ATLAS: A scalable and high-performance scheduling algorithm for multiple memory controllers
01 Apr 2010-pp 1-12
TL;DR: It is shown that the implementation of least-attained-service thread prioritization reduces the time the cores spend stalling and significantly improves system throughput, and ATLAS's performance benefit increases as the number of cores increases.
Abstract: Modern chip multiprocessor (CMP) systems employ multiple memory controllers to control access to main memory. The scheduling algorithm employed by these memory controllers has a significant effect on system throughput, so choosing an efficient scheduling algorithm is important. The scheduling algorithm also needs to be scalable — as the number of cores increases, the number of memory controllers shared by the cores should also increase to provide sufficient bandwidth to feed the cores. Unfortunately, previous memory scheduling algorithms are inefficient with respect to system throughput and/or are designed for a single memory controller and do not scale well to multiple memory controllers, requiring significant finegrained coordination among controllers. This paper proposes ATLAS (Adaptive per-Thread Least-Attained-Service memory scheduling), a fundamentally new memory scheduling technique that improves system throughput without requiring significant coordination among memory controllers. The key idea is to periodically order threads based on the service they have attained from the memory controllers so far, and prioritize those threads that have attained the least service over others in each period. The idea of favoring threads with least-attained-service is borrowed from the queueing theory literature, where, in the context of a single-server queue it is known that least-attained-service optimally schedules jobs, assuming a Pareto (or any decreasing hazard rate) workload distribution. After verifying that our workloads have this characteristic, we show that our implementation of least-attained-service thread prioritization reduces the time the cores spend stalling and significantly improves system throughput. Furthermore, since the periods over which we accumulate the attained service are long, the controllers coordinate very infrequently to form the ordering of threads, thereby making ATLAS scalable to many controllers. We evaluate ATLAS on a wide variety of multiprogrammed SPEC 2006 workloads and systems with 4–32 cores and 1–16 memory controllers, and compare its performance to five previously proposed scheduling algorithms. Averaged over 32 workloads on a 24-core system with 4 controllers, ATLAS improves instruction throughput by 10.8%, and system throughput by 8.4%, compared to PAR-BS, the best previous CMP memory scheduling algorithm. ATLAS's performance benefit increases as the number of cores increases.
Citations
More filters
09 Jun 2012
TL;DR: This paper proposes RAIDR (Retention-Aware Intelligent DRAM Refresh), a low-cost mechanism that can identify and skip unnecessary refreshes using knowledge of cell retention times and group DRAM rows into retention time bins and apply a different refresh rate to each bin.
Abstract: Dynamic random-access memory (DRAM) is the building block of modern main memory systems. DRAM cells must be periodically refreshed to prevent loss of data. These refresh operations waste energy and degrade system performance by interfering with memory accesses. The negative effects of DRAM refresh increase as DRAM device capacity increases. Existing DRAM devices refresh all cells at a rate determined by the leakiest cell in the device. However, most DRAM cells can retain data for significantly longer. Therefore, many of these refreshes are unnecessary. In this paper, we propose RAIDR (Retention-Aware Intelligent DRAM Refresh), a low-cost mechanism that can identify and skip unnecessary refreshes using knowledge of cell retention times. Our key idea is to group DRAM rows into retention time bins and apply a different refresh rate to each bin. As a result, rows containing leaky cells are refreshed as frequently as normal, while most rows are refreshed less frequently. RAIDR uses Bloom filters to efficiently implement retention time bins. RAIDR requires no modification to DRAM and minimal modification to the memory controller. In an 8-core system with 32 GB DRAM, RAIDR achieves a 74.6% refresh reduction, an average DRAM power reduction of 16.1%, and an average system performance improvement of 8.6% over existing systems, at a modest storage overhead of 1.25 KB in the memory controller. RAIDR's benefits are robust to variation in DRAM system configuration, and increase as memory capacity increases.
520 citations
Cites methods from "ATLAS: A scalable and high-performa..."
...13 For these results we only evaluated the 32 workloads with 50% memory-intensive benchmarks, as this scenario of balanced memory-intensive and non-memory-intensive benchmarks is likely to be common in future systems [22]....
[...]
07 Dec 2013
TL;DR: RowClone is proposed, a new and simple mechanism to perform bulk copy and initialization completely within DRAM — eliminating the need to transfer any data over the memory channel to perform such operations.
Abstract: Several system-level operations trigger bulk data copy or initialization. Even though these bulk data operations do not require any computation, current systems transfer a large quantity of data back and forth on the memory channel to perform such operations. As a result, bulk data operations consume high latency, bandwidth, and energy — degrading both system performance and energy efficiency. In this work, we propose RowClone, a new and simple mechanism to perform bulk copy and initialization completely within DRAM — eliminating the need to transfer any data over the memory channel to perform such operations. Our key observation is that DRAM can internally and efficiently transfer a large quantity of data (multiple KBs) between a row of DRAM cells and the associated row buffer. Based on this, our primary mechanism can quickly copy an entire row of data from a source row to a destination row by first copying the data from the source row to the row buffer and then from the row buffer to the destination row, via two back-to-back activate commands. This mechanism, which we call the Fast Parallel Mode of RowClone, reduces the latency and energy consumption of a 4KB bulk copy operation by 11.6× and 74.4×, respectively, and a 4KB bulk zeroing operation by 6.0× and 41.5×, respectively. To efficiently copy data between rows that do not share a row buffer, we propose a second mode of RowClone, the Pipelined Serial Mode, which uses the shared internal bus of a DRAM chip to quickly copy data between two banks. RowClone requires only a 0.01% increase in DRAM chip area. We quantitatively evaluate the benefits of RowClone by focusing on fork, one of the frequently invoked system calls, and five other copy and initialization intensive applications. Our results show that RowClone can significantly improve both single-core and multi-core system performance, while also significantly reducing main memory bandwidth and energy consumption.
385 citations
Additional excerpts
...Maximum Slowdown [12, 29, 30] Reduction 6% 12% 23%...
[...]
...NumberofCores 2 4 8 NumberofWorkloads 138 50 40 WeightedSpeedup[14]Improvement 15% 20% 27% InstructionThroughputImprovement 14% 15% 25% HarmonicSpeedup[35]Improvement 13% 16% 29% MaximumSlowdown[12,29,30]Reduction 6% 12% 23% MemoryBandwidth/Instruction[49]Reduction 29% 27% 28% MemoryEnergy/InstructionReduction 19% 17% 17% Table 7:Effect ofRowClone onmulti-core performance, fairness,bandwidth,andenergy To provide more insightintothebenefitsofRowClone on multi-core systems, we classify our copy/initializationintensive benchmarks into two categories: 1) Moderately copy/initialization-intensive (compile,mcached,andmysql)and highlycopy/initialization-intensive(bootup,forkbench,andshell)....
[...]
04 Dec 2010
TL;DR: This paper presents a new memory scheduling algorithm that addresses system throughput and fairness separately with the goal of achieving the best of both, and evaluates TCM on a wide variety of multiprogrammed workloads and compares its performance to four previously proposed scheduling algorithms, finding that TCM achieves both the best system throughputand fairness.
