Data-centric applications are increasingly more common, causing issues brought on by the discrepancy between processor and memory technologies to be increasingly more apparent. Near-Data Processing (NDP) is an approach to mitigate this issue. It proposes moving some of the computation close to the memory, thus allowing for reduced data movement and aiding data-intensive workloads. Analytical database queries are very commonly used in NDP research due to their intrinsics usage of very large volumes of data. In this paper, we investigate the migration of most time-consuming database operators to VIMA, a novel 3D-stacked memory-based NDP architecture. We consider the selection, projection, and bloom join database query operators, commonly used by data analytics applications, comparing Vector-In-Memory Architecture (VIMA) to a high-performance x86 baseline. We pitch VIMA against both a single-thread baseline and a modern 16-thread x86 system to evaluate its performance. Against a single-thread baseline, our experiments show that VIMA is able to speed up execution by up to 5× for selection, 2.5× for projection, and 16× for join while consuming up to 99% less energy. When considering a multi-thread baseline, VIMA matches the execution time performance even at the largest dataset sizes considered. In comparison to existing state-of-the-art NDP platforms, we find that our approach achieves superior performance for these operators.

Near-Data Processing (NDP) modifies the traditional computer system design by placing logic near the memory, bringing computation to the data. One NDP approach places such elements on the logic layer of 3D-stacked memories to quickly access data while avoiding reliance on narrow buses and better accessing the parallelism these devices offer. However, NDP architectures often fail to fully leverage available memory resources. In this work, we propose adding an instruction buffer to a common NDP design with large vector instructions. This modification allows the NDP to fetch instruction operands out of program order and delegates some responsibility regarding precise exceptions to the near-data device. Our results show our modifications cause a reduction in execution time of up to 28% while consuming up to 25% less energy.

As applications become more data-intensive, issues like von Neumann’s bottleneck and the memory wall became more apparent since data movement is the main source of inefficiency in computer systems. Looking to mitigate this issue, Near-Data Processing (NDP) moves computation from the processor to the memory, thus reducing the data movement required by many data-intensive workloads. In this paper, we look to database query operators, common targets of NDP research as database systems often need to deal with large amounts of data. We investigate the migration of most time-consuming database operators to Vector-In-Memory Architecture (VIMA), a novel 3D-stacked memory-based NDP architecture. We consider the selection, projection, and bloom join database query operators, commonly used by data analytics applications, comparing VIMA to a high-performance x86 baseline. Our results show speedups of up to 8× for selection, 6× for projection, and 16× for join while consuming up to 99% less energy. To the best of our knowledge, these results outperform the state-of-the-art for these operators on NDP platforms.

In big data environments, measuring performance interference while sharing resources is complex and not well covered by current benchmarks and tools. Different applications can compete for the same resources (e.g., persistent storage) until they become overloaded without notice either by the benchmark or the big data system. We unveil the case when RocksDB (a big data management system) shares disk flashes among multiple tenants and how YCSB and db_bench (two state-of-the-art big data benchmarks) fail to catch performance interference. In this paper, we define a methodology to measure the problem of performance interference over time and show how RocksDB quickly degrades throughput until it turns almost inoperative while sharing the persistent storage with other tenants.

The rapid growth of "big-data" intensified the problem of data movement when processing data analytics: Large amounts of data need to move through the memory up to the CPU before any computation takes place. To tackle this costly problem, Processing-in-Memory (PIM) inverts the traditional data processing by pushing computation to memory with an impact on performance and energy efficiency. In this paper, we present an experimental study on processing database SIMD operators in PIM compared to current x86 processor (i.e., using AVX512 instructions). We discuss the execution time gap between those architectures. However, this is the first experimental study, in the database community, to discuss the trade-offs of execution time and energy consumption between PIM and x86 in the main query execution systems: materialized, vectorized, and pipelined. We also discuss the results of a hybrid query scheduling when interleaving the execution of the SIMD operators between PIM and x86 processing hardware. In our results, the hybrid query plan reduced the execution time by 45%. It also drastically reduced energy consumption by more than 2x compared to hardware-specific query plans.