Sybase has quietly racked up a string of successful growth years, riding its pioneering status in commercial analytic databases (ADBMS) and holding on to its loyal base in everyday DBMS after being elbowed aside by Oracle a decade ago. Its steady market performance has not been driven by dramatic innovations: Sybase has seemed to lag the Big Three (Oracle, Microsoft and IBM) in new feature/function. But it has innovated: IQ has grown into a key revenue source, and Sybase RAP has established itself as one of the more successful event processing offerings, with a string of Wall Street customers creating a new class of applications.
In the current (5-year-old) major release level of its flagship Adaptive Server Enterprise (ASE) product, Sybase has added user-defined SQL functions, support for plugin Java Runtime Environment (JRE) and JVM components, xml tables, SQL statement replication, new statistical aggregate functions, and a shared disk cluster edition. And now, Sybase is about to add new in-memory database capabilities and step up its support for external storage management. I’ve spent some time recently with the Sybase team to discuss their plans for the upcoming 15.5 release (currently available in a developer version), and found palpable excitement about the possibilities of their new work.
One of the most obvious opportunities is to think about the role of I/O and disk storage in performance. Much of a database’s code line is spent logging, writing to disk, managing buffers, optimizing access strategies, and ensuring ACID properties for transactions. ACID properties (Atomicity, Consistency, Isolation and Durability) have been the hallmark of DBMS thinking since the earliest days of computing for transaction processing. But the assumptions they were designed for are not universally necessary – not all actions taken with data are transactional, and persistence may already be assured for non-transactional data elsewhere. Moreover, computing architectures have changed dramatically since the earliest days of transactional theory, and much of the thinking was designed for worst case scenarios occurring in system designs constrained by technology and economics that have since changed.
In-memory databases have been shown to create a substantial uplift in performance, but adoption has lagged. The shift in processor architectures to 64-bit is hardly new; decades have passed since MIPS and DEC introduced 64-bit microprocessors into mainstream computing, followed by Sun, IBM and HP. But software architectures that leveraged the opportunity lagged; operating systems took a long time to exploit the possibilities of a theoretical address space measured in terabytes. Even with the limitations of today’s systems and the cost of memory (which is continuing to fall rapidly), this is exciting stuff. But it’s been over a decade since Solaris 7 arrived in 1998, following on DEC’s Alpha-based OSs. And while Unix and Linux variants running on 64-bit architectures arrived, databases continued to be slow to re-architect for the new paradigm: DBMS vendors ported their products to run there, but didn’t dramatically re-engineer the parts of the codeline that could profitably take advantage of the new environment.
Many database products have complex features and options that can (in fact, must) be managed to achieve optimal performance, by relaxing assumptions and changing processing defaults where it is possible to do so. But the arcana associated with this has been the province of specialists – like those who tune the systems used for TPC benchmarks, or the top-notch staff at financial institutions responsible for trading systems. Sybase is exploiting the opportunities in a different way, by creating (rather than acquiring) a product whose codeline has been explicitly designed for in-memory use, and configured for two specific use cases where it is appropriate: one for data that does not need to be persisted at all ( a completely in-memory, no-disk instance) and one with “relaxed durability.” In this way, Sybase expects to dramatically expand the number of customers who can take advantage of the high-performance opportunity – and reach a new group of customers who may not have been considering ASE for many of their new projects.
Sybase has taken a different path here: instead of acquiring in-memory technology, like IBM’s SOLID DB or Oracle’s TimesTen, it has built an in-memory version of its existing code, reductively, ensuring compatibility and ease of interoperation much more rapidly than its competitors. This will make it far easier for its customers to build applications that leverage multiple persistence profiles in different database instances. In an age where many applications depend on look-aside processing, compliance management, and similar out-of-transactional band activities, this could be a powerful new attractor for modern workloads. Databases that are kept in memory for lookup purposes can be recreated automatically at startup from “templates,” so assumptions about their recovery can be completely different. Database elements like tables, stored procedures, datatypes, etc., are included; Sybase’s TSQL and all APIs are supported immediately. Management, monitoring, and the like will also be seamless from the product’s release, because no retrofitting will be required and no interim skill acquisition will be required for customers.
Another major step forward in systems architectures has been the growth of smart storage and associated management software, a requisite in a world where data volumes continue to explode and managing data is a task that goes on outside, as well as inside, the DBMS. Sybase’s customers, like those of its competitors, have heterogeneous environments with multiple database products and storage options. Recognizing that external storage management is frequently chosen, Sybase has partnered with IBM to leverage Tivoli Storage Manager (TSM)’s ability to work with multiple DBMSs and store across multiple types of disk and tape storage. Sybase’s Backup Server will integrate with TSM to hand this off.
These are solid steps forward, and should allow Sybase to continue to move forward on its successful growth path. While they are neither dramatic enough nor likely to be marketed aggressively enough to push Sybase into the top revenue tier, they are well-designed to allow the company’s steady growth to continue and even accelerate. Sybase has picked up market share in the past few years, but larger changes are afoot as the rise of open source, cloud computing, appliances, and proliferating analytics workloads change business and deployment models in the decade ahead. And Sybase’s “Unwired Enterprise” positioning around mobility, as powerful as it has been, hasn’t been connected to the ASE story here. New strategies will be needed from all the vendors as they confront the industry transitions ahead. Dramatic innovations at the edge will continue to nip at the heels of the largest players, picking off carefully targeted use cases. I’ll talk about many of them in the months ahead.