In-memory Database Management Systems Research Articles

Cost modelling for optimal data placement in heterogeneous main memory

The cost of DRAM contributes significantly to the operating costs of in-memory database management systems (IMDBMS). Persistent memory (PMEM) is an alternative type of byte-addressable memory that offers --- in addition to persistence --- higher capacities than DRAM at a lower price with the disadvantage of increased latencies and reduced bandwidth. This paper evaluates PMEM as a cheaper alternative to DRAM for storing table base data, which can make up a significant fraction of an IMDBMS' total memory footprint. Using a prototype implementation in the SAP HANA IMDBMS, we find that placing all table data in PMEM can reduce query performance in analytical benchmarks by more than a factor of two, while transactional workloads are less affected. To quantify the performance impact of placing individual data structures in PMEM, we propose a cost model based on a lightweight workload characterization. Using this model, we show how to place data pareto-optimally in the heterogeneous memory. Our evaluation demonstrates the accuracy of the model and shows that it is possible to place more than 75% of table data in PMEM while keeping performance within 10% of the DRAM baseline for two analytical benchmarks.

Mainlining databases

The proliferation of modern data processing tools has given rise to open-source columnar data formats. These formats help organizations avoid repeated conversion of data to a new format for each application. However, these formats are read-only, and organizations must use a heavy-weight transformation process to load data from on-line transactional processing (OLTP) systems. As a result, DBMSs often fail to take advantage of full network bandwidth when transferring data. We aim to reduce or even eliminate this overhead by developing a storage architecture for in-memory database management systems (DBMSs) that is aware of the eventual usage of its data and emits columnar storage blocks in a universal open-source format. We introduce relaxations to common analytical data formats to efficiently update records and rely on a lightweight transformation process to convert blocks to a read-optimized layout when they are cold. We also describe how to access data from third-party analytical tools with minimal serialization overhead. We implemented our storage engine based on the Apache Arrow format and integrated it into the NoisePage DBMS to evaluate our work. Our experiments show that our approach achieves comparable performance with dedicated OLTP DBMSs while enabling orders-of-magnitude faster data exports to external data science and machine learning tools than existing methods.

Permutable compiled queries

Just-in-time (JIT) query compilation is a technique to improve analytical query performance in database management systems (DBMSs). But the cost of compiling each query can be significant relative to its execution time. This overhead prohibits the DBMS from employing well-known adaptive query processing (AQP) methods to generate a new plan for a query if data distributions do not match the optimizer's estimations. The optimizer could eagerly generate multiple sub-plans for a query, but it can only include a few alternatives as each addition increases the compilation time. We present a method, called Permutable Compiled Queries (PCQ), that bridges the gap between JIT compilation and AQP. It allows the DBMS to modify compiled queries without needing to recompile or including all possible variations before the query starts. With PCQ, the DBMS structures a query's code with indirection layers that enable the DBMS to change the plan even while it is running. We implement PCQ in an in-memory DBMS and compare it against non-adaptive plans in a microbenchmark and against state-of-the-art analytic DBMSs. Our evaluation shows that PCQ outperforms static plans by more than 4X and yields better performance on an analytical benchmark by more than 2X against other DBMSs.

Taurus

Existing single-stream logging schemes are unsuitable for in-memory database management systems (DBMSs) as the single log is often a performance bottleneck. To overcome this problem, we present Taurus, an efficient parallel logging scheme that uses multiple log streams, and is compatible with both data and command logging. Taurus tracks and encodes transaction dependencies using a vector of log sequence numbers (LSNs). These vectors ensure that the dependencies are fully captured in logging and correctly enforced in recovery. Our experimental evaluation with an in-memory DBMS shows that Taurus's parallel logging achieves up to 9.9X and 2.9X speedups over single-streamed data logging and command logging, respectively. It also enables the DBMS to recover up to 22.9X and 75.6X faster than these baselines for data and command logging, respectively. We also compare Taurus with two state-of-the-art parallel logging schemes and show that the DBMS achieves up to 2.8X better performance on NVMe drives and 9.2X on HDDs.

Relaxed operator fusion for in-memory databases

In-memory database management systems (DBMSs) are a key component of modern on-line analytic processing (OLAP) applications, since they provide low-latency access to large volumes of data. Because disk accesses are no longer the principle bottleneck in such systems, the focus in designing query execution engines has shifted to optimizing CPU performance. Recent systems have revived an older technique of using just-in-time (JIT) compilation to execute queries as native code instead of interpreting a plan. The state-of-the-art in query compilation is to fuse operators together in a query plan to minimize materialization overhead by passing tuples efficiently between operators. Our empirical analysis shows, however, that more tactful materialization yields better performance. We present a query processing model called "relaxed operator fusion" that allows the DBMS to introduce staging points in the query plan where intermediate results are temporarily materialized. This allows the DBMS to take advantage of inter-tuple parallelism inherent in the plan using a combination of prefetching and SIMD vectorization to support faster query execution on data sets that exceed the size of CPU-level caches. Our evaluation shows that our approach reduces the execution time of OLAP queries by up to 2.2× and achieves up to 1.8× better performance compared to other in-memory DBMSs.

Relational Joins on GPUs: A Closer Look

The problem of scaling out relational join performance for large data sets in the database management system (DBMS) has been studied for years. Although in-memory DBMS engines can reduce load times by storing data in the main memory, join queries still remain computationally expensive. Modern graphics processing units (GPUs) provide massively parallel computing and may enhance the performance of such join queries; however, it is not clear yet in what condition relational joins perform well on GPUs. In this paper, we identify the performance characteristics of GPU computing for relational joins by implementing several well-known GPU-based join algorithms under various configurations. Experimental results indicate that the speedup ratio of GPU-based relational joins to CPU-based counterparts depends on the number of compute cores, the size of data sets, join conditions, and join algorithms. In the best case, the speedup ratios are up to 6.67 times for non-index joins, 9.41 times for sort index joins, and 2.55 times for hash joins. The execution time of GPU-based implementation for index joins, on the other hand, is only about 0.696 times less than the execution time of the CPU’s counterparts.

An empirical evaluation of in-memory multi-version concurrency control

Multi-version concurrency control (MVCC) is currently the most popular transaction management scheme in modern database management systems (DBMSs). Although MVCC was discovered in the late 1970s, it is used in almost every major relational DBMS released in the last decade. Maintaining multiple versions of data potentially increases parallelism without sacrificing serializability when processing transactions. But scaling MVCC in a multi-core and in-memory setting is non-trivial: when there are a large number of threads running in parallel, the synchronization overhead can outweigh the benefits of multi-versioning. To understand how MVCC perform when processing transactions in modern hardware settings, we conduct an extensive study of the scheme's four key design decisions: concurrency control protocol, version storage, garbage collection, and index management. We implemented state-of-the-art variants of all of these in an in-memory DBMS and evaluated them using OLTP workloads. Our analysis identifies the fundamental bottlenecks of each design choice.

LAYER: A cost-efficient mechanism to support multi-tenant database as a service in cloud

This paper presents a novel mechanism to cost-efficiently support multi-tenant database as a service (MTDBaaS) in cloud for small businesses. We aim at the scenarios where a large number of small tenants are served but only some of them are active simultaneously. By small tenants, we mean that a tenant may have many small-sized tables while only a small number of those tables are accessed concurrently for each query. As most MTDBaaS providers, we consolidate multiple tenants’ data into the same database management system (DBMS) to reduce the cost of operation. However, our solution distinguishes itself from the existing solutions by a novel mechanism: Load As You quERy (LAYER in short). Concretely, tenants can define and create their own tables with LAYER, and set up possible reference constraints between any two tables. A shared table is used to store all data for all tenants, but only a moderate number of working tables are maintained for answering queries from active tenants. When a new query is submitted, tables involved in the query but not yet in the DBMS will be restored: tables are created, and data are loaded to these newly-created tables. If an active tenant becomes inactive (logs out or no query is issued in a specified time period), tables belonging to the tenant could be dropped when necessary, and updates to these tables would be mirrored to the shared table for backup. We provide two implementations of the LAYER mechanism, one is LAYER-MySQL, which is based on the traditional disk-based relational DBMS MySQL, and can yield high consolidation and acceptable performance; the other is LAYER-VoltDB, which is based on the in-memory relational DBMS VoltDB, and can provide much higher performance. Experimental results validate the feasibility of the proposed mechanism.

In-Memory Modeling and Analytics as-a Service in Cloud

The aim is to conceptualize and create a platform for building complex data models and executing huge data processing tasks in a more efficient manner. We try to conceptualize a solution which caters to cloud infrastructure and memory-resident storage platform by providing a virtualized environment to end-users for carrying analytics on large scale without investing huge cost from infrastructure setup and maintenance point of view. The solution will support an analyst to replicate data and compute-intensive job running on his/her personal machine, to run on cloud with minimal effort and providing scope for future scale-out. There are several applications which are intensive in terms of data storage and computation cycles. For such applications hardware infrastructure and maintenance costs considerably increase with the increase in data volumes. These applications will vastly benefit with large scale in-memory database management system in cloud where the computations can be done with minimal effort. The idea is to enable provisioning for end users to custom build the skeleton of their data models for data manipulation and then export them using analytics service on the cloud. The ability of the system to expose results of computation in various data format will enable different unique insights to the data. The platform will enable users to harness the power of cloud and in-memory database to solve complex real-time problem with minimal effort and reduced cost. 

The Optimization of In-Memory Space Partitioning Trees for Cache Utilization

In this paper, a novel cache conscious indexing technique based on space partitioning trees is proposed. Many researchers investigated efficient cache conscious indexing techniques which improve retrieval performance of in-memory database management system recently. However, most studies considered data partitioning and targeted fast information retrieval. Existing data partitioning-based index structures significantly degrade performance due to the redundant accesses of overlapped spaces. Specially, R-tree-based index structures suffer from the propagation of MBR (Minimum Bounding Rectangle) information by updating data frequently. In this paper, we propose an in-memory space partitioning index structure for optimal cache utilization. The proposed index structure is compared with the existing index structures in terms of update performance, insertion performance and cache-utilization rate in a variety of environments. The results demonstrate that the proposed index structure offers better performance than existing index structures.