Ethernet Sans: The best choice for enterprise-class storage and cloud

Business needs force IT technology to move forward, but businesses cannot afford to waste in order to ensure return on investment. This forces it to become a model of survival of the fittest, and hopes to get rid of data center performance can not http://www.aliyun.com/zixun/aggregation/8869.html "> meet the needs of products- This makes the current hot storage in the wind wave mouth above.

Two trends in the mainstream industry are continuing to drive enterprise-class storage for change. First is the enterprise, the rapid growth of data and the growing complexity of the analytical needs of the Government and the science sector have created a new noun-large data that includes sensor data, photos and videos of social media, satellite imagery, genome sequences, and data from various mass and modern equipment. The organization found that their data storage requirements were growing at a rate of more than 50% a year; the second trend was that traditional storage patterns were evolving towards a highly resilient cloud computing model. Businesses need the IT department to be more agile and want to provide immediately available applications and data on demand.

These two trends put enormous pressure on the traditional storage systems designed for the mainframe era, which are not well adapted to the dynamic virtualized environment of the cloud and are not designed to meet the data growth requirements of more than 50% a year. Storage costs have increased dramatically, with most organizations storing budgets that reach more than 40% of the overall it budget, and administrators increasingly grappling with the complex management requirements of traditional storage networks. These requirements are all pressing for the storage industry to require a redesigned shared storage architecture.

With the continuous development of the Internet, the server industry has undergone the same painful transformation. The rise of various web sites, online searches and E-commerce forced the server to expand, and the mainframe was quickly replaced by a scalable X86 platform that eventually formed a virtualized server. This shift fundamentally changes the economics of computing, lowers costs, and allows the scale of computing to grow linearly with business growth. An Ethernet SAN brings the same vision to the storage industry. With the development of Ethernet and standardized hardware platforms, Ethernet Sans provide a flexible, scale-out storage architecture that eliminates the complexities of traditional San tiers. This allows organizations to cruise and expand their storage infrastructure without the heavy management burden of initial inputs and traditional storage.

The Ethernet San leverages the most Pervasive data center network Technology-Ethernet-to provide better performance than traditional 10GbpsFCSAN. And, this advantage will only increase, because 40Gbps and 100Gbps Ethernet technology is already in the planning. In addition, the economics of the bulk data centers provide a huge advantage, and the cost of Ethernet will continue to fall, far below Fibre Channel. Ultimately, a large, parallel connection is eliminated, eliminating the complex multi-level, and easily scale-out, Ethernet sans of traditional sans. Simply adding storage devices to a storage network increases capacity and its performance grows linearly as each storage device has its own processor.

In contrast, traditional storage networks rely on the fixed data storage and initial design size that each server connected to the switch can provide. As a result, when storage requirements change, administrators have to manage complex, multi-tiered storage, including multiple paths, aggregation ports, switch partitions, controller load Balancing, and array management across multiple tiers of workloads. This approach can only manage a certain number of devices, but the size and pace of growth in modern data centers makes traditional management more difficult to maintain.

In addition, the dynamic virtualization workload based on the controller design is also one of the challenges. A large number of small block random I/O and multiple virtual machines share a connection, resulting in hot data problems and performance problems are very difficult to solve. The only solution to the performance or capacity constraints faced by many traditional arrays is forklift-type upgrades, while larger-scale storage involves more capital and operating expenses.

(Responsible editor: The good of the Legacy)