Schemes for avoiding starvation in transactional memory systems

Waliullah, M. M.; Stenström, Per

doi:10.1002/cpe.1363

Cited by 7 publications

(4 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For performance evaluation of KSFTM with the state-of-the-art STMs, we implemented the the algorithms PKTO, SV-SFTM [9,27,26] along with KSFTM in C++ 2 . We used the available implementations of NOrec STM [6], and ESTM [7] developed in C++.…”

Section: Experimental Evaluationmentioning

confidence: 99%

“…r 1 (y, 0) r 1 (x, 0) Figure 1: Limitation of Single-version Starvation Free Algorithm Related work on the starvation-free STMs: Starvation-freedom in STMs has been explored by a few researchers in literature such as Gramoli et al [9], Waliullah and Stenstrom [27], Spear et al [26]. Most of these systems work by assigning priorities to transactions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Achieving Starvation-Freedom in Multi-version Transactional Memory Systems

et al. 2019

View full text Add to dashboard Cite

Software Transactional Memory systems (STMs) have garnered significant interest as an elegant alternative for addressing synchronization and concurrency issues with multi-threaded programming in multi-core systems. For STMs to be efficient, they must guarantee some progress properties. This work explores the notion of one of the progress property, i.e., starvation-freedom, in STMs. An STM system is said to be starvation-free if every thread invoking a transaction gets the opportunity to take a step (due to the presence of a fair scheduler) such that the transaction eventually commits.A few starvation-free algorithms have been proposed in the literature in context of single-version STMs. These algorithms are priority based i.e. if two transactions are in conflict, then the transaction with lower priority will abort. A transaction running for a long time will eventually have the highest priority and hence commit. But the drawback with this approach is that if a set of high-priority transactions become slow, then they can cause several other transactions to abort. So, we propose multi-version starvation-free STM system which addresses this issue.Multi-version STMs maintain multiple-versions for each transactional object. By storing multiple versions, these systems can achieve greater concurrency. In this paper, we propose multi-version starvation-free STM, KSFTM, which as the name suggests achieves starvation-freedom while storing K-versions of each t-object. Here K is an input parameter fixed by the application programmer depending on the requirement. Our algorithm is dynamic which can support different values of K ranging from one to infinity. If K is infinite, then there is no limit on the number of versions. But a separate garbage-collection mechanism is required to collect unwanted versions. On the other hand, when K is one, it becomes the same as a single-version starvation-free STM system. We prove the correctness and starvation-freedom property of the KSFTM algorithm.To the best of our knowledge, this is the first multi-version STM system that satisfies starvation-freedom. We implement KSFTM and compare its performance with single-version starvation-free STM system (SV-SFTM) which works on the priority principle. Our experiments show that KSFTM gives an average speedup on the worst-case time to commit of a transaction by a factor of 1.22, 1.89, 23.26 and 13.12 times over PKTO, SV-SFTM, NOrec STM and ESTM respectively for counter application. KSFTM performs 1.5 and 1.44 times better than PKTO and SV-SFTM but 1.09 times worse than NOrec for low contention KMEANS application of STAMP benchmark whereas KSFTM performs 1.14, 1.4 and 2.63 times better than PKTO, SV-SFTM and NOrec for LABYRINTH application of STAMP benchmark which has high contention with long-running transactions. * A preliminary version of this work was accepted in AADDA 2017 as work in progress. † A part of this work was submitted towards the fulfillment of M.Tech thesis requirement by the author. ‡ Author sequence follows a lexical order of last...

show abstract

Section: Experimental Evaluationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Achieving Starvation-Freedom in Multi-version Transactional Memory Systems

et al. 2019

View full text Add to dashboard Cite

show abstract

“…LogTM [17] employed a requester-stall approach when resolving conflicts, with a software handler to break deadlocks. VTM [20] used agebased priorities while Waliullah et al [28] adapted a squashcount based priority scheme to avoid starvation in lazy conflict resolution HTMs. Bobba et al [2] identified performancedegrading interactions between workloads and the underlying TM design.…”

