Enterprises are looking to NAS systems to solve administrative problems concerning file serving. In Part one of this series, we discussed how NAS systems are moving higher in the enterprise hierarchy in deployment and described the need for a high-end NAS system to meet those needs. We also discussed the high-end NAS system characteristics needed for scale, administration, availability and features. With that foundation of understanding, we'll examine various approaches to provide a high-end NAS system for those enterprises.
Differing implementations for high-end NAS systems
Vendors have taken several approaches to produce high-end NAS systems and they vary greatly. All of these implementations lead to different characteristics in administration, price, maximum scalability (capacity and performance) and implementation ease (time to deployment). Each product would need an evaluation before a customer makes a decision on which direction to go in, but there are some considerations for the number of implementations that can be pointed out.
The complexity of clustering is also an issue. Is clustering a specific implementation for that NAS product, transparent to the customer for administration and operation? Is it a standard server cluster software implementation with additional administrative tasks and potentially some limitations for storage failover transparency and speed? The interconnect technology may also have some implications in performance (especially failover and failback speed) and extensibility.
Very large NAS systems are usually devices that have been built much larger internally with custom hardware or much more processor power for server-based products to avoid clustering. Availability is usually achieved with the pairing of these large systems, but some may have redundant components internally to achieve fault tolerance. These have the advantage in that they just look like a "bigger" NAS device and have the same administrative efforts. There will be some limitation on the amount of scalability, and the granularity of increment from one system to another may be very large and expensive.
Federated nodes are similar in concept to a clustered system but the communication and data movement between nodes is implemented quite differently. These federated nodes allow more nodes to be attached non-disruptively and data and access distributed dynamically for performance and capacity scaling. Many vendors are beginning to call these products a storage grid. These relatively new implementations appear to have great potential in scaling and presenting a single image. The differentiating factor initially appears to be the extent of self-management of data that is included.
Consolidation boxes are another NAS device that can sit in front other NAS devices to achieve a single image where the consolidating NAS device distributes data across the attached NAS and manages those devices. The amount of device management that still remains for those attached devices may be more than initially desired by a customer and there may be other variations in performance and feature support. As a consolidation platform, this may be a cost-effective product for many customers.
First and foremost, customers need to have a storage strategy. The storage strategy is based upon the business requirements, both immediate and those in the future. This strategy can be used to plan for a high-end NAS system that will meet the requirements. Part of the strategy includes the requirements for long-term support from a vendor and the risks that a customer is willing to take with new technologies or products.
References for a product are the most valuable information for a customer and sometimes the hardest information to obtain. They should be pursued and used to judge real value of the product.
Incorporating new technologies usually has economic advantages over time that greatly outweigh the risk that new technology brings. The values derived from vendors' products vary greatly. Evaluation is necessary in very critical high-end enterprise environments.
For more information:
About the author: Randy Kerns is vice president of strategy and planning for storage at Sun Microsystems Inc., and covers storage and storage management software including SAN and NAS analysis. He has been in the computer industry for over 30 years.
This was first published in September 2005