(Mis)understanding the NUMA memory system performance of multithreaded workloads

Majó, Zoltán; Groß, Thomas

doi:10.1109/iiswc.2013.6704666

Cited by 33 publications

(21 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Á threads on i .T 1 C T 2 C : : : C T k /%i processors in step 4 (lines [11][12][13][14] if Cores idle is larger than N i.…”

Section: Algorithmmentioning

confidence: 99%

“…There are four sources of data accesses on NUMA systems (i.e., local/remote cache access and local/remote dynamic random access memory (DRAM) access). In general, local DRAM access has the same latency as remote cache access [14]. For simplicity, we classify the type of data access into local node and remote node without regard to whether the data access hits on cache or memory.…”

Section: Modelmentioning

confidence: 99%

“…When a multi-threaded program (e.g., PARSEC benchmark [13]) is deployed to the NUMA system, it brings new challenges besides the two aforementioned problems because of inter-thread interactions across processors. Some work has concluded that the inter-thread data sharing has a MCCSDN15 SPECIAL ISSUE 5811 large influence on the performance of multi-threaded program on the NUMA multi-core system [14][15][16][17]. Moreover, parallel applications with inter-thread synchronization (e.g., barrier) require additional scheduling mechanisms to balance the time of the costly remote data accesses for threads.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

SymS: a symmetrical scheduler to improve multi‐threaded program performance on NUMA systems

Zhu

Jin

Liao

2015

Concurrency and Computation

View full text Add to dashboard Cite

SUMMARYThe nonuniform memory access (NUMA) architecture has been used extensively in data centers. Most of the previous works used single-threaded multiprogrammed workloads to study the performance of NUMA systems, which mainly focus on two classes of problems: resource contention and data locality. However, when running multi-threaded programs on NUMA systems, the critical thread of these programs significantly influences the system performance and brings new challenges that are different from those in a single-threaded situation. In particular, an additional scheduling scheme is desired to avoid the performance degradation caused by the critical thread of multi-threaded programs running on NUMA systems. This work presents a scheduler, Symmetrical Scheduler, which successfully solves the lagging problem by balancing the number of the costly remote shared data accesses for threads on NUMA systems. To the best of our knowledge, little work has been conducted to examine the performance impacted by the critical thread of multi-threaded programs on NUMA systems. By running the PARSEC benchmark on such systems, our methodology can improve the program performance by a factor of 6% on average and achieve maximally 25.3% improvement compared with Linux kernel scheduling mechanism.

show abstract

“…Á threads on i .T 1 C T 2 C : : : C T k /%i processors in step 4 (lines [11][12][13][14] if Cores idle is larger than N i.…”

Section: Algorithmmentioning

confidence: 99%

Section: Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

SymS: a symmetrical scheduler to improve multi‐threaded program performance on NUMA systems

Zhu

Jin

Liao

2015

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…Afterwards, instrumentation code to perform data mapping is inserted into the application. Other proposals suggest changing the application source code to optimize it for NUMA systems [4].…”

Section: Related Workmentioning

confidence: 99%

“…Locality is mostly important when the memory access time to access a remote NUMA node is much higher than an access to a local node. However, when the difference between local and remote accesses is low, or the overhead to access a remote node is hidden by cache memories or cache line prefetchers [4], the influence of memory locality becomes lower. In these cases, another important factor to be considered is the load of the memory controllers.…”

Section: Introductionmentioning

confidence: 99%

Locality vs. Balance: Exploring Data Mapping Policies on NUMA Systems

Diener

Cruz

Navaux

2015

2015 23rd Euromicro International Conference on Parallel, Distributed, and Network-Based Processing

View full text Add to dashboard Cite

In parallel architectures that have a Non-Uniform Memory Access (NUMA) behavior, the mapping of memory pages to NUMA nodes influences the performance of parallel applications. In order to improve traditional data mapping policies, two basic strategies can be employed: optimizing locality or balance of memory accesses. In a locality-based policy, memory pages are mapped to nodes that access the page the most. In a balance-based policy, memory pages are mapped such that the number of memory accesses resolved by each memory controller is similar.In this paper, we perform an in-depth exploration of these data mapping policies on the performance of parallel applications. We introduce metrics that describe their memory access behavior and evaluate their suitability for data mapping. We also present new mapping policies that focus on locality, balance or both. These policies were evaluated on three different NUMA architectures with applications from the NAS-OMP and PARSEC benchmark suites. Results show that the performance improvements of each policy depend on the characteristics of the applications and machines. Choosing the wrong policy can actually hurt the performance compared to the default first-touch mapping. Compared to traditional mapping policies and to policies that only focus on either locality or balance, taking into account both locality and balance results in the highest improvements. Furthermore, it avoids the performance reduction caused by the wrong data mapping.

show abstract

PerfMemPlus: A Tool for Automatic Discovery of Memory Performance Problems

Helm

Taura

2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

(Mis)understanding the NUMA memory system performance of multithreaded workloads

Cited by 33 publications

References 18 publications

SymS: a symmetrical scheduler to improve multi‐threaded program performance on NUMA systems

SymS: a symmetrical scheduler to improve multi‐threaded program performance on NUMA systems

Locality vs. Balance: Exploring Data Mapping Policies on NUMA Systems

PerfMemPlus: A Tool for Automatic Discovery of Memory Performance Problems

Contact Info

Product

Resources

About