Director – Power Systems
Originally, there were file systems like ISAM, VSAM, and Sequential, which was after card decks went away. They were sufficient for technology and data processing (remember that old term??)… These were needs of the 60’s and 70’s. With faster and larger storage capabilities, on-line processing (OLTP…. another oldie), and evolving business needs, relational (aka SQL… Sequential Query Language access) databases were created, and are now the major implementation and organization of all business data storage needs. Database solutions named Oracle DB, IBM DB2, PostgreSQL, Openedge Database (Progress), Microsoft SQL, etc. are key players in the relational database arena. Relational databases provide a row-column structure, and can generally contain most information that users store, access, and manipulate on a daily basis, along with acceptable performance metrics. Consider a simple row-column structure of an inventory system. The rows may be each item and then columns might be the quantity on hand, list price, colors, etc. Then, additional dimension might be who bought the item, number bought, location bought, etc.
Using this row-column structure and cross references of the schemas, data could be accessed quickly for specific orders, inventory control, etc. Also, using this row-column structure enabled the advent of analytics to do inquires and ”what if” type analysis. All was fine until the world got more complex, and customers needed more performance, data structures, and complex analytics on complex data. In relational databases, each element basically has a fixed size, 24 characters, for example. So, how do you put an element into a database that has an indeterminate size (say a picture or an audio clip or a video clip) or an indeterminate number of elements (say you wanted to store the feedback of all customers from internet feeds on twitter)? You can do it in a relational DB, but performance and capacity needs would not be optimized, and standard processes like backup, calculation of storage, etc. would be difficult and time consuming. Today, there is an evolution and growth in database systems specifically addressing different business needs and new solutions areas. At a high level, these are called document databases, no-sql databases, in-memory databases, scripted databases, etc. It is no longer the normal situation where relational databases are the default database architecture, regardless of the application and data structure needs. Today, use-cases are driving the acceptance and implementation of multiple database architectures that are matched to application needs, including performance, capacity, expandability, content management, compression, redundancy, etc. One estimate is, as we move forward, 50% of the current relational databases will be augmented or migrated to use-case specific databases.
What becomes interesting is recognizing that the infrastructure to manage and contain this data (as well as the database structure) is also evolving. Nutanix, XIV, V7000, SVCs, flash, SSDs, Spectrum Scale, ESS, etc. are all moving forward as physical storage advancements. These systems are costing less, having higher performance, and they have moved functionality into the hardware as automated functions (like compression, redundancy, etc.). There are also advances in servers and server architecture, such as larger memories, greater bandwidth, improved reliability/redundancy, more cores and threads/cores, that enable more processing on the data than ever before. Power Systems is leading the way in data-oriented computing with a wide range of systems matched to various data needs. Analytics and cognitive computing are at new levels to drive businesses, and are being adopted across all industries. They will benefit from advanced databases, storage and Power Systems.
But, taking a line from an Indiana Jones movie, “you must choose wisely.” There are a large number of both technologies and vendors from which to choose. And, most of the new software solutions, including databases, are implemented on Linux. This is because the OpenSource model has cross-platform capabilities on Linux, and is the force in rapid evolution of database and scripting capabilities. Today, if you need an in-memory database with high performance, Redis Labs has a solution based on the OpenSource Redis, but with enterprise level support. Running on a Power System enabled with CAPI and Flash, very high performance of in-memory no-SQL and caching of relational databases is possible. Mongo is a document database for unstructured data. While there is a Community edition to test with and learn from, there is also an Enterprise version that is optimized for Power Systems, which addresses the performance required for processing the quantity of data becoming available. Another solution, which is moving to new in-memory structure, is SAP HANA. SAP HANA is primarily targeted for analytics, but the future is for a centralized data structure that all the SAP suites can act upon at high speed, to support higher needs in transaction processing. SAP HANA also benefits from the performance and capacities of Power Systems. There are many more databases being developed to support specific business solutions… too many to mention, but a few to possibly watch are Orient DB, Extreme DB, and ScaleOut DB.
But, let us not ignore relational databases, which are also getting updates for Linux from the OpenSource community. Marie DB is the logical follow-on to MySQL. It has high performance and runs on Linux to provide consistency across a multiplatform environment. Enterprise DB is also growing in acceptance as an alternate relational database at lower cost. It runs on Linux for multiplatform consistency, with high performance and compatibility, as well as ease of migration from existing relational databases.
In summary, there are new solutions being enabled by new data structures, and corresponding databases enabling solutions never before imaginable. Coupled with new high performance systems like Power Systems and new storage systems, all Linux based, the IT industry is moving from a unique focus on relational databases to one of choosing the database to optimize the solution.
What about the cloud? Should I put my data in the cloud? That is another area that needs discussion and study.