Roger Bitar's Sun BluePrints Online paper, which he entitles "Deploying Hybrid Storage Pools with Sun Flash Technology and the Solaris ZFS File System," provides an overview of Flash technology and discusses its introduction into a new tier of storage infrastructure. Bitar explains how companies can utilize Flash technology and the Solaris ZFS file system to take advantage of the high performance of enterprise solid state drives (SSDs) and the low cost of high-capacity hard disks to create hybrid storage pool solutions that help balance system performance and cost.

The paper points out the economical ineffectiveness of replacing all hard disk drives in a system with enterprise SSDs in typical datacenter storage infrastructures. The right approach, Bitar suggests is to combine the strengths of both technologies. He recommends placing enterprise SSDs in a new storage tier to assist hard disk drives by holding frequently accessed data to minimize the impact of disk latencies and improve application performance. Utilizing enterprise SSDs to handle CPU I/O, and hard disk drives to store massive data sets, enables a hybrid storage pool to deliver significant performance gains without sacrificing capacity.

The writer illustrates his point with a graphical representation of such a combination in which a quad-core server with a maximum memory configuration is connected to enterprise SSDs and hard disk drive storage via multiple host adapters. Both the resulting balanced system and combined storage pool can generate hundreds of thousands of IOPS, he explains.

The Solaris ZFS, an enterprise-class, general-purpose file system that provides virtually unlimited file system scalability and increased data integrity to large-scale solutions, is the solution Bitar recommends in the combined storage pool. With up to 21 billion YottaBytes of capacity, this 128-bit, open source file system integrates traditional file system functionality with built-in volume management techniques, Bitar notes, adding that, by automatically allocating space from pooled storage when needed, Solaris ZFS simplifies storage management and gives organizations the flexibility to optimize data for performance.

There are several key capabilities of the Solaris ZFS that qualify it for such a role in the hybrid storage pool, including:

• Virtual storage pools
• Data integrity
• High performance
• Simplified administration

"Solaris ZFS provides a seamless and easy way to administer hybrid storage pools, taking advantage of the performance of enterprise-class SSDs and inexpensive hard disk drive capacity. Unlike traditional volume managers that simply use SSDs in a RAID stripe, Solaris ZFS integrates the volume manager with the file system and can use enterprise SSDs more effectively," Bitar writes. He adds that Solaris ZFS also accelerates both read and write operations and allows administrators to tailor system configuration to workload demands.

The Solaris ZFS, Bitar continues, combines main memory and enterprise SSDs into a large read cache and uses an Adaptive Replacement Cache (ARC) for its cache replacement algorithm. The ARC manages and balances the cache content using most frequently used (MFU) and most recently used (MRU) algorithms for storing data to, and retrieving data from, memory.

He explains that a second-level ARC (L2ARC) with smart caching and pre-fetching techniques lets Solaris ZFS use enterprise SSDs as a second-level cache to speed read performance even further. Defective Flash blocks are treated as a cache miss rather than data loss, with information retrieved from hard disk to satisfy the request. The checksums built into Solaris ZFS are used to catch cache inconsistencies, he adds.

The benefits of using Solaris ZFS and an L2ARC stored on Flash devices helps users to:

• Eliminate disk latency
• Speed access to working sets
• Enhance cache performance
• Speed system readiness by warming caches
• Reduce the volatility of cache content

Bitar goes on to explain how Solaris ZFS can deliver reduced write latency through its ZFS Intent Log (ZIL), which allows applications that demand synchronous writes to a permanent storage medium to benefit from apparent latency reductions and get work done while data is written asynchronously in the background.

Small transactions are stored to the file system in a dedicated enterprise SSD pool before committing the transaction to disk. The ZIL stores enough information to replay the transaction, if needed, freeing these records after the data is committed to disk. The ZIL handles small and large writes differently, including small writes in the log record while synchronizing large writes to disk.

Applications that can benefit from implementation of a hybrid storage pool include database applications, distributed file systems and Web 2.0 applications that use a memchached distributed memory system, Bitar suggests.

Read the full text of Bitar\'s paper here. [...read more...]

		fullsource
Print Email Current Profile Archive Blog Generate newsletter