Storage 102: Convergence of NAS and SAN

Network Attached Storage

Those NAS solutions were pre-configured boxes that included a server, storage and software and were optimized for simplicity of installation and operation. All customers had to do was take the hardware out of the box, plug it into their IP network, turn it on, and they were ready to go. All right, well maybe it wasn't quite that simple, but it was pretty close--so appliance was the right term. Data was backed up directly from the appliance to the tape drive, preventing backup traffic from hitting the local area network (LAN).

Customers loved their NAS appliances, and when the first appliance filled up, they bought another, and another, and then another. Suddenly, they had multitudes. In an environment with many NAS appliances, the following limitations must be recognized:

Scaling: Once the appliance fills up, the only way to increase storage is to buy another appliance and split the data over the pair.

Backup: Each appliance must be backed up separately, either to its own, dedicated tape drive or across the LAN to a tape library (which defeats the original advantage). More advanced backup functionality, like flash backups, and fast restores are not available.

id
unit-1659132512259
type
Sponsored post

Replication: Replication capability is limited. It uses the LAN.

Availability: If the server in the NAS appliance fails, the whole appliance fails.

Storage Attached Networks
Storage Attached Networks (SAN) also emerged in the 1990's with the development of Fibre Channel technology. As storage array capacity increases to hundreds of terabytes, efficient utilization of that storage requires that more and more hosts be able to access the array. Fibre Channel switches allow multiple hosts to be connected to a storage array without degrading performance, addressing limitations of existing technology like SCSI and UltraSCSI, making continuous application availability realistic by using clustered server configurations. Rather than making data available to servers in file form as NAS appliances do, SAN storage is served up in block format.

During the past two years, Fibre Channel has emerged as an extremely reliable technology that is commonly deployed into production by enterprise IT organizations.

Frequently, initial SAN deployments are project-driven, resulting in a SAN island--a dedicated server or small cluster of servers attached through Fibre Channel switches to a storage array. Multiple projects may lead to multiple, unconnected islands. Ultimately, in order to optimize flexibility, connectivity, and utilization, companies look to merge the SAN islands, so that most servers and arrays are interconnected. Once the SAN infrastructure is in place, new applications are implemented using SAN resources if they are available. If necessary, new resources can be added to the SAN.+

Network Convergence
The approach of 10-gigabit Ethernet opens the door for the next shift in technology. The increased connection speed of IP-based networks will provide an opportunity for SAN and NAS to utilize a single network rather than requiring companies to build and maintain both IP and Fibre Channel networks. However, don't be fooled into thinking that this shift will be either automatic or immediate--storage architects are by nature a cautious lot and take their data very seriously. They approached Fibre Channel technology carefully, implementing in a lab setting and testing thoroughly before moving it into production. They will no doubt follow similar procedures with iSCSI.

While IP network protocols have been built to expect packet loss, storage protocols have not. Packet loss in a storage environment can lead to a significant drop in storage performance. Intermittent network outages are annoying to users. In business critical situations, intermittent network outages are more than annoying; they are unacceptable. The resolution of these and other problems will be key to the success of iSCSI storage.

Consolidation--Servers and Storage
As more and more applications are moved off the mainframe and onto distributed servers, companies find themselves with a plethora of servers resident in multiple data centers spread throughout the organization. The business critical nature of many of these applications requires that they be managed with process and discipline reminiscent of the mainframe environment, which leads to centralization of management, which leads to centralization of resource, which ultimately leads to consolidation of servers and storage.

Companies are finding that it costs less to maintain fewer, bigger servers and fewer, bigger storage arrays that run multiple applications, rather than having dedicated resources for each application. Studies show that acquisition costs represent only 10 to 15 percent of the total cost of hardware ownership. Maintenance and operation consume the vast majority of those costs and represent the biggest opportunity for savings.

Consolidation achieves four goals: