DRAM Main Memory Research Articles

Boosting access parallelism to PCM-based main memory

Despite its promise as a DRAM main memory replacement, Phase Change Memory (PCM) has high write latencies which can be a serious detriment to its widespread adoption. Apart from slowing down a write request, the consequent high latency can also keep other chips of the same rank, that are not involved in this write, idle for long times. There are several practical considerations that make it difficult to allow subsequent reads and/or writes to be served concurrently from the same chips during the long latency write. This paper proposes and evaluates several novel mechanisms -- re-constructing data from error correction bits instead of waiting for chips currently busy to serve a read, rotating word mappings across chips of a PCM rank, and rotating the mapping of error detection/correction bits across these chips -- to overlap several reads with an ongoing write (RoW) and even a write with an ongoing write (WoW). The paper also presents the necessary micro-architectural enhancements nee-ded to implement these mechanisms, without significantly changing the current interfaces. The resulting PCM access parallelism (PCMap) system incorporating these enhancements, boosts the intra-rank-level parallelism during such writes from a very low baseline value of 2.4 to an average and maximum values of 4.5 and 7.4, respectively (out of a maximum of 8.0), across a wide spectrum of both multiprogrammed and multithreaded workloads. This boost in parallelism results in an average IPC improvement of 15.6% and 16.7% for the multi-progra-mmed and multi-threaded workloads, respectively.

Asteroid: Scalable Online Memory Diagnostics for Multi-core, Multi-socket Servers

Memory diagnostics are important to improving the resilience of DRAM main memory. As bit cell size reaches physical limits, DRAM memory will be more likely to suffer both transient and permanent errors. Memory diagnostics that operate online can be a component of a comprehensive strategy to allay errors. This paper presents a novel approach, Asteroid, to integrate online memory diagnostics during workload execution. The approach supports diagnostics that adapt at runtime to workload behavior and resource availability to maximize test quality while reducing performance overhead. We describe Asteroid’s design and how it can be efficiently integrated with a hierarchical memory allocator in modern operating systems. We also present how the framework enables control policies to dynamically configure a diagnostic. Using an adaptive policy, in a 16-core server, Asteroid has modest overhead of 1–4 % for workloads with low to high memory demand. For these workloads, Asteroid’s adaptive policy has good error coverage and can thoroughly test memory.

Opportunities for Nonvolatile Memory Systems in Extreme-Scale High-Performance Computing

For extreme-scale high-performance computing systems, system-wide power consumption has been identified as one of the key constraints moving forward, where DRAM main memory systems account for about 30 to 50 percent of a node's overall power consumption. As the benefits of device scaling for DRAM memory slow, it will become increasingly difficult to keep memory capacities balanced with increasing computational rates offered by next-generation processors. However, several emerging memory technologies related to nonvolatile memory (NVM) devices are being investigated as an alternative for DRAM. Moving forward, NVM devices could offer solutions for HPC architectures. Researchers are investigating how to integrate these emerging technologies into future extreme-scale HPC systems and how to expose these capabilities in the software stack and applications. Current results show several of these strategies could offer high-bandwidth I/O, larger main memory capacities, persistent data structures, and new approaches for application resilience and output postprocessing, such as transaction-based incremental checkpointing and in situ visualization, respectively.

CIDR: A Cache Inspired Area-Efficient DRAM Resilience Architecture against Permanent Faults

Faulty cells have become major problems in cost-sensitive main-memory DRAM devices. Conventional solutions to reduce device failure rates due to cells with permanent faults, such as populating spare rows and relying on errorcorrecting codes, have had limited success due to high area overheads. In this paper, we propose CIDR, a novel cache-inspired DRAM resilience architecture, which substantially reduces the area overhead of handling bit errors from these faulty cells. A DRAM device adopting CIDR has a small cache next to its I/O pads to replace accesses to the addresses that include the faulty cells with ones that correspond to the cache data array. We minimize the energy overhead of accessing the cache tags for every read or write by adding a Bloom filter in front of the cache. The augmented cache is programmed once during the testing phase and is out of the critical path on normal accesses because both cache and DRAM arrays are accessed in parallel, making CIDR transparent to existing processor-memory interfaces. Compared to the conventional architecture relying on spare rows, CIDR lowers the area overhead of achieving equal failure rates over a wide range of single-bit error rates, such as 23.6× lower area overhead for a bit-error rate of 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-5</sup> and a device failure rate of 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-3</sup> .

