New horizons in Enterprise Storage: NAS gateway precursors SAN/NAS convergence. (Cover story).
NAS and SAN in the Enterprise
As storage needs grew in the enterprise, network-attached storage (NAS) devices, or filers, became a popular alternative to direct-attached and network storage. Well-funded data centers could afford high-end NAS filers from NetApp or EMC, which were scaleable and provided sophisticated management options. But departments, workgroups and smaller divisions bought quantities of economical and simple NAS devices to handle their file storage. NAS was very popular--so popular that it created a giant corporate headache for IT groups charged with centralizing storage operations across the enterprise. And as NAS proliferated within divisions, the local IT departments began to struggle. Many of them were not used to managing big chunks of storage with provisioning and different operating systems, and few NAS devices lent themselves to centralized management.
No one wanted to replace NAS with SAN--the two serve distinct and valuable purposes. But storage administrators struggled with managing both types of storage. They have distinct management interfaces, different (and often incompatible) backup procedures, cannot share capacity, cannot load-balance application workloads across each other, and migrating application storage between environments is challenging.
SAN and NAS are similar in some ways: They are storage devices that combine multiple disks into arrays for availability and performance, and usually have a buffer cache to provide additional performance. NAS is excellent for providing heterogeneous and simultaneous file sharing, but suffers from network protocol overhead: The protocols maintain a large amount of native error-checking and integrity features. The overhead can impact NAS performance. This is not an issue with channels or SANs, so using SANs for larger transfers offers high performance and more predictable response. SANs lack NAS file sharing capabilities.
SANs live on dedicated networks, most commonly Fibre Channel--Fibre Channel provides any-to-any connections between hosts and storage. Like DAS, I/O requests access devices directly. SANs are optimized to handle storage traffic at high speeds with single points of control, and offload significant processing tasks from the LAN. SANs divide the storage up into pieces called logical units, or LUNs (Logical Unit Number). LUNs are made to be manageable:
* They're relatively large so there are typically just a few per zone, which makes LUN management a simpler prospect for zone administrators. For example, a LUN may be one entire array, or a very large array may be divided into two or three LUNs.
* They are similarly sized and can only be accessed in whole blocks, which speeds up data transfer. Using command level interfaces, wizards, or semi-automated procedures, IT administrators map LUNs to blocks on the disk subsystems.
* Block-level data on the LUNs view the SAN as direct-attached storage, making SANs a good choice for database storage.
NAS consists of a specialized processor with its own disk storage. It attaches to the LAN or WAN using specialized file access and sharing protocols, and uses its processor to service file requests instead of allowing direct access to its storage. NAS is optimized for simple management and file sharing across workgroups and platforms, and is a wild and woolly environment compared to the staid SAN. NAS systems differ from SANs because they present their storage as a set of files instead of LUNs:
* Files are more dynamic than LUN's steady data blocks--they come in radically different sizes and are easily created, modified or deleted.
* This dynamic and hierarchical environment supports many thousands of files instead of tens or hundreds of LUNs.
* NAS developers configure the systems to handle sophisticated space management and performance using full file systems, a very different environment than SANs.
Ken Steinhardt, EMC's director of technology analysis, believes that application requirements will ultimately dictate NAS or SAN choices and spending: "NAS is very good at doing many, many, many small things. SAN is inherently better at doing large things very well. It is really at the application level that the functional difference and the use for either will come into play." Since both storage-networking technologies serve application requirements, it makes sense to consolidate them as much as possible. Today's most common convergence technology is a NAS gateway.
A NAS gateway forces the two into a workable partnership and increases the value of both. The gateway consists of a NAS device with a Fibre Channel HBA (host bus adapter) added to it, allowing it to connect to the Fibre Channel SAN. NAS retains its unique file sharing capabilities, but its storage becomes more flexible and manageable on the SAN. For example, a company installed EMC's Celerra NAS/SAN gateway so it could back up file-level NAS data to its Symmetrix arrays. This sped up backup considerably and the company could manage the NAS gateway as well as the SAN through a single interface. Other SAN/NAS gateway vendors include Auspex, HDS, LeftHand Networks, IBM, LSI, NetApp and Xiotech. They are far from perfect--for example, EMC, HDS and LSI gateways can only store data on arrays from the same vendor, and IT managers require the server interfaces to look at file-level data on the SAN. But gateways are clearly an improvement over chaotic collections of NAS devices.
NAS gateways help storage administrators to centralize distributed storage, and users do not have to give up NAS advantages. Ruth Colombo, Veritas' ServPoint marketing manager, put it this way: "The storage group is widening its responsibility to include both distributed as well as centralized storage. So once they increase their responsibility, they start looking at those appliances. Wait a second, this is too costly to manage! And then they say, Why do I have to treat my storage differently from my other storage? They're already used to managing arrays, to provisioning storage. They are not intimidated by a NAS gateway."
Gateways enable storage administrators to reallocate storage between NAS and SAN. They also enhance the value of an existing SAN by allowing it to service NAS applications through the gateway. This is a clear advantage to backup operations: Without a gateway, most NAS devices must be backed up separately. By using NAS gateways they can be backed up to the SAN using the same backup application that IT is already used to. This not only makes for a more effective use of backup resources, it also makes it possible for companies to consolidate their storage. Instead of buying three small arrays for three NAS devices, the company can buy one larger array for multiple NAS gateways. (When an IT administrator views the SAN's zones, it presents the NAS-deployed storage as DAS. Since NAS cannot share this disk space, a large array must be carved into three separate LUNs--still an improvement on individual back-up procedures and devices.)
There are disadvantages to NAS gateways. A gateway does not truly converge NAS and SAN, so companies must still spend money to store file- and block-based data on different devices. A NAS gateway can still suffer from the same latency and bottleneck issues a NAS appliance might suffer, and requires extra data movement--data must move from the SAN array to the NAS and to the requesting host. However, fast networks and sophisticated arrays lessen data movement and latency problems.
NAS gateways give users the ability to have both file and block access, without having to choose between the two on a given storage device. Russ Bailey, Dell's senior manager of enterprise storage product marketing, reports that NAS gateways benefit companies that are looking to consolidate storage. "Customers have been telling us for some time that they're tired of managing separate storage silos, the SAN environment or a NAS environment, or a SAN environment or a DAS environment. They're tired of making separate storage decisions between a filer or SAN storage array. They're looking to combine the two, and make it easier to add storage and to manage storage." Bailey sees value around ROI and consolidation discussions. "It's really based around a return on investment and storage consolidation discussion. Reduce the number of software management packages you have, because we're going to reduce the types of storage you have installed."
NAS gateways are a precursor to true NAS/SAN convergence. But vendors differ on the form that convergence will take: It may be transparent gateways that can share files on the back-end, hybrid devices or global file systems running over storage area networks. Whatever the path forward, John Toor, director of product marketing at ClariStor, said that they'd like to move beyond the fact of a NAS device (or server, for that matter) connected to dedicated storage. "What we're trying to move away from is this idea of having storage tied to this particular serving device."
Convergence--gaining any-to-any connectivity with NAS as well as SAN devices--is the way to get the full benefit of the SAN. In the meantime, the NAS gateway is a solid interim step.