Home > Workload Solutions > SQL Server > White Papers > Dell EMC PowerMax Storage for Mission-Critical SQL Server Databases > Adaptive Compression Engine
PowerMax Adaptive Compression Engine (ACE) offers high performance and maximum storage efficiency for application environments. ACE compresses data and efficiently optimizes system resources to balance overall system performance. ACE features include:
ABC marks the busiest data in the storage resource pool (SRP) to skip the compression flow regardless of the related SG compression setting. This function differentiates busy data from idle or less-busy data and only accounts for up to 20 percent of the allocations in the SRP. It marks up to 20 percent of the busiest allocations to skip the compression action, ensuring optimal response time and reducing the overhead that can result from the act of compressing data to save space.
The mechanism that determines the busiest data does not add CPU load to the system. ABC uses statistics that are collected from front-end devices to determine which datasets are the best candidates for compression. The system can then maintain balance across the resources, providing an optimal environment for both the best possible compression savings and the best performance. Effectively, this balance avoids compression and decompression latency for the busiest data and reduces system overhead.
Fine-grained data packing offers performance benefits for both the compression function and the overall performance of the system. Included in this process is a zero reclaim function that prevents the allocation of buffers with all zeros or no actual data. Pairing the zero reclaim function with fine-grained data packing enables the compression function to operate efficiently with minimal impact on performance. The 128 KB I/O are compressed in four buffers individually, and in parallel, enabling each section to be handled independently, although they are still part of the initial 128 KB I/O. The main benefit comes in the case of partial write updates or read I/O. If only one or two of the sections are updated or read, only that data is decompressed.