be aborted. Further, TWM guarantees Virtual World Consistency, a safety property that is deemed as particularly relevant in the context of TM. We demonstrate the practicality of this approach through an extensive experimental study, where we compare TWM with four other TMs, and show an average performance improvement of 65% in high concurrency scenarios.

show abstract

“…Opacity is a safety criteria for transactional memory, and this work presents a tool that automatically identifies two types of bugs that lead to non-opacity. Li Lu (probably the shortest name I've heard) presented "Generic Multiversion STM" [7], a generic way to implement a multiversion STM on top of existing STM systems. Multiversion STM allows readers to view old values of the data and "commit in the past" (personally, I've been a fan of time traveling ever since "Back to the Future").…”

mentioning

confidence: 99%

DISC 2013 review

Timnat

2013

SIGACT News

View full text Add to dashboard Cite

The 27th International Symposium on Distributed Computing, DISC 2013, was held this year in Begin Center, Jerusalem, Israel. The event included two workshops days and three days of the main conference, from October 14th to the 18th. Naturally, this review cannot cover all the works that has been presented, but I will try to give you a glimpse of the event.As one might expect from a symposium on distributed computing, the first workshop day included several workshops, which were executed concurrently. At the end of the day, all of the participants synced in a joint session named: "Love and Fear in the Lab", which was the highlight of the day. Uri Alon and his guitar led the session. Uri talked about how important it is to enjoy the research, and how advisors should encourage their students to do so. The session included some hit songs, most notably: "Sunday at the Lab". If you look it up on youtube, you can find Uri performing it (but not in DISC).The next day was the first day of the main conference. The session started with an invited talk, given by Ravi Rajwar from Intel. The talk's main subject was the concept of lock-elision. Lock-elision is a hardware mechanism that enables avoiding locks when different threads do not actually collide in their execution. This effect is achieved using hardware transactional memory: threads attempt to execute a transaction instead of acquiring a lock, and only revert and acquire a lock when necessary. Ravi emphasized the greater benefit of lock-elision is that it comes free: no change is required in the programming model. Programmers think with locks. Lock-elision enables them to continue to program this way, and yet achieve more parallelism. The new Intel Haswell processor supports lock-elision, and thus, as Yehuda Afek (the session chair) said, it is likely to live forever.This talk was followed by the presentation of the paper "Distributed Minimum Cut Approximation" [4], which was the winner of the best paper award. The paper was presented by Mohsen Ghaffari, and it is a joint work by him and Fabian Kuhn. While the minimum cut is a central problem in graph theory, it was not yet studied thoroughly in the context of distributed computing. The authors presented two new probabilistic approximation algorithms for the problem, and a lower bound.

show abstract

Generic Multiversion STM

Cited by 13 publications

References 20 publications

Time-Warp

Time-Warp

Time-warp

DISC 2013 review

Contact Info

Product

Resources

About