The Sun Storage F5100 flash array is the world’s first solid-state flash storage array delivering more than 1.2 million read and write IOPS and up to 2 TB of solid-state capacity in 1U of space. That’s the equivalent of more than 3,000 enterprise disk drives in I/O performance while using less than three lightbulbs (300 watts) of power. A single rack of systems can deliver more than 50 million IOPS with more than 80 TB of capacity. This blazing performance can help you accelerate applications while improving productivity and ecoefficiency.
The Sun Storage F5100 flash array redefines storage performance and efficiency with an incredible 1.6 million read and 1.2 million write IOPS in just 1.75 inches of rack space. It sets new benchmark records for IOPS per dollar, IOPS per watt, and IOPS per space. I/O-intensive database applications with heavy I/O read and 4 K block-aligned write workloads can take advantage of the Sun Storage F5100 flash array to accelerate their performance, improve response times, and increase transactional scalability, while also reducing power and space costs.
The Sun Storage F5100 flash array offers a breakthrough in storage economics with industry-leading price/performance, power/ performance, and space/performance efficiency. By using the Sun Storage F5100 flash array to address your performance needs, you can supplant costly and inefficient 15 K rpm disk drives, which drive up power and storage costs. This can help you significantly reduce both storage costs and operating costs, resulting in much-lower TCO.
The Sun Storage F5100 flash array provides a very high level of reliability with an all-solid-state durable design consisting of 80 nonvolatile enterprise-class single-level cell (SLC) flash modules, each with MTBF exceeding that of enterprise disk drives. The system uses sophisticated wear leveling, advanced bad block mapping, and enhanced write endurance to help ensure the highest level of reliability and longevity. Redundant power and cooling helps reduce the risk of downtime, and integrated super capacitors help prevent interruptions in write operations should a power failure occur. Host-based mirroring can also be used for mirroring internal domains or individual flash arrays.
The flash array takes advantage of data integrity features in the Oracle Solaris Zettabyte File System (ZFS), including ZFS RAID, automatic data integrity checking, and correction with block-level checksums. When a corrupt block is identified, self-healing features in Oracle Solaris ZFS will direct another copy to be written as an automatic repair, thus helping to prevent silent data corruption.
The Sun Storage F5100 flash array appears as a normal storage device enabling easy deployment with instant performance benefits into new or existing environments, whether using the Oracle Solaris operating system, Windows, or Linux. The flash array can be managed together with other Sun storage arrays using Sun StorageTek Common Array Manager (CAM) software, which provides a common, simple-to-use management interface. Sun StorageTek CAM’s fully integrated Service Advisor software offers proactive health checking and supports quick time to service. System error messages are connected directly to specific repair procedures in the patch knowledgebase, making common repairs simple and easy.
Today more and more applications, especially databases, are being choked by disk drives that can no longer keep up with CPU performance, causing latencies and I/O bottlenecks. The traditional approach of using large numbers of mechanical disk drives to address growing storage performance needs can greatly increase power, cooling, and space costs. IT managers are looking for more-cost-effective and highly scalable storage solutions that can quickly accelerate application performance while also reducing operating costs.
Many of today’s enterprise applications require a high volume of I/O throughput to deliver the high service levels needed to keep users productive. Oracle database applications, such as enterprise resource planning solutions, often serve thousands of users with a single central database instance that is accessed for each transaction. High-performance computing and financial trading applications represent additional categories of applications that require low-latency access to large data stores to support adequate application performance. Scaling to meet the performance requirements of these applications can be a very costly undertaking using traditional storage architectures.
For I/O-intensive applications, performance is not bound by the CPU, but by I/O throughput. Disk access speeds remain one of the major bottlenecks. While CPU performance has been doubling every year with Moore’s law and continues to grow rapidly with today’s multicore CPUs, disk drive performance has not kept pace.
Seek times on today’s mechanical disk drives are more than 200 times slower than today’s servers, causing application performance to suffer from storage latencies and I/O bottlenecks. To compensate for this performance gap and slow seek times, application data is often spread across a large pool of high-performance 15 K rpm disk drives that are “short stroked” for better performance.
This effectively multiplies I/O throughput by enabling read and write operations to take advantage of multiple spinning disks. However, it also results in inefficiencies, because the need for performance results in configurations with more disk drives than would be warranted based solely on capacity requirements. The high power requirements of 15 K rpm drives and the poor space use of partially filled disks also result in unnecessary datacenter power, space, and capital costs.
Another approach to compensating for slow disk access speeds is to deploy a large buffer of DRAM, so that an entire application’s working set can be stored in memory, thus reducing latency. Although DRAM offers very high performance, it is also quite costly and requires a battery backup due to its volatile nature.
Both of these traditional storage architectures are costly to acquire and operate. Furthermore, as CPU performance continues its exponential growth, these architectures will eventually become impractical. A new approach is needed that can meet today’s demanding storage performance requirements at a minimum cost and with minimum power, space, and cooling requirements.
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 tiered storage. Solid-state flash and SSDs 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 hard disk drives (HDDs).
Reliability characteristics of enterprise-class flash and SSDs have also improved, yielding mean time between failures (MTBF) ratings that exceed those of HDDs. Like HDDs, enterprise SSDs also support bad block management, wear leveling, and error correction codes to foster the highest level of data integrity and reduce service downtime. The solid-state nature of flash allows enterprise SSDs to withstand significantly higher shock and vibrations than HDDs.