Editor's Note: We asked Randy to share his top five predictions on the future of the NAS market with our readers earlier in the month. We also asked him to elaborate here about the background behind each of his predictions. (If you would like to know more about NAS, please join Randy for our LiveExpert Q&A Online Event on the future of NAS, this Friday, on January 25, 2002.)1. DAFS entering the NAS market will create new levels of performance with database software that is VI enabled.
The first major vendor to offer a NAS product in this area will be Network Appliance with an update to their Data ONTAP operating system for support of DAFS. Being a major force in the NAS marketspace, Network Appliance will command attention and be able to install DAFS appliances in customer environments. There are two methods for capitalizing on the use of Remote Direct Memory Access employed by the Virtual Interface Architecture which is the underpinning of DAFS: a layer to change a file system access to a remote DMA direct into that address space or VI enabled applications. Both will provide improvement over the normal execution using a redirector and the transversal of the TCP/IP protocol stack with its multiple memory moves of data and interrupts.
Applications coded to specifically do the remote DMA direct into their own address space will see the greatest performance gains. Given that the challenge for NAS has been to be used for high-transaction database operations, the database software that has been changed to take advantage of the new technology will benefit the most and will add potential customers for DAFS enabled NAS that were out of reach previously. More than one database vendor has made these changes so when DAFS enabled products are released, you can expect to see a great deal of activity here.
2. NAS vendors will start selling TCP/IP accelerators (sometimes called offload engines) with NAS products to provide significantly enhanced performance.
TCP/IP accelerators take much of the function of handling the TCP/IP protocol stack and move it from being executed by the processor on the server or client to an adaptor card (called either a NIC or Storage NIC or HBA depending on your orientation and who you want to confuse). The functions still need to be performed that include check summing, acknowledgements, segmentation and reassembly, etc. but now they are done in a special purpose piece of hardware. Will that special purpose piece of hardware be as inexpensive as network cards? No. Volume drives cost when it comes to hardware. It will take a long time and a very successful rate of adoption to get the cost down.
Will all TCP/IP accelerators from different vendors be the same? Not initially. There will be differences with a new implementation. There always is but that will settle out over time. Much of the publicity around the TCP/IP accelerator (or offload engines if you want an ugly acronym) was focused on trying to improve iSCSI to have acceptable levels of performance. The often over looked jewel here is that they can provide significant performance improvement for NAS devices. These will be delivered to customers soon and will improvement the weakest performance area.
3. SAN and NAS convergence will continue. Driven by administrative cost issues and the usage of NAS in enterprise environments by storage professionals, various forms of convergence will occur. The most popular will be a NAS gateway or HEAD providing file access into data managed within a SAN.
More vendors will announce products in this area and every major system vendor will have a NAS gateway offering in their portfolio for storage networking solutions. The big win here for customers will be when an integrated software package can manage multiple NAS gateways (from different vendors) on the SAN along with the SAN block level storage. With the increasing deployment of virtualization solutions, including the NAS gateway, management will yield the greatest gains for customers in administrative reduction. The first vendor to come out with a complete solution here will have a significant advantage over the competition.
4. NAS management aggregation becomes more predominant with tie-in to SRM packages.
Automation of storage management driven by policies is where we're headed. Incorporating NAS (with support for heterogeneous NAS devices) into that automation driven by a higher-level Storage Resource Manager product will provide the greatest economic gain in administrative cost reduction for customers.
5. Confusion over when to install NAS or SAN continues. Understanding the application and business requirements is the issue that a salesman probably can't help with.
The sale is not cookbook. A good salesman works with the customer to understand the requirements, both current and future. Unfortunately most salesmen aren't given the time (or may have the capability) to work with the customer in a long-standing relationship to attain that knowledge. We'll be left with the customer really having to understand the issues and what technologies are best. Educational needs for customers will continue.===========================================================
To view all Storage Evaluator analyses and commentaries, go to Storage Evaluator Tips ===========================================================
The Evaluator Series and Evaluator Series On-Line (ES/OL) are trademarks of Evaluator Group, Inc.