It was the best of storage protocols, it was the worst of storage protocols. It was the age of SAN, it was the age of NAS. Okay, I’ll stop with the Tale of Two Cities spoof. This post will cover SAN and NAS, the characteristics of these storage protocols, and why someone might pick one protocol over the other.
All About Those Blocks
SAN (Storage Area Network) is a block-level storage protocol, meaning data is stored and accessed in blocks. All of these blocks of data are stored in Logical Unit Numbers or LUNs. LUNs are then presented to an initiator (host) from a target (storage system). To an end device like a server, these LUNs look like direct-attached-storage (DAS) or a local drive. But unlike with DAS, you get portability and a slew of other benefits from centralized management.
A caveat to block storage is that you need to format the drive with an operating system-specific file system. For a Windows server, you’ll probably be using NTFS. Another caveat is, depending on the file system, you may be limited to having a single host read and write data. However, a clustered file system like VMFS (vSphere datastores) handles locking for writes from multiple hosts to prevent data corruption.
If you serve data using the SAN protocol, you may have specialized hardware for transporting storage area network traffic. For instance, if you serve your SAN data using Fibre Channel Protocol, you have Fibre Channel switches and HBAs (host-based adapters) dedicated for this traffic. If you are using iSCSI for transporting this traffic, you can get away with using the same switches and network cards that you use for your network traffic. However, you probably will have specialized converged network adapters (CNAs) on your hosts that offload some of the processing from the host’s CPU. These CNAs offer improved performance.
Separate hardware is actually an advantage to using SAN as a storage protocol. Typically, storage traffic is segregated from general purpose network traffic. Because SAN offers fast, resilient storage, it is frequently the storage protocol of choice for databases and other high-performance applications. With the advent of the NVMe protocol and development of NVMe over Fabric, the performance benefits of SAN storage protocols are game-changing.
Files, Files, Files Baby
Network Attached Storage (NAS) is more abstracted than SAN storage. From this abstraction comes one of NAS’ biggest advantages: simplicity. Instead of blocks, data lives in volumes that are made up of files and folders. It’s the typical client-server architecture. Clients access folders and files from a UNC path that the NAS server shares. Because the NAS server handles the file system, data is completely independent of the devices to which they are connected. Multiple clients or devices can connect to the same files simultaneously.
NFS, SMB, and CIFS (legacy SMB 1) all fall into the NAS protocol category. NAS is a storage protocol of unstructured data. File shares and home directories make up a lot of this unstructured data. NFS datastores comprise another part of NAS data.
Advantages to using NAS protocols include easier management and the ability to leverage your existing network infrastructure. Another perk is that because data is stored at the file level, you can easily shrink and autogrow volumes. The ability to shrink volumes to reclaim over-allocated space is a pretty big benefit. When using SAN as a storage protocol, you really can only grow those LUNs. To get a smaller LUN, you would need to re-create that LUN and move all of the storage to the new smaller LUN.
SAN and NAS each have their own benefits and trade-offs. Like most IT questions on which approach to take when selecting a storage protocol to use, the answer is, “It depends.” Fortunately, it’s not an either-or question. NetApp doesn’t limit you to using only SAN or only NAS. NetApp pioneered the unified storage architecture over a decade ago. Choose what works best for your data—that’s what NetApp is all about.