An Oracle White Paper March Enabling Greater Performance and Efficiency with Hybrid Storage Pools - PDF

Please download to get full document.

View again

of 15
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report



Views: 7 | Pages: 15

Extension: PDF | Download: 0

Related documents
An Oracle White Paper March 2010 Enabling Greater Performance and Efficiency with Hybrid Storage Pools Executive Overview... 2 Introduction... 2 A New Approach to Storage Architectures... 3 Business Benefits
An Oracle White Paper March 2010 Enabling Greater Performance and Efficiency with Hybrid Storage Pools Executive Overview... 2 Introduction... 2 A New Approach to Storage Architectures... 3 Business Benefits of Hybrid Storage Pools... 3 Challenges with Traditional Storage Approaches... 4 Flash Technology and Hybrid Storage Pools... 5 Cost and Performance Trade-offs... 6 Using Flash Technology in a Tier 0 Configuration... 7 Hybrid Storage Pools Provide a New Approach... 8 Hybrid Storage Pools and Oracle Solaris ZFS... 8 Transparent Storage Optimization... 9 Business Value of Hybrid Storage Pools Improving Productivity Enabling Flexibility for New Types of Solutions Radically Simplifying Enterprise Storage Lowering Acquisition Costs and Ongoing Costs Conserving Datacenter Resources Business Value Scenarios Conclusion... 13 Executive Overview Today s enterprise applications require fast access to large volumes of information, and are pushing the performance limits of traditional storage architectures. Although it is easy to add storage capacity by adding more disk drives and arrays, it is not so easy to scale storage I/O performance. Hybrid storage pools enable a new approach to storage architectures, increasing performance, reducing costs, and simplifying management. Introduction Enterprise applications range from high-volume online transaction processing (OLTP) and database applications to Web 2.0 and high-performance computing (HPC) solutions. Many of these modern applications have become dependent on I/O throughput for application performance. As the number of users and the amount of data stored continue to grow, traditional storage architectures require more and more disk drive spindles to generate enough I/O throughput to keep up with the I/O processing speed of servers. This white paper provides an overview of hybrid storage pools (HSPs) and how they can help businesses meet the growing demands for storage I/O throughput in today s data-intensive applications. 2 A New Approach to Storage Architectures Recent advances in the production of flash technology have made solid-state drives (SSDs) and flash array products much more cost-effective, enabling a new approach to storage architectures using HSPs. HSPs enable multiple types of storage media to be deployed together and managed as a single pool of storage. Applications that access the HSP can then automatically take advantage of the different performance and capacity characteristics of the different types of storage media. HSPs utilize Oracle Solaris ZFS to transparently optimize data placement across different types of storage media. Frequently used data is directed to high-performance storage devices such as enterprise SSDs, and less-frequently used data is placed in high-capacity storage devices such as mechanical disks. By combining the high performance of SSDs with high-capacity disks, HSPs enable dramatic performance gains while also providing tremendous savings in cost including savings in datacenter space, power, and cooling. Management of an HSP is transparent to users and simple for storage administrators. HSPs can be quickly and easily deployed, allowing organizations to immediately tackle pressing application performance bottlenecks. HSPs can also necessitate fewer components in the storage infrastructure, thus reducing complexity and saving on storage administration costs. Business Benefits of Hybrid Storage Pools Key business benefits of HSPs include the following: Increased application performance and extraordinary scalability as a result of utilizing flash technology to buffer application reads and writes, enabling business applications to process more revenue-based transactions Radically simplified management, reducing labor costs by eliminating the need for highly skilled storage administrators, improving productivity, and increasing service levels through reduced risk of errors Reduced acquisition costs as a result of using high-capacity drives and flash storage rather than expensive 15,000 rpm disk drives, which often require excess capacity just to meet I/O throughput requirements Lower ongoing costs because business applications can be scaled without overprovisioning capacity and because administrative costs as well as datacenter space, power, and cooling requirements can be reduced Architectures based on HSPs can cost as little as one-tenth and consume as little as one-fifth the power of traditional monolithic storage pools. Organizations can immediately begin benefiting from HSPs by deploying Oracle s Sun open storage solutions with Oracle Solaris ZFS. Oracle s Sun Storage 