Data Classification Management with its Interfacing Structure for Hybrid SLC/MLC PRAM Main Memory

This research aims to design a new phase-change RAM (PRAM)-based main memory structure, supporting the advantages of PRAM while providing performance similar to that of conventional DRAM main memory. To replace conventional DRAMs with non-volatile PRAMs as the main memory components, comparable memory access latency, overall cost, power dissipation, and memory cell endurance should be supported. For these goals, we propose a new main memory system consisting of a DRAM converter and an array of single-level cell (SLC)/multi-level cell (MLC) PRAMs. The DRAM converter consists of an aggressive fetching superblock buffer to assure better use of spatial locality and a selective filtering buffer for better use of temporal locality. The array of the SLC/MLC hybrid PRAM structure includes a combination of SLC and MLC PRAMs to enhance the lifetime of the MLC PRAM main memory and hide asymmetric read/write access latency. The proposed structure is evaluated by a trace-driven simulator using SPEC CPU 2006 and SPLASH-2 traces. Experimental results show that the proposed DRAM converter can reduce the miss rate by ∼37% and write count by ∼55% in comparison with the uniform buffer case. Also, the SLC/MLC PRAM with DRAM converter shows performance in terms of access latency and power consumption close to that of the conventional memory architecture. Thus, our proposed memory architecture can be used to replace the current DRAM-based main memory system.

PMSS: A programmable memory system and scheduler for complex memory patterns

HPC industry demands more computing units on FPGAs, to enhance the performance by using task/data parallelism. FPGAs can provide its ultimate performance on certain kernels by customizing the hardware for the applications. However, applications are getting more complex, with multiple kernels and complex data arrangements, generating overhead while scheduling/managing system resources. Due to this reason all classes of multi threaded machines–minicomputer to supercomputer–require to have efficient hardware scheduler and memory manager that improves the effective bandwidth and latency of the DRAM main memory. This architecture could be a very competitive choice for supercomputing systems that meets the demand of parallelism for HPC benchmarks. In this article, we proposed a Programmable Memory System and Scheduler (PMSS), which provides high speed complex data access pattern to the multi threaded architecture. This proposed PMSS system is implemented and tested on a Xilinx ML505 evaluation FPGA board. The performance of the system is compared with a microprocessor based system that has been integrated with the Xilkernel operating system. Results show that the modified PMSS based multi-accelerator system consumes 50% less hardware resources, 32% less on-chip power and achieves approximately a 19x speedup compared to the MicroBlaze based system.

Adaptive wear-leveling algorithm for PRAM main memory with a DRAM buffer

Phase Change RAM (PRAM) is a candidate to replace DRAM main memory due to its low idle power consumption and high scalability. However, its latency and endurance have generated problems in fulfilling its main memory role. The latency can be treated with a DRAM buffer, but the endurance problem remains, with three critical points that need to be improved despite the use of, existing wear-leveling algorithms. First, existing DRAM buffering schemes do not consider write count distribution. Second, swapping and shifting operations are performed statically. Finally, swapping and shifting operations are loosely coupled with a DRAM buffer. As a remedy to these drawbacks, we propose an adaptive wear-leveling algorithm that consists of three novel schemes for PRAM main memory with a DRAM buffer. The PRAM-aware DRAM buffering scheme reduces the write count and prevents skewed writing by considering the write count and clean data based on the least recently used (LRU) scheme. The adaptive multiple swapping and shifting scheme makes the write count even with the dynamic operation timing, the number of swapping pages being based on the workload pattern. Our DRAM buffer-aware swapping and shifting scheme reduces overhead by curbing additional swapping and shifting operations, thus reducing unnecessary write operations. To evaluate the wear-leveling effect, we have implemented a PIN-based wear-leveling simulator. The evaluation confirms that the PRAM lifetime increases from 0.68 years with the previous wear-leveling algorithm to 5.32 years with the adaptive wear-leveling algorithm.

A circuit-architecture co-optimization framework for exploring nonvolatile memory hierarchies

Many new memory technologies are available for building future energy-efficient memory hierarchies. It is necessary to have a framework that can quickly find the optimal memory technology at each hierarchy level. In this work, we first build a circuit-architecture joint design space exploration framework by combining RC circuit analysis and Artificial Neural Network (ANN)-based performance modeling. Then, we use this framework to evaluate some emerging nonvolatile memory hierarchies. We demonstrate that a Resistive RAM (ReRAM)-based cache hierarchy on an 8-core Chip-Multiprocessor (CMP) system can achieve a 24% Energy Delay Product (EDP) improvement and a 36% Energy Delay Area Product (EDAP) improvement compared to a conventional hierarchy with SRAM on-chip caches and DRAM main memory.