Section: Impact Of Lv* Configurationsmentioning

confidence: 99%

LV<sup>∗</sup>: A low complexity lazy versioning HTM infrastructure

Negi

Waliullah

Stenström

2010

2010 International Conference on Embedded Computer Systems: Architectures, Modeling and Simulation

Self Cite

View full text Add to dashboard Cite

Transactional memory (TM) promises to unlock parallelism in software in a safer and easier way than lock-based approaches but the path to deployment is unclear for several reasons. First of all, since TM has not been deployed in any machine yet, experience of using it is limited. While software transactional memory implementations exist, they are too slow to provide useful experience. Existing hardware transactional memory implementations, on the other hand, can provide the efficiency required but they require a significant effort to integrate in cache coherence infrastructures or freeze critical policy parameters. This paper proposes the LV* (lazy versioning and eager/lazy conflict resolution) class of hardware transactional memory protocols. This class of protocols has been implemented with ease of deployment in mind. LV* can be integrated with low additional complexity in standard snoopy-cache MESI-protocols and can be accommodatedin a directory-based cache coherence infrastructure. Since the optimal conflict resolution policy (lazy or eager) depends on transactional characteristics of workloads, LV* supports a set of conflict resolution policies that range from LazEr -a family of Lazy versioning Eager conflict resolution protocolsto LL-MESI which provides lazy resolution. We show that LV* can be hosted in a MESI protocol through straightforward extensions and that the flexibility in the choice of conflict resolution strategy has a significant impact on performance.

show abstract

“…Related work on Starvation-free STMs: Some researchers Gramoli et al [12], Waliullah and Stenstrom [13], Spear et al [14], Chaudhary et al [15] have explored starvationfreedom in RWSTMs. Most of them assigned priority to the transactions.…”

Section: Introductionmentioning

confidence: 99%

Achieving Starvation-Freedom with Greater Concurrency in Multi-Version Object-based Transactional Memory Systems

Juyal

Kulkarni

Kumari

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

To utilize the multi-core processors properly concurrent programming is needed. The main challenge is to design a correct and efficient concurrent program. Software Transactional Memory Systems (STMs) provide ease of multithreading to the programmer without worrying about concurrency issues as deadlock, livelock, priority inversion, etc. Most of the STMs work on read-write operations known as RWSTMs. Some STMs work at higher-level operations and ensure greater concurrency than RWSTMs. Such STMs are known as Single-Version Object-based STMs (SVOSTMs). The transactions of SVOSTMs can return commit or abort. Aborted SVOSTMs transactions retry. But in the current setting of SVOSTMs, transactions may starve. So, we propose a Starvation-Freedom in SVOSTM as SF-SVOSTM that satisfies the correctness criteria conflict-opacity. Databases and STMs say that maintaining multiple versions corresponding to each shared data-item (or key) reduces the number of aborts and improves the throughput. So, to achieve greater concurrency further, we propose Starvation-Freedom in Multi-Version OSTM as SF-MVOSTM algorithm. The number of versions maintains by SF-MVOSTM either be unbounded with garbage collection as SF-MVOSTM-GC or bounded with latest K-versions as SF-KOSTM. SF-MVOSTM satisfies the correctness criteria as local opacity and shows the performance benefits as compared with state-of-the-art STMs.

show abstract

Schemes for avoiding starvation in transactional memory systems

Cited by 7 publications

References 10 publications

Achieving Starvation-Freedom in Multi-version Transactional Memory Systems

Achieving Starvation-Freedom in Multi-version Transactional Memory Systems

LV<sup>∗</sup>: A low complexity lazy versioning HTM infrastructure

Achieving Starvation-Freedom with Greater Concurrency in Multi-Version Object-based Transactional Memory Systems

Contact Info

Product

Resources

About

Schemes for avoiding starvation in transactional memory systems

Cited by 7 publications

References 10 publications

Achieving Starvation-Freedom in Multi-version Transactional Memory Systems

Achieving Starvation-Freedom in Multi-version Transactional Memory Systems

LV<sup>&#x2217;</sup>: A low complexity lazy versioning HTM infrastructure

Achieving Starvation-Freedom with Greater Concurrency in Multi-Version Object-based Transactional Memory Systems

Contact Info

Product

Resources

About

LV<sup>∗</sup>: A low complexity lazy versioning HTM infrastructure