Sanbolic’s products create a foundation for networked storage that enable storage to be unified into a single pool, and then be flexibly allocated and shared. The logical architecture of storage resources can be greatly simplified, storage can be dynamically allocated to applications, multiple users can be given concurrent read and write access to files, and data can be more effectively managed over its lifecycle. The inherent redundancy within storage networks can be fully utilized because it is no longer necessary to have specific disks accessible only through a single server. Sanbolic’s products enable users to build a highly reliable and flexible single storage pool from modular components.

The rapid growth of storage volume over the past decade has been accompanied by a shift in storage architecture from direct-attached storage (DAS) to networked storage. In a direct-attached storage environment, each server utilizes disk drives within the server enclosure or in a nearby dedicated storage array. Network storage, on the other hand, separates storage from applications servers, often physically centralizing the storage. This facilitates storage management and growth, and allows multiple servers to share storage hardware.

Sanbolic’s products enable all computer systems connected to the storage in a SAN to share a single file system, which greatly simplifies data storage management, increases the flexibility to easily reassign tasks among computers, and allows shared access to information in a reliable and secure environment. This becomes particularly important as more customers shift to using clusters of small, inexpensive Intel-based servers to replace large proprietary Unix servers.

Changes in Storage Architecture Create Need for New File
System Capability

Direct Attached Storage

• Each server has its own storage hardware
• Separate file system for each server
• Storage-server “silo”— data is “owned” by a single server

SAN Attached Storage with Local File System

• Servers share storage hardware
• Separate file system for each server
• Storage-server “silo”— data is “owned” by a single server

SAN Attached Storage with Melio FS

Limitations of Local File Systems

In almost all network storage installations today, each application server maintains its own local file system. Although SANs enable sharing of storage hardware, typically each server on a SAN needs to be provisioned with unique disk space on the storage array, which is managed by the server’s file system. It cannot directly share its data with another server on the SAN. Although storage is physically consolidated, logically it still looks like a complex direct attached environment. Each server has a separate file system to maintain, and none of the servers have a single view of the entire storage system. Each NAS server also maintains its own file system.

Single-node file systems create an inherent complexity and scalability issue in large installations, since each file system requires management and data is not easily shared. Hence, IT infrastructure consists of a collection of relatively inflexible storage/server/application “silos”. Applications grow by replacing small servers with large proprietary multiprocessor servers, which typically don’t take advantage of the cost advantage of the extremely high volume Intel/AMD-based servers. When clusters of smaller servers are used, they need to create a separate copy of all data for each server. This adds complexity, large quantities of redundant storage, and a lot of network and server overhead for file replication which offsets much of the cost advantages.

The inflexible storage/server/application “silos” result in very limited ability to quickly reassign resources in respond to demand changes, for example for a month end closing or an event that creates a spike in web server demand. Hardware is provisioned for peak demand, resulting in poor average utilization. Sun has published numbers showing that 10-15% server utilization is not uncommon.

In addition, the fact that each server maintains its own file system adds great complexity to the task of protecting data in the event of hardware failure, in effectively managing data in response to regulatory compliance requirements, as well as in managing the cost-performance trade-off of how data is stored as its value changes over time. Today, many of the tools used to manage data are either deployed on each server and work in conjunction with that server’s file system, or are deployed on a storage array and manage the block based data on that one array. Many of these tools are proprietary, incompatible, and expensive. While the industry has put significant effort into improving compatibility, the fact remains that many environments have hundreds of heterogeneous servers and dozens of heterogeneous storage arrays.