7000 Unified Storage Systems, for example, are appliances that can include flash technology and have HSPs already implemented within the 3 appliance. HSPs can also be incorporated into an existing or customized storage infrastructure by leveraging the Oracle Solaris ZFS capabilities for HSPs that are present in the open source OpenSolaris operating system (OS). This enables HSPs to be implemented on storage offerings from other vendors as well. Challenges with Traditional Storage Approaches Today s faster servers allow business and technical applications to serve thousands of users and generate millions of data records. As data volumes and user activity continue to grow, however, traditional storage solutions are presenting severe limitations. Today s storage administrators and IT managers are faced with three key challenges for their datacenter storage: Application Performance Bottlenecks Advances in server technology have yielded extremely fast multicore servers that are approaching I/O processing capabilities in excess of one million I/O operations per second (IOPS). Hard disk drive performance, however, has lagged behind the recent dramatic gains in server performance. Disk drive performance is largely constrained by disk drive rotational speeds, which have increased by a small fraction compared to server performance gains. Today s fastest disk drives are capable of 300 to 400 IOPS. Thus, nearly 3,000 drives would be required to handle the one million IOPS of today s servers. Unsustainable Costs Scaling to meet the performance requirements of today s I/O-intensive applications can be a very costly undertaking when using traditional storage architectures. To compensate for the performance gap and slow seek times of hard disk drives (HDDs), IT managers generally employ one of two approaches (see Figure 1). In the first approach, application data is either spread across a large pool of high-performance 15,000 rpm disk drives, or some portion of the application data set is maintained in a large buffer of dynamic random access memory (DRAM) or solid-state drive (SSD) storage to reduce latency. Figure 1. Traditional options for extending storage performance are becoming too costly. 4 Deploying a large pool of spinning disks effectively multiplies storage I/O throughput by enabling read and write operations to be performed in parallel. However, it can require large quantities of HDDs to keep up with today s server performance needs and maintain service-level agreement (SLA) requirements. This approach results in high costs, due to the excessive power requirements of highspeed drives. And, because drives are often only partially filled, this results in poor utilization of datacenter space, power, and cooling resources. The second approach of implementing a Tier 0 storage solution using DRAM or SSDs can speed performance by providing lower-latency access. Sometimes an entire application working set can be held in DRAM. However, DRAM and SSDs are much more costly to acquire than disk drives and can add significantly to storage management complexity when used in a Tier 0 storage environment where administrators often manually move data to help optimize performance. Complex Storage Management Not only are today s applications processing greater volumes of data, but much of the data must also be maintained over time and securely shared throughout a global community of users. Traditional means of deploying and managing storage are becoming outdated as the volume of data under management increases. Storage administrators must manage vast arrays of storage and are finding it increasingly difficult to identify and quickly fix storage performance and availability issues. This difficulty not only affects application service levels but also results in higher administration costs. Most of today s storage systems require a highly trained administrator to provision and actively manage the environment. Often, there is a need to manually provision storage or to manually move data to continually adjust application performance as business needs change. Even in cases where hierarchical storage management (HSM) solutions are deployed, the degree of manual intervention needed to optimize performance for high-throughput applications can make storage administration costly. HSM solutions are very cost-effective for managing data migration to an archival tier. However, they are less effective when attempting to maintain frequently used data in a Tier 0 cache storage solution. They often require manual intervention for performance optimization or a costly deployment strategy that includes overprovisioning the high-performance storage media so that more data can be cached. It is becoming impractical to meet today s performance needs with traditional storage architectures. The next level of application performance requires a new and more cost-effective storage architecture that can help prevent storage I/O bottlenecks and help bring down storage costs. Flash Technology and Hybrid Storage Pools Flash memory is low-cost, nonvolatile computer memory that can be electronically