It should be easy to understand what it means to run storage management tools on a storage network switch, right? Wrong. In fact, there are already 10 or more different terms to describe these new "intelligent switches" and what they will be able to do down the road. You won't have to wait long for feedback, because a number of them are rolling out this year, and even more are becoming available in 2004.
This column attempts to tackle this issue by trying to ferret out what vendors plan on delivering, as well as discussing the good and bad news about these products.
Which brings up my first point. My wife and I went car shopping this past weekend, and like good consumers, we compared prices and features of a number of cars and visited a number of dealerships. As I am sure you are already too aware, we got the full-court press on why we should buy a car at each dealership. Of course, they all claimed to have the best price, but when we started reading the fine print, the truth started coming into focus. We found that some dealerships charged more for certain delivery fees. We also found that others that originally looked pricier were actually less expensive.
One dealer had either the hubris or the stupidity (I'd opt for the latter) to tell us that the 2004 cars weren't available - just after we had test-driven a 2004 model at another dealer of the same manufacturer. It just goes to demonstrate the importance of "buyer beware."
The same goes for these intelligent switches. You need to look at each switch on its own merit and examine the features that best fit your current environment. Here are some key points to consider in your evaluation process.
First, understand that for the short term, vendors will use different terms to mean the same thing when describing their intelligent switches.
I'd call these new devices "storage application switches" or "Layer 7 storage routers." Neither term will likely stick, but they best describe the core missions of these architectures, which is to migrate storage management functionality from a number of different levels onto a common switching platform. This platform will add these software tools in a network service approach to route data and user access.
Whether we like it or not, as the storage networking market has matured, we are seeing more and more enterprise networking environments in which data services are layered from Layers 1-7. As a result, more advanced network services, such as Voice Over IP, SSL acