When FlashSystem V840 Real-time Compression is used on a compressible workload, the resulting solution has up to a 5:1 reduction in cost and capacity, while maintaining up to a 5x reduction in latency compared to disk drive based systems. Redpaper Accelerate with FlashSystem V840 Compression discusses “In a nutshell” how to effectively implement FlashSystem V840 with Real-time Compression (RtC). It walks you through planning, set up and operations, and provides performance guidance needed to exploit V840 performance and capacity savings. Candidate data sets and workloads best suited for compression are discussed. Examples are also provided on using the Comprestimator Utility to estimate expected compression rates for V840 with various workloads.
The Operations and analysis chapter explains the V840 7.4 software stack, performance monitoring, using synthetic workloads with RtC, V840 with RtC compared with disk, as well as topics on analysis and verification.
It is important to understand the FlashSystem V840 software stack and the flow of read and write requests as it pertains to Real-time Compression. The FlashSystem V840 software stack is shown in the figure below. Compression is transparently integrated with existing system management design. All of the FlashSystem V840 advanced features are supported on compressed volumes. You can create, delete, migrate, map (assign), and unmap (unassign) a compressed volume as though it were a fully allocated volume. In addition, you can utilize Real-time Compression along with Easy Tier on the same volumes. This compression method provides non disruptive conversion between compressed and decompressed volumes. This conversion provides a uniform user-experience and eliminates the need for special procedures when dealing with compressed volumes.
The Real-time Compression software component sits below the upper-level fast write cache and above the lower-level advanced read/write cache. There are several advantages to this dual-level model regarding Real-time Compression:
- Host writes, whether to compressed or decompressed volumes, are still serviced directly via the upper-level write cache, preserving low host write I/O latency. Response time can improve with this model as the upper cache flushes less data to Real-time Compression component more frequently.
- The performance of the destaging of compressed write I/Os to storage is improved because these I/Os are now destaged via the advanced lower-level cached, as opposed to directly to storage.
- The existence of a lower-level write cache below the Real-time Compression component in the software stack allows for the coalescing of compressed writes, and as a result, a reduction in back-end I/Os due to the ability to perform full-stride writes for compressed data.
- The existence of a lower-level read cache below the Real-time Compression component in the software stack allows the temporal locality nature of Real-time Compression to benefit from pre-fetching from the backend storage.
- The main (lower level) cache stores compressed data for compressed volumes, increasing the effective size of the lower-level cache.
Read the publicly available draft of Redpaper, Accelerate with FlashSystem V840 Compression for more detail and use cases that demonstrate how to implement and immediately start gaining the benefits of FlashSystem V840 with Real-time Compression.