erased and rewritten. Several advances in flash technology are making it possible to utilize flash technology and enterprise SSDs to optimize the storage hierarchy in enterprise storage implementations to offer improved throughput: 5 High performance. Flash technology completes operations in microseconds, placing it between HDDs (with operations in milliseconds) and DRAM (with operations in nanoseconds) for access time. Because flash technology contains no moving parts, enterprise SSDs can avoid the seek times and rotational latencies associated with traditional HDDs and thus provide tens of thousands of IOPS, compared to hundreds of IOPS for HDDs. Low power consumption. Because flash integrated circuits have no motors or other mechanical parts, enterprise SSDs consume a fraction of the power of conventional HDDs. In fact, enterprise SSDs use only 5 percent of the power used by HDDs when idle and as little as 15 percent when performing operations. Enterprise SSDs also produce less heat in the system chassis, providing relief for both power and cooling costs. Low cost. Although flash devices are more expensive per gigabyte than a comparable disk drive, they are much less expensive in terms of cost per IOPS. Hard disk drives cost approximately US$1.25 per IOPS, compared to only US$0.02 per IOPS for enterprise SSDs. When used with HSPs, flash technology enables dramatic reductions in acquisition cost, compared to a similarly performing solution based solely on HDDs. High reliability. While enterprise SSDs provide similar functionality to traditional hard drives, they offer improved reliability. Both HDDs and enterprise SSDs support bad block management, wear leveling, and error correction codes (ECC) to foster data integrity. However, unlike hard drives, enterprise SSDs contain no moving parts. Data is stored on integrated circuits that can withstand significant shock and vibration. In fact, enterprise SSDs operate in a wider thermal operating range and in a wider operational vibration range than HDDs do, to deliver significantly higher mean time between failures (MTBF) (2.0 million hours versus 1.2 million hours). These characteristics of enterprise flash technology make it a strong alternative or adjunct to HDDs, providing a means to rebalance system and storage I/O performance by using HSPs. Cost and Performance Trade-offs Today s organizations often use multiple media types in their storage infrastructure because there are cost and performance trade-offs with different media types. As shown in Figure 2, there is an inverse correlation between cost per GB of storage and the time required to retrieve the first byte of data. Tape libraries, for example, clearly provide the lowest cost per GB, but they can be used only for archival purposes due to their slow access speed. 6 Figure 2. Hybrid storage pools enable organizations to take advantage of the different cost and performance characteristics of different media types. Enterprise flash and SSD technologies fall in a cost and performance sweet spot between mechanical drives and DRAM. They are nonvolatile and significantly cheaper than DRAM. They also offer much higher performance and greater power efficiency than HDDs. Deploying new architectures with flash technology can enable significantly greater performance while also reducing the number of servers and storage devices required to support today s applications. Using Flash Technology in a Tier 0 Configuration Replacing all HDDs with enterprise SSDs is not economical for most datacenter storage infrastructures. One strategy that is commonly deployed is the use of a Tier 0 configuration that acts as a read cache for frequently used data. Deploying flash technology to hold actively used data in Tier 0 can have a big impact on performance, as long as the most-active data is actually kept in this tier. However, in most cases, there is no automatic way to determine which data sets (and from which applications) will be most actively used and thus should be maintained in Tier 0. As business needs change and different sets of data become the most active, something must be done to recalibrate storage performance. Storage administrators must therefore monitor storage usage patterns and make decisions about which data sets should be maintained in Tier 0. This often involves manually moving data into or out of Tier 0. As mentioned earlier, HSM solutions are sometimes used for this type of data management but have proven to be only moderately effective at reducing the manual effort involved. Thus, although the Tier 0 approach can improve performance, it can add to management complexity and does not eliminate the need for high-performance disk drives that are used for Tier 1. Thus, it provides only a moderate cost savings and can add complexity. 