Implications of Emerging 3D GPU Architecture on the Scan Primitive

Abstract

Analytic database workloads are growing in data size and query complexity. At the same time, computer architects are struggling to continue the meteoric increase in performance enabled by Moore's Law. We explore the impact of two emerging architectural trends which may help continue the Moore's Law performance trend for analytic database workloads, namely 3D die-stacking and tight accelerator-CPU integration, specifically GPUs. GPUs have evolved from fixed-function units, to programmable discrete chips, and now are integrated with CPUs in most manufactured chips. Past efforts to use GPUs for analytic query processing have not had widespread practical impact, but it is time to re-examine and re-optimize database algorithms for massively data-parallel architectures. We argue that high-throughput data-parallel accelerators are likely to play a big role in future systems as they can be easily exploited by database systems and are becoming ubiquitous. Using the simple scan primitive as an example, we create a starting point for this discussion. We project the performance of both CPUs and GPUs in emerging 3D systems and show that the high-throughput data-parallel architecture of GPUs is more efficient in these future systems. We show that if database designers embrace emerging 3D architectures, there is possibly an order of magnitude performance and energy efficiency gain.

Near-Memory Address Translation

“…We focus on workloads whose working set fits (almost) entirely in memory. We believe this scenario is the common case as memory is growing exponentially cheaper and bigger [36], while modern online services and analytic engines consistently require low response times [7], [9], [21], [25], [35], [41], [57], [67], [76], [95]. Nevertheless, prior work on multi-GB caches backed up with an order-of-magnitude larger DRAM memory, has shown little sensitivity to associativity [48], [78].…”

“…With storage devices in recent decades dramatically lagging behind processors and memory in performance, and DRAM continuously improving in density and cost, many online services and analytic engines are carefully engineered to fit their working set in memory [7], [9], [21], [25], [35], [41], [57], [67], [76], [95]. Due to rare page swapping, contiguous virtual pages are often mapped to contiguous physical pages [71], [72], and hence the conventional page placement flexibility provided by full associativity remains largely unused.…”

“…In this work, we leverage the observation that most modern online services and analytic engines are practically memory resident [7], [9], [21], [25], [35], [41], [57], [67], [76], [95]. These systems experience page swapping extremely infrequently and have much less fragmented memory layouts.…”

When to use 3D Die-Stacked Memory for Bandwidth-Constrained Big Data Workloads

“…As a case study, we focus on an analytic database workload. Due to recent advances in analytic database query algorithms [10,19,20,27], the performance of analytic database queries is now constrained by the system's main memory bandwidth [28]. We find that the scan operator (a major contributor to the total time in analytic queries [19]) requires fetching about 4 bytes to the main memory per instruction when using SIMD.…”

“…There are new algorithms for analytic database queries that convert complex queries into simple operations like scans and aggregates [20]. Additionally, new hardware accelerators, like generalpurpose graphics processing units (GPGPUs) show potential to be both higher performance and lower energy than traditional multicore CPUs for these simpler algorithms [27,28]. Emerging hardware like GPGPUs are more efficient than CPUs because they have less complex hardware than multicore CPUs, which works together with the simplified analytic database query algorithms to increase performance and decrease energy.…”