IBM Systems Magazine, Mainframe - September/October 2017 - 52
ADMINISTRATOR Pause-Less Garbage Collection Java is one of the premier platforms for deploying online transaction processing workloads with strict response time service-level agreements (SLA) and analytics applications with large heaps. In both scenarios, tuning of the Java heap and GC policies of the JVM present one of the biggest opportunities and challenges in extracting better Java performance. GC is an automated memory management feature of the JVM that kicks in when the Java heap becomes exhausted. During GC phases, application threads are paused to allow GC to coalesce live objects and free up heap memory. Such periods, known as stop-the-world pause times-where application progress is stalled-can impact both application throughput and response time. IBM z14 introduces specialized hardware that enables the JVM's GC to relocate objects concurrently with the user application. The JVM can mark ranges of the Java heap from which objects are being migrated during a GC cycle. The z14 hardware provides a lightweight mechanism of detecting objects that reference those regions. The combination of JVM exploitation and hardware allows for both GC and application to execute in parallel, resulting in significant reductions in stop-theworld pause times. Figure 1 (right) shows a Java Store Inventory and Point of Sale application that demonstrates an average pause-time of 300 ms with the default generation concurrent collector GC policy. Using the pause-less GC mode with IBM Java 8 Service Refresh (SR) 5 running on z14, the average GC pause-times have reduced to 30 ms, a 10x improvement. Furthermore, the Figure 1: Garbage Collection Pause Times Figure 2: DayTrader on WebSphere Liberty Profile pause-less GC mode delivers up to 3x better throughput under specific response-time SLAs, as the shorter GC pause times help ensure more consistent application response-times. High-Performance Encryption Today's business climate demands trusted computing and the utmost security for data and transactions. Pervasive encryption 52 // SEPTEMBER/OCTOBER 2017 ibmsystemsmag.com has become the norm, and high performance of encryption and cryptography has become crucial in many applications. IBM Java 8 introduced extensive hardware exploitation of IBM Z Central Processor Assist for Cryptographic Functions (CPACF) instructions to accelerate many aspects of encryption for connections secured via Transport Layer Security (TLS) protocols, including SHA-1/SHA-2 hashing algorithms, elliptic
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.