7 Hybrid Storage Pools Provide a New Approach In order to fully utilize the strengths of both SSD technology and HDDs, a new approach is needed. What is needed is a file system that recognizes different types of storage media and can transparently optimize data placement to drive better application and file system performance. With Oracle s HSPs, enterprise SSDs can be used to assist HDDs by caching frequently accessed data to minimize the impact of disk latencies and to improve application performance. Because enterprise SSDs offer low latency, they can generally keep up with CPU I/O throughput, and HDDs can then be used to store massive data sets. HSPs provide the means to automate the process of keeping the most active data in the enterprise SSD read cache while also buffering data writes in a separate cache area. The most significant effect of using HSPs for this task is that it can drive enough performance to enable high-speed Fibre Channel (FC) drives to be replaced by high-capacity drives. These highcapacity drives, which are often serial ATA (SATA) drives, enable dramatic cost reductions in storage acquisition costs. At the same time, storage administrators are freed from the laborious task of ongoing storage optimization efforts. Thus, HSPs give organizations an economical way to achieve significant performance gains without sacrificing capacity, and they also help to reduce storage administration costs. The following section provides additional information about the cost savings when HSPs are used instead of a Tier 0 approach. Hybrid Storage Pools and Oracle Solaris ZFS Unlike less-sophisticated file systems, Oracle Solaris ZFS recognizes different media types and will optimize how it handles each type to help maximize system throughput. Oracle Solaris ZFS also automates the process of data placement so that administration efforts and the risk of human errors are greatly reduced. Oracle Solaris ZFS uses HSPs to provide both simplicity of management and greater flexibility for optimizing performance. HSPs typically include DRAM, a read cache, a write cache, and a highcapacity storage area. Each of these different storage categories may be implemented with different types of storage media, or they may all be made up of the same media. Each category within the HSP storage hierarchy is defined according to its capabilities (such as highcapacity or write cache). It is most common for the high-capacity storage area to be implemented with media such as 1 TB SATA drives spinning at 7,200 rpm and for the read and write cache areas to be implemented with enterprise SSDs. However, there is no reason that the high-capacity storage area could not be implemented with enterprise SSDs as well. In some applications, such as a database or OLTP application, it may be appropriate to use SSDs in the high-capacity area if the data set is large and performance is also a critical success factor. Regardless of media types used, Oracle Solaris ZFS transparently manages the combined storage areas as a single pool of storage. HSPs essentially isolate applications from the underlying physical media, as shown in Figure 3. Applications can thus transparently utilize multiple types of storage media to take advantage of their different price and performance characteristics. When flash technology is used in the 8 read cache area, for example, applications can access the most-active data at speeds that are up to 100 times faster than with HDDs alone. This enables HSPs to deliver the high performance of SSDs while also enabling higher return on investment through the use of potentially different media for highcapacity storage. Figure 3. Hybrid storage pools isolate applications from the physical storage media, enabling Oracle Solaris ZFS to optimize performance across multiple media types. Transparent Storage Optimization In Oracle Solaris ZFS, synchronous writes are first executed to a write cache that typically consists of DRAM and write-optimized SSDs. The writes can then be acknowledged quickly, allowing the application to continue processing while data is automatically flushed to high-capacity storage. Data migration to the high-capacity area is managed in Oracle Solaris ZFS as a background task that is transparent to both users and administrators. Oracle Solaris ZFS also manages the process of copying frequently accessed data into a highperformance read cache. The read cache typically consists of read-optimized SSDs that enable data to be transparently retrieved by applications with very low latency. Oracle Solaris ZFS looks at usage patterns to determine whether and how to use the different storage media. For example, large synchronous writes, such as video streaming, do not benefit from caching, so Oracle Solaris ZFS does not try to copy this type of data to write cache. Similarly, the read cache is populated based on an intelligent algorithm that takes into account not only the most recently used data but also anticipated read requests and estimated data to be held in DRAM. Whereas other solutions generally require a storage administrator to monitor and manage the environment to continually optimize performance, Oracle Solaris ZFS virtually eliminates the need for manual data-placement optimization. With Oracle Solaris ZFS, all data movement between media types is managed seamlessly and transparently, without interventi
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!