Abstract: In a modern chip-multiprocessor system, memory is a shared resource among multiple concurrently executing threads. The memory scheduling algorithm should resolve memory contention by arbitrating memory access in such a way that competing threads progress at a relatively fast and even pace, resulting in high system throughput and fairness. Previously proposed memory scheduling algorithms are predominantly optimized for only one of these objectives: no scheduling algorithm provides the best system throughput and best fairness at the same time. This paper presents a new memory scheduling algorithm that addresses system throughput and fairness separately with the goal of achieving the best of both. The main idea is to divide threads into two separate clusters and employ different memory request scheduling policies in each cluster. Our proposal, Thread Cluster Memory scheduling (TCM), dynamically groups threads with similar memory access behavior into either the latency-sensitive (memory-non-intensive) or the bandwidth-sensitive (memory-intensive) cluster. TCM introduces three major ideas for prioritization: 1) we prioritize the latency-sensitive cluster over the bandwidth-sensitive cluster to improve system throughput, 2) we introduce a ``niceness'' metric that captures a thread's propensity to interfere with other threads, 3) we use niceness to periodically shuffle the priority order of the threads in the bandwidth-sensitive cluster to provide fair access to each thread in a way that reduces inter-thread interference. On the one hand, prioritizing memory-non-intensive threads significantly improves system throughput without degrading fairness, because such ``light'' threads only use a small fraction of the total available memory bandwidth. On the other hand, shuffling the priority order of memory-intensive threads improves fairness because it ensures no thread is disproportionately slowed down or starved. We evaluate TCM on a wide variety of multiprogrammed workloads and compare its performance to four previously proposed scheduling algorithms, finding that TCM achieves both the best system throughput and fairness. Averaged over 96 workloads on a 24-core system with 4 memory channels, TCM improves system throughput and reduces maximum slowdown by 4.6%/38.6% compared to ATLAS (previous work providing the best system throughput) and 7.6%/4.6% compared to PAR-BS (previous work providing the best fairness).
375 citations
Cites background or methods from "ATLAS: A scalable and high-performa..."
...If there are multiple memory controllers, this information is sent to a centralized meta-controller at the end of a quantum, similarly to what is done in ATLAS [5]....
[...]
...Grouping of threads into clusters happens in a synchronized manner across all memory controllers to better exploit bank-level parallelism [5, 14]....
[...]
...Compared to ATLAS [5], the best previous algorithm in terms of system throughput, TCM improves system throughput and reduces maximum slowdown by 4....
[...]
..., by forming a priority order based on a metric that corresponds to memory intensity, as done in [5])....
[...]
...Finally, as previous work has shown, it is desirable that scheduling decisions are made in a synchronized manner across all banks [5, 14, 12], so that concurrent requests of each thread are serviced in parallel, without being serialized due to interference from other threads....
[...]
01 Jan 2010
TL;DR: TCM as discussed by the authors dynamically groups threads with similar memory access behavior into either the latency-sensitive (memory non-intensive) or the bandwidth-intensive (memory intensive) clusters, and introduces a ''niceness'' metric that captures a thread's propensity to interfere with other threads.
Abstract: In a modern chip-multiprocessor system, memory is a shared resource among multiple concurrently executing threads. The memory scheduling algorithm should resolve memory contention by arbitrating memory access in such a way that competing threads progress at a relatively fast and even pace, resulting in high system throughput and fairness. Previously proposed memory scheduling algorithms are predominantly optimized for only one of these objectives: no scheduling algorithm provides the best system throughput and best fairness at the same time. This paper presents a new memory scheduling algorithm that addresses system throughput and fairness separately with the goal of achieving the best of both. The main idea is to divide threads into two separate clusters and employ different memory request scheduling policies in each cluster. Our proposal, Thread Cluster Memory scheduling (TCM), dynamically groups threads with similar memory access behavior into either the latency-sensitive (memory-non-intensive) or the bandwidth-sensitive (memory-intensive) cluster. TCM introduces three major ideas for prioritization: 1) we prioritize the latency-sensitive cluster over the bandwidth-sensitive cluster to improve system throughput, 2) we introduce a ``niceness'' metric that captures a thread's propensity to interfere with other threads, 3) we use niceness to periodically shuffle the priority order of the threads in the bandwidth-sensitive cluster to provide fair access to each thread in a way that reduces inter-thread interference. On the one hand, prioritizing memory-non-intensive threads significantly improves system throughput without degrading fairness, because such ``light'' threads only use a small fraction of the total available memory bandwidth. On the other hand, shuffling the priority order of memory-intensive threads improves fairness because it ensures no thread is disproportionately slowed down or starved. We evaluate TCM on a wide variety of multiprogrammed workloads and compare its performance to four previously proposed scheduling algorithms, finding that TCM achieves both the best system throughput and fairness. Averaged over 96 workloads on a 24-core system with 4 memory channels, TCM improves system throughput and reduces maximum slowdown by 4.6%/38.6% compared to ATLAS (previous work providing the best system throughput) and 7.6%/4.6% compared to PAR-BS (previous work providing the best fairness).
358 citations
09 Jun 2012
TL;DR: Three new mechanisms (SALP-1, SALP-2, and MASA) mitigate the negative impact of bank serialization by overlapping different components of the bank access latencies of multiple requests that go to different subarrays within the same bank.
Abstract: Modern DRAMs have multiple banks to serve multiple memory requests in parallel. However, when two requests go to the same bank, they have to be served serially, exacerbating the high latency of off-chip memory. Adding more banks to the system to mitigate this problem incurs high system cost. Our goal in this work is to achieve the benefits of increasing the number of banks with a low cost approach. To this end, we propose three new mechanisms that overlap the latencies of different requests that go to the same bank. The key observation exploited by our mechanisms is that a modern DRAM bank is implemented as a collection of subarrays that operate largely independently while sharing few global peripheral structures. Our proposed mechanisms (SALP-1, SALP-2, and MASA) mitigate the negative impact of bank serialization by overlapping different components of the bank access latencies of multiple requests that go to different subarrays within the same bank. SALP-1 requires no changes to the existing DRAM structure and only needs reinterpretation of some DRAM timing parameters. SALP-2 and MASA require only modest changes (< 0.15% area overhead) to the DRAM peripheral structures, which are much less design constrained than the DRAM core. Evaluations show that all our schemes significantly improve performance for both single-core systems and multi-core systems. Our schemes also interact positively with application-aware memory request scheduling in multi-core systems.
338 citations
References
More filters
12 Jun 2005
TL;DR: The goals are to provide easy-to-use, portable, transparent, and efficient instrumentation, and to illustrate Pin's versatility, two Pintools in daily use to analyze production software are described.
Abstract: Robust and powerful software instrumentation tools are essential for program analysis tasks such as profiling, performance evaluation, and bug detection. To meet this need, we have developed a new instrumentation system called Pin. Our goals are to provide easy-to-use, portable, transparent, and efficient instrumentation. Instrumentation tools (called Pintools) are written in C/C++ using Pin's rich API. Pin follows the model of ATOM, allowing the tool writer to analyze an application at the instruction level without the need for detailed knowledge of the underlying instruction set. The API is designed to be architecture independent whenever possible, making Pintools source compatible across different architectures. However, a Pintool can access architecture-specific details when necessary. Instrumentation with Pin is mostly transparent as the application and Pintool observe the application's original, uninstrumented behavior. Pin uses dynamic compilation to instrument executables while they are running. For efficiency, Pin uses several techniques, including inlining, register re-allocation, liveness analysis, and instruction scheduling to optimize instrumentation. This fully automated approach delivers significantly better instrumentation performance than similar tools. For example, Pin is 3.3x faster than Valgrind and 2x faster than DynamoRIO for basic-block counting. To illustrate Pin's versatility, we describe two Pintools in daily use to analyze production software. Pin is publicly available for Linux platforms on four architectures: IA32 (32-bit x86), EM64T (64-bit x86), Itanium®, and ARM. In the ten months since Pin 2 was released in July 2004, there have been over 3000 downloads from its website.
4,019 citations
TL;DR: It is found that user-initiated TCP session arrivals, such as remote-login and file-transfer, are well-modeled as Poisson processes with fixed hourly rates, but that other connection arrivals deviate considerably from Poisson.
