Highly Consistent Response Times with New Threaded Kernel in ASE 15.7
Sybase’s latest version of Adaptive Server Enterprise introduced a new thread-based kernel that significantly streamlines Input/Output handling and improves application partitioning capabilities. What’s more, ASE 15.7 dramatically improves system scalability and reduces query latency, all while improving the predictability and consistency of performance.
The new Threaded Kernel makes this possible by using the best of the VSA architecture — with its native DBMS threads for user requests — and complementing it with native OS threads for system services, according to Peter Thawley, senior director and architect in Sybase’s CTO Group. Thawley hosted last week’s webcast “Introducing ASE 15.7’s New Threaded Kernel.”
“Many applications, especially those in capital markets, demand highly consistent response times, often in the face of unpredictable and peak load conditions,” Thawley said. “This chart (at right) clearly demonstrates ASE 15.7’s new K21 Threaded Kernel meets this requirement.”
The fifth and final webcast in a series on ASE 15.7 covered the new kernel architecture, focusing on upgrades from the previous version, as well as performance monitoring and tuning. Thawley discussed workloads that show the greatest improvements; how to use thread pools for workload partitioning; new tunable parameters; new and modified MDA tables; and changes to sp_sysmon.
The Process Kernel was highly innovative for SMP systems of the 1990s, when it debuted in version 4.9. But advances in hardware parallelism of modern processor architectures (think large core counts and hardware threads, as well as virtualization’s increasing maturity and popularity) indicated that it was time to rethink the ASE kernel, according to Thawley.
“The Process Kernel balances query throughput with latency by periodically running system services, such as polling for disk and network I/O,” Thawley said. “This could create additional latency as user requests waited for these services to complete, but ASE 15.7’s new Threaded Kernel eliminates this latency by leveraging OS thread pools to do this work.”
Click on the links below for replays of previous webcasts.
And click here for a replay of last week’s webcast on the Threaded Kernel.