7 Big Ideas
IBM BLU Acceleration is a group of technologies that enhance DB2 10.5 and help make it extremely suitable for Data Warehouse and Analytics jobs. The video above and the points below illustrate seven of the key ideas behind this technology.
1. Column Organized Storage
Traditional row-organized tables with index access are, generally speaking, well suited to meet the needs of transactional workloads. For analytics/OLAP workloads, BLU Acceleration provides DB2 10.5 with column organized tables. This in turn provides significant improvements to storage, query performance, and ease of use through simplified design and tuning. In some circumstances, queries on column-organized tables have been reported to be over a hundred times faster than the equivalent row-based queries.
Simplicity is the driving principle behind BLU. Each feature has been designed to simplify the operational model for our customers. For example, conventional row-based SQL queries require a great deal of consideration and fine tuning to ensure that the queries are fast, use the right indexes, and so on. BLU Accelerator provides queries that are optimized out of the box, without as much configuration and tuning. Likewise, every other part of the ecosystem, including load, backup, and restore operations have been greatly simplified so as to be easy to use.
3. Actionable Compression
BLU Acceleration introduces several exclusive techniques that enable DB2 to very efficiently compress and store data. Moreover, it allows operations to be performed on the compressed data (without the need to uncompress the data first). In fact, BLU Acceleration applies predicates, performs joins and aggregates, all on the encoded values of column-organized tables.
4. Core Friendly Parallelism
The computer industry moved away from an emphasis on simply faster CPUs, in favor of grouping together an ever increasing number of cores. Servers with dozens of cores are routinely available in OLAP and analytics environments. BLU Acceleration bakes in automated multi-core parallelism, ensuring that all your core CPUs will be used by your queries on column-organized tables. This ensures very high performances without the complication of manually partition and handle how to parallelize the data.
5. Vector Processing (SIMD)
Leveraging the power of Single Instruction Multiple Data (SIMD) available in modern CPUs, DB2 with BLU Acceleration further guarantees that all the available cores are busy processing your queries as fast as possible. In fact, using said hardware instructions, DB2 with BLU Acceleration can automatically apply a single instruction to many data elements simultaneously.
6. Scan Friendly Caching
Memory and cache management have been optimized to avoid I/O access as much as possible through a smart “victim selection” algorithm, which selects pages in the bufferpool. This means that your performance is going to be great whether the data fits in your memory or not.
7. Data Skipping
Sticking with the theme of providing incredible performance for your queries, BLU Accelerator provides the ability to automatically skip large sections of data that do not qualify for your query, and as such it is akin to an automatic “negative index”. This approach, which requires no additional steps or work on the user’s part, translates into an order of magnitude savings at the CPU, RAM, and I/O levels. Faster queries with no fine tuning, you’ve got to love that.