Abstract: Network arrivals are often modeled as Poisson processes for analytic simplicity, even though a number of traffic studies have shown that packet interarrivals are not exponentially distributed. We evaluate 24 wide area traces, investigating a number of wide area TCP arrival processes (session and connection arrivals, FTP data connection arrivals within FTP sessions, and TELNET packet arrivals) to determine the error introduced by modeling them using Poisson processes. We find that user-initiated TCP session arrivals, such as remote-login and file-transfer, are well-modeled as Poisson processes with fixed hourly rates, but that other connection arrivals deviate considerably from Poisson; that modeling TELNET packet interarrivals as exponential grievously underestimates the burstiness of TELNET traffic, but using the empirical Tcplib interarrivals preserves burstiness over many time scales; and that FTP data connection arrivals within FTP sessions come bunched into "connection bursts", the largest of which are so large that they completely dominate FTP data traffic. Finally, we offer some results regarding how our findings relate to the possible self-similarity of wide area traffic. >
3,915 citations
TL;DR: Content-Centric Networking (CCN) is presented which uses content chunks as a primitive---decoupling location from identity, security and access, and retrieving chunks of content by name, and simultaneously achieves scalability, security, and performance.
Abstract: Current network use is dominated by content distribution and retrieval yet current networking protocols are designed for conversations between hosts. Accessing content and services requires mapping from the what that users care about to the network's where. We present Content-Centric Networking (CCN) which uses content chunks as a primitive---decoupling location from identity, security and access, and retrieving chunks of content by name. Using new approaches to routing named content, derived from IP, CCN simultaneously achieves scalability, security, and performance. We describe our implementation of the architecture's basic features and demonstrate its performance and resilience with secure file downloads and VoIP calls.
3,122 citations
"ATLAS: A scalable and high-performa..." refers background or methods in this paper
...propose deployable modifications to IP to enhance its evolvability [37]; but does not admit the expressiveness afforded by XIA....
[...]
...RPT over CCN: RPT also can be integrated with a broad class of content-aware networks, including CCN [37] and SmartRE [15]....
[...]
...CCN is designed to operate on top of unreliable packet delivery service, and thus Interest and Data packets may be lost [37]....
[...]
...We use the term contentaware networks to refer to the variety of architectural proposals [14, 32, 37, 42] and devices [2, 4, 10, 35] that cache data and remove duplicates to alleviate congestion (i....
[...]
...In the CCN-over-jumbo-UDP protocol case [37], a client generates three Interest packets (325 bytes) and receives five 1500-byte packets (6873 bytes) to fetch a Web page [37]....
[...]
TL;DR: In this article, a fair gateway queueing algorithm based on an earlier suggestion by Nagle is proposed to control congestion in datagram networks, based on the idea of fair queueing.
Abstract: We discuss gateway queueing algorithms and their role in controlling congestion in datagram networks. A fair queueing algorithm, based on an earlier suggestion by Nagle, is proposed. Analysis and s...
2,639 citations
TL;DR: It is shown that the self-similarity in WWW traffic can be explained based on the underlying distributions of WWW document sizes, the effects of caching and user preference in file transfer, the effect of user "think time", and the superimposition of many such transfers in a local-area network.
Abstract: The notion of self-similarity has been shown to apply to wide-area and local-area network traffic. We show evidence that the subset of network traffic that is due to World Wide Web (WWW) transfers can show characteristics that are consistent with self-similarity, and we present a hypothesized explanation for that self-similarity. Using a set of traces of actual user executions of NCSA Mosaic, we examine the dependence structure of WWW traffic. First, we show evidence that WWW traffic exhibits behavior that is consistent with self-similar traffic models. Then we show that the self-similarity in such traffic can be explained based on the underlying distributions of WWW document sizes, the effects of caching and user preference in file transfer, the effect of user "think time", and the superimposition of many such transfers in a local-area network. To do this, we rely on empirically measured distributions both from client traces and from data independently collected at WWW servers.
2,608 citations