High-performance and low-energy buffer mapping method for multiprocessor DSP systems

When implementing digital signal processing (DSP) applications onto multiprocessor systems, one significant problem in the viewpoints of performance is the memory wall. In this paper, to help alleviate the memory wall problem, we propose a novel, high-performance buffer mapping policy for SDF-represented DSP applications on bus-based multiprocessor systems that support the shared-memory programming model. The proposed policy exploits the bank concurrency of the DRAM main memory system according to the analysis of hierarchical parallelism. Energy consumption is also a critical parameter, especially in battery-based embedded computing systems. In this paper, we apply a synchronization back-off scheme on the top of the proposed high-performance buffer mapping policy to reduce energy consumption. The energy saving is attained by minimizing the number of non-essential synchronization transactions. We measure throughput and energy consumption on both synthetic and real benchmarks. The simulation results show that the proposed buffer mapping policy is very useful in terms of performance, especially in memory-intensive applications where the total execution time of computational tasks is relatively small compared to that of memory operations. In addition, the proposed synchronization back-off scheme provides a reduction in the number of synchronization transactions without degrading performance, which results in system energy saving.

Re-architecting DRAM memory systems with monolithically integrated silicon photonics

The performance of future manycore processors will only scale with the number of integrated cores if there is a corresponding increase in memory bandwidth. Projected scaling of electrical DRAM architectures appears unlikely to suffice, being constrained by processor and DRAM pin-bandwidth density and by total DRAM chip power, including off-chip signaling, cross-chip interconnect, and bank access energy. In this work, we redesign the DRAM main memory system using a proposed monolithically integrated silicon photonics technology and show that our photonically interconnected DRAM (PIDRAM) provides a promising solution to all of these issues. Photonics can provide high aggregate pin-bandwidth density through dense wavelength-division multiplexing. Photonic signaling provides energy-efficient communication, which we exploit to not only reduce chip-to-chip interconnect power but to also reduce cross-chip interconnect power by extending the photonic links deep into the actual PIDRAM chips. To complement these large improvements in interconnect bandwidth and power, we decrease the number of bits activated per bank to improve the energy efficiency of the PIDRAM banks themselves. Our most promising design point yields approximately a 10x power reduction for a single-chip PIDRAM channel with similar throughput and area as a projected future electrical-only DRAM. Finally, we propose optical power guiding as a new technique that allows a single PIDRAM chip design to be used efficiently in several multi-chip configurations that provide either increased aggregate capacity or bandwidth.

A durable and energy efficient main memory using phase change memory technology

Using nonvolatile memories in memory hierarchy has been investigated to reduce its energy consumption because nonvolatile memories consume zero leakage power in memory cells. One of the difficulties is, however, that the endurance of most nonvolatile memory technologies is much shorter than the conventional SRAM and DRAM technology. This has limited its usage to only the low levels of a memory hierarchy, e.g., disks, that is far from the CPU. In this paper, we study the use of a new type of nonvolatile memories -- the Phase Change Memory (PCM) as the main memory for a 3D stacked chip. The main challenges we face are the limited PCM endurance, longer access latencies, and higher dynamic power compared to the conventional DRAM technology. We propose techniques to extend the endurance of the PCM to an average of 13 (for MLC PCM cell) to 22 (for SLC PCM) years. We also study the design choices of implementing PCM to achieve the best tradeoff between energy and performance. Our design reduced the total energy of an already low-power DRAM main memory of the same capacity by 65%, and energy-delay 2 product by 60%. These results indicate that it is feasible to use PCM technology in place of DRAM in the main memory for better energy efficiency.

A Comprehensive Memory Modeling Tool and Its Application to the Design and Analysis of Future Memory Hierarchies

In this paper we introduce CACTI-D, a significant enhancement of CACTI 5.0. CACTI-D adds support for modeling of commodity DRAM technology and support for main memory DRAM chip organization. CACTI-D enables modeling of the complete memory hierarchy with consistent models all the way from SRAM based L1 caches through main memory DRAMs on DIMMs. We illustrate the potential applicability of CACTI-D in the design and analysis of future memory hierarchies by carrying out a last level cache study for a multicore multithreaded architecture at the 32nm technology node. In this study we use CACTI-D to model all components of the memory hierarchy including L1, L2, last level SRAM, logic process based DRAM or commodity DRAM L3 caches, and main memory DRAM chips. We carry out architectural simulation using benchmarks with large data sets and present results of their execution time, breakdown of power in the memory hierarchy, and system energy-delay product for the different system configurations. We find that commodity DRAM technology is most attractive for stacked last level caches, with significantly lower energy-delay products.

The case for SRAM main memory

The growing CPU-memory gap is resulting in increasingly large cache sizes. As cache sizes increase, associativity becomes less of a win. At the same time, since costs of going to DRAM increase, it becomes more valuable to be able to pin critical data in the cache---a problem if a cache is direct-mapped or has a low degree of associativity. Something else which is a problem for caches of low associativity is reducing misses by using a better replacement policy. This paper proposes that L2 cache sizes are now starting to reach the point where it makes more sense to manage them as the main memory of the computer, and relegate the traditional DRAM main memory to the role of a paging device. The paper details advantages of an SRAM main memory, as well as problems that need to be solved, in managing an extra level of virtual to physical translation.