Jinwoong Kim

Parallel multi-dimensional range query processing with R-trees on GPU

The general purpose computing on graphics processing unit (GP-GPU) has emerged as a new cost effective parallel computing paradigm in high performance computing research that enables large amount of data to be processed in parallel. Large scale scientific data intensive applications have been playing an important role in modern high performance computing research. A common access pattern into such scientific data analysis applications is multi-dimensional range query, but not much research has been conducted on multi-dimensional range query on the GPU. Inherently multi-dimensional indexing trees such as R-Trees are not well suited for GPU environment because of its irregular tree traversal. Traversing irregular tree search path makes it hard to maximize the utilization of massively parallel architectures. In this paper, we propose a novel MPTS (Massively Parallel Three-phase Scanning) R-tree traversal algorithm for multi-dimensional range query, that converts recursive access to tree nodes into sequential access. Our extensive experimental study shows that MPTS R-tree traversal algorithm on NVIDIA Tesla M2090 GPU consistently outperforms traditional recursive R-trees search algorithm on Intel Xeon E5506 processors.

NVWAL: Exploiting NVRAM in Write-Ahead Logging

Emerging byte-addressable non-volatile memory is considered an alternative storage device for database logs that require persistency and high performance. In this work, we develop NVWAL (NVRAM Write-Ahead Logging) for SQLite. The contribution of NVWAL consists of three elements: (i) byte-granularity differential logging that effectively eliminates the excessive I/O overhead of filesystem-based logging or journaling, (ii) transaction-aware lazy synchronization that reduces cache synchronization overhead by two-thirds, and (iii) user-level heap management of the NVRAM persistent WAL structure, which reduces the overhead of managing persistent objects. We implemented NVWAL in SQLite and measured the performance on a Nexus 5 smartphone and an NVRAM emulation board - Tuna. Our performance study shows the following: (i) the overhead of enforcing strict ordering of NVRAM writes can be reduced via NVRAM-aware transaction management. (ii) From the application performance point of view, the overhead of guaranteeing failure atomicity is negligible; the cache line flush overhead accounts for only 0.8~4.6% of transaction execution time. Therefore, application performance is much less sensitive to the NVRAM performance than we expected. Decreasing the NVRAM latency by one-fifth (from 1942 nsec to 437 nsec), SQLite achieves a mere 4% performance gain (from 2517 ins/sec to 2621 ins/sec). (iii) Overall, when the write latency of NVRAM is 2 usec, NVWAL increases SQLite performance by at least 10x compared to that of WAL on flash memory (from 541 ins/sec to 5812 ins/sec).

Exploiting Massive Parallelism for IndexingMulti-Dimensional Datasets on the GPU

Inherently multi-dimensional n-ary indexing structures such as R-trees are not well suited for the GPU because of their irregular memory access patterns and recursive back-tracking function calls. It has been known that traversing hierarchical tree structures in an irregular manner makes it difficult to exploit parallelism and to maximize the utilization of GPU processing units. Moreover, the recursive tree search algorithms often fail with large indexes because of the GPUs tiny runtime stack size. In this paper, we propose a novel parallel tree traversal algorithm-massively parallel restart scanning (MPRS) for multi-dimensional range queries that avoids recursion and irregular memory access. The proposed MPRS algorithm traverses hierarchical tree structures with mostly contiguous memory access patterns without recursion, which offers more chances to optimize the parallel SIMD algorithm. We implemented the proposed MPRS range query processing algorithm on n-ary bounding volume hierarchies including R-trees and evaluated its performance using real scientific datasets on an NVIDIA Tesla M2090 GPU. Our experiments show braided parallel SIMD friendly MPRS range query algorithm achieves at least 80 percent warp execution efficiency while task parallel tree traversal algorithm shows only 9-15 percent efficiency. Moreover, braided parallel MPRS algorithm accesses 7-20 times less amount of global memory than task parallel parent link algorithm by virtue of minimal warp divergence.

A Performance Study of Traversing Spatial Indexing Structures in Parallel on GPU

CUDA is a parallel programming environment that enables significant performance improvement by leveraging the massively parallel processing capability of the GPU. Inherently spatial indexing structures such as R-Trees are not well suited for CUDA environment due to its irregular tree traversal for range queries. Traversing irregular tree search paths makes it hard to maximize the utilization of many-core architectures. In this paper, we propose assigning an individual sub-tree to each SMP (streaming multi-processor) in GPGPU, such that CUDA cores in the same SMP co-operate to navigate tree index nodes. This parallel partitioned-indexing improves the utilization of many cores in GPGPU significantly. Also, we propose a new range query search algorithm - three-phase-search that avoids non-sequential random access to tree nodes and accelerates the search performance of spatial indexing structures on GPU. Our experimental results show that GPU-based parallel spatial indexing scheme on NVIDA Tesla M2090 GPGPU outperforms the CPU-based multi-threaded R-trees on AMD Opteron 6128HE processor by two times.

High-throughput query scheduling with spatial clustering based on distributed exponential moving average

In distributed scientific query processing systems, leveraging distributed cached data is becoming more important. In such systems, a front-end query scheduler distributes queries among many application servers rather than processing queries in a few high-performance workstations. Although many query scheduling policies exist such as round-robin and load-monitoring, they are not sophisticated enough to exploit cached results as well as balance the workload. Efforts were made to improve the query processing performance using statistical methods such as exponential moving average. However, existing methods have limitations for certain query patterns: queries with hotspots, or dynamic query distributions. In this paper, we propose novel query scheduling policies that take into account both the contents of distributed caching infrastructure and the load balance among the servers. Our experiments show that the proposed query scheduling policies outperform existing policies by producing better query plans in terms of load balance and cache-hit ratio.

