ANALYST'S CORNER

Revisiting NAS and iSCSI SAN

When network attached storage (NAS) and storage area networks (SAN) were first introduced to IT networking, press releases examined these options and in many cases contrasted the two in an "us versus them" philosophy of coverage. From a journalistic point of view, the adversarial process has a way of making people think and perhaps discover something that they hadn't considered before. After that, with editorial wisdom, the reader would vote with dollars based on which would answer their need.

The well publicized battle ultimately resolved itself into a clear difference: NAS was based on file-level transactions, and SAN was based on block-level transactions. Effectively, NAS is a file server front-end to a SAN. Thus, product differentiation was realized, and the dramatic battle faded out of memory as storage vendors looked at other competitive targets.

Then iSCSI came along and put a dent in the very neat compartmentalizing of SAN and NAS. iSCSI is a disruptive technology in that, unlike Fibre Channel, iSCSI and NAS operate over the same type of network and therefore serve the same business need. The question to be resolved, then, is whether an iSCSI SAN can do the traditional job of NAS better than NAS can.

Technical Details

Although SAN and NAS can both run over existing TCP/IP networks, there are differences. Figure 1 illustrates that NAS is essentially a file server with dedicated storage - either DAS or SAN. A NAS box interfaces to the LAN and provides managed and provisioned storage to client hosts, whereas a SAN box interfaces to the SAN and provides managed and provisioned storage to servers. Furthermore, Figure 1 also shows that SAN is flexible in supporting all server storage, including dedicated NAS storage. A NAS gateway can utilize a SAN just like a file server, and an allocated SAN volume can be dedicated NAS storage.

NAS is technically not simply "storage attached to a network." SCSI storage hardware block I/O occurs between the NAS processor and the storage arrays to which it connects and not to the actual network. Rather, it is the file serving that is network-attached. A NAS processor receives a file request, works through the file system's metadata for the target file, identifies the blocks on disk that make up the file, reassembles the file, and encapsulates the content in TCP/IP for network transaction. NAS differs from the file server basically in that the file server OS is stripped of non-essential functions and sends files via IP protocols like NFS or CIFS. A NAS filer sends files onto the network while doing block SCSI I/O in the background.

SAN sends block I/O directly onto the network infrastructure and does not have the file assembly steps that the NAS device must support. This dramatically improves SAN performance over NAS, as the file system processing remains on the server and is not replicated in the storage device. iSCSI SANs encapsulate SCSI commands, status and data over TCP/IP.

SAN also provides a security advantage, as SANs use zoning and LUN-based security tools to control server access to the networked storage. NAS security is typically at the file system level through user access control utilizing the domain server and an LDAP directory. Administering server security is much simpler than user security, because there are typically an order of magnitude fewer servers than users. Some SAN implementations also provide more flexibility for the SAN, in that the storage administrator can mix RAID levels within the same device to match application performance and data protection requirements.


Traditionally, an important differentiator had been cost when deciding between SAN versus NAS. iSCSI's cost advantages over Fibre Channel have changed the playing field for making storage decisions. It is also important to consider that NAS is a dedicated storage device and may not support all storage solutions. Many networked storage installations incorporate several NAS devices, including NAS gateways. Clustering multiple NAS devices may be required to increase storage bandwidth.

SAN can be leveraged to support all networked servers, including NAS gateways. Thus, the decision point evolved from comparing just a single NAS appliance to a single iSCSI RAID device when considering how to grow a storage network for cost efficiency, simplicity and flexibility.

Simply stated, you can leverage an existing SAN to act like a NAS with a low-cost server appliance and save some serious money.

Likely Outcome

In the end, the winning value proposition may be cost. But if NAS and SAN follow the history of other hardware storage options, complexity will also influence storage purchasing decisions. The drive toward simplicity is not a strategic business decision, but a very practical and tactical consideration as complexity costs time and money. The tremendous amount of information spawned by the growth of data on networks must be managed. Simplicity has become the pivotal measure of business value, and in the storage networking world, simplicity is another name for iSCSI.

iSCSI SAN decouples the capacity growth of storage from computing resources, providing flexibility in addition to simplicity. IT managers can independently add servers to handle user loads or storage to handle increased data growth. SAN enables provisioning, allocating and managing storage independent of the server resources, thus creating a more efficient use of storage with increased performance.

Whereas NAS couples storage to file services requiring addition of discrete filing components for expanding capacity. NAS adds another layer of file services between the network and storage interfaces, decreasing bandwidth and increasing latency.

Separating file servers from storage devices, providing decoupled capacity growth, and sharing storage services across all networked servers using iSCSI SAN is simply more flexible and cost-efficient than adding discrete NAS components on a more ad-hoc basis.


» Return to the front page