Non-Volatile RAM (NVRAM) is a novel class of hardware technology which is an interesting blend of two storage paradigms: byte-addressable DRAM and block-addressable storage (e.g. HDD/SSD). Most of the existing enterprise relational data management systems such as SAP HANA have their internal architecture based on the inherent assumption that memory is volatile and base their persistence on explicit handling of block-oriented storage devices. In this paper, we present the early adoption of Non-Volatile Memory within the SAP HANA Database, from the architectural and technical angles. We discuss our architectural choices, dive deeper into a few challenges of the NVRAM integration and their solutions, and share our experimental results. As we present our solutions for the NVRAM integration, we also give, as a basis, a detailed description of the relevant HANA internals.

Recent work has shown that perfect hashing and retrieval of data values associated with a key can be done in such a way that there is no need to store the keys and that only a few bits of additional space per element are needed. We present FiRe -a new, very simple approach to such data structures. FiRe allows very fast construction and better cache efficiency. The main idea is to substitute keys by small fingerprints. Collisions between fingerprints are resolved by recursively handling those elements in an overflow data structure. FiRe is dynamizable, easily parallelizable and allows distributed implementation without communicating keys. Depending on implementation choices, queries may require close to a single access to a cache line or the data structure needs as low as 2.58 bits of additional space per element.

No abstract

We present an overview of SAP HANA's Native Store Extension (NSE). This extension substantially increases database capacity, allowing to scale far beyond available system memory. NSE is based on a hybrid in-memory and paged column store architecture composed from data access primitives. These primitives enable the processing of hybrid columns using the same algorithms optimized for traditional HANA's in-memory columns. Using only three key primitives, we fabricated byte-compatible counterparts for complex memory resident data structures (e.g. dictionary and hash-index), compressed schemes (e.g. sparse and run-length encoding), and exotic data types (e.g. geo-spatial). We developed a new buffer cache which optimizes the management of paged resources by smart strategies sensitive to page type and access patterns. The buffer cache integrates with HANA's new execution engine that issues pipelined prefetch requests to improve disk access patterns. A novel load unit configuration, along with a unified persistence format, allows the hybrid column store to dynamically switch between inmemory and paged data access to balance performance and storage economy according to application demands while reducing Total Cost of Ownership (TCO). A new partitioning scheme supports load unit specification at table, partition, and column level. Finally, a new advisor recommends optimal load unit configurations. Our experiments illustrate the performance and memory footprint improvements on typical customer scenarios.

Units of the Simpson series of formations are evaluated by means of computing both a "textural" parameter, W, and the volume of mud filtrate that has displaced fluids originally present in the formation. This analysis is performed primarily through the use of response equations for an electromagnetic wave propagating through the formation. The first computation uses the interval transit time and the resulting solution for Sxo. The second one uses the attenuation of the wave. The textural parameter is used as both m and n in Archie's equation for water saturation. The m and n values agree with the range of values reported in core measurements for the Simpson. The resulting water saturation computations agree with production from examples. Computing the volume of moved fluids leads to a prediction of the amount of connate water displaced by the mud filtrate. This, in turn, leads to predictions about the water that the well will produce and a rough idea about the bulk volume of irreducible water. These predicted and computed values agree fairly well with production from examples. The Simpson series in Oklahoma is often characterized by low-resistivity pays, and the resistivity from an induction log is typically low in these formations. The water saturations computed from Archie's equation using m=n=2 are very high and are often in disagreement with production. Two very real possibilities for this disagreement are the use of wrong values for m and n and the presence of high bulk volumes of irreducible water. The techniques described in this paper help solve these two problems.