Parallel Tree Traversal for Nearest Neighbor Query on the GPU

The similarity search problem is found in many application domains including computer graphics, information retrieval, statistics, computational biology, and scientific data processing just to name a few. Recently several studies have been performed to accelerate the k-nearest neighbor (kNN) queries using GPUs, but most of the works develop brute-force exhaustive scanning algorithms leveraging a large number of GPU cores and none of the prior works employ GPUs for an n-ary tree structured index. It is known that multi-dimensional hierarchical indexing trees such as R-trees are inherently not well suited for GPUs because of their irregular tree traversal and memory access patterns. Traversing hierarchical tree structures in an irregular manner makes it difficult to exploit parallelism since GPUs are tailored for deterministic memory accesses. In this work, we develop a data parallel tree traversal algorithm, Parallel Scan and Backtrack (PSB), for kNN query processing on the GPU, this algorithm traverses a multi-dimensional tree structured index while avoiding warp divergence problems. In order to take advantage of accessing contiguous memory blocks, the proposed PSB algorithm performs linear scanning of sibling leaf nodes, which increases the chance to optimize the parallel SIMD algorithm. We evaluate the performance of the PSB algorithm against the classic branch-and-bound kNN query processing algorithm. Our experiments with real datasets show that the PSB algorithm is faster by a large margin than the branch-and-bound algorithm.

Multi-dimensional multiple query scheduling with distributed semantic caching framework

It is becoming more important to leverage a large number of distributed cache memory seamlessly in modern large scale systems. Several previous studies showed that traditional scheduling policies often fail to exhibit high cache hit ratio and to achieve good system load balance with large scale distributed caching facilities. To maximize the system throughput, distributed caching facilities should balance the workloads and leverage cached data at the same time. In this work, we present a distributed job processing framework that yields high cache hit ratio while achieving balanced system load. Our framework employs a scheduling policy—DEMA that considers both cache hit ratio and system load and it supports geographically distributed multiple job schedulers. We show collaborative task scheduling and the data migration can even further improve the performance by increasing the cache hit ratio while achieving good load balance. Our experiments show that the proposed job scheduling policies outperform legacy load-based job scheduling policy in terms of job response time, load balancing, and cache hit ratio.

clfB-tree: Cacheline Friendly Persistent B-tree for NVRAM

Emerging byte-addressable non-volatile memory (NVRAM) is expected to replace block device storages as an alternative low-latency persistent storage device. If NVRAM is used as a persistent storage device, a cache line instead of a disk page will be the unit of data transfer, consistency, and durability. In this work, we design and develop clfB-tree—a B-tree structure whose tree node fits in a single cache line. We employ existing write combining store buffer and restricted transactional memory to provide a failure-atomic cache line write operation. Using the failure-atomic cache line write operations, we atomically update a clfB-tree node via a single cache line flush instruction without major changes in hardware. However, there exist many processors that do not provide SW interface for transactional memory. For those processors, our proposed clfB-tree achieves atomicity and consistency via in-place update, which requires maximum four cache line flushes. We evaluate the performance of clfB-tree on an NVRAM emulation board with ARM Cortex A-9 processor and a workstation that has Intel Xeon E7-4809 v3 processor. Our experimental results show clfB-tree outperforms wB-tree and CDDS B-tree by a large margin in terms of both insertion and search performance.

Improving Multi-dimensional query processing with data migration in distributed cache infrastructure

In distributed query processing systems where caching infrastructure is distributed and scales with the number of servers, it is becoming more important to orchestrate and leverage a large number of cached objects in distributed caching systems seamlessly as the present trend is to build large scalable distributed systems by connecting small heterogeneous machines. With a large scale distributed caching system, a scheduling policy must consider both cache hit ratio and system load balance to optimize multiple queries. A scheduling policy that considers system load but not cache hit ratio often fails to reuse cached data by not assigning a query to the sever that has data objects the query needs. On the contrary, a scheduling policy that considers cache hit ratio but not system load balance may suffer from system load imbalance. To maximize the overall system throughput and to reduce query response time, a multiple query scheduling policy must balance system load and also leverage cached objects. In this paper, we present a distributed query processing framework that exhibits high cache hit ratio while achieving good system load balance. In order to seamlessly manage our distributed scalable caching system, our framework performs autonomic cached data migrations to improve cache hit ratio. Our experiments show that our proposed query scheduling policy and data migration policy significantly improve system throughput by achieving high cache hit ratio while avoiding system load imbalance.

Collaborative Multi-dimensional Dataset Processing with Distributed Cache Infrastructure in the Cloud

As modern large scale systems are built with a large number of independent small servers, it is becoming more important to orchestrate and leverage a large number of distributed buffer cache memory seamlessly. Several previous studies showed that with large scale distributed caching facilities, traditional resource scheduling policies often fail to exhibit high cache hit ratio and to achieve good system load balance. A scheduling policy that solely considers system load results in low cache hit ratio, and a scheduling policy that puts more emphasis on cache hit ratio than load balance suffers from system load imbalance. To maximize the overall system throughput, distributed caching facilities should balance the workloads and also leverage cached data at the same time. In this work, we present a distributed job processing framework that yields high cache hit ratio while achieving good system load balance, the two of which are most critical performance factors to improve overall system throughput and job response time. Our framework is a component-based distributed data analysis framework that supports geographically distributed multiple job schedulers. The job scheduler in our framework employs a distributed job scheduling policy -- DEMA that considers both cache hit ratio and system load. In this paper, we show collaborative task scheduling can even further improve the performance by increasing the overall cache hit ratio while achieving load balance. Our experiments show that the proposed job scheduling policies outperform legacy load-based job scheduling policy in terms of job response time, load balancing, and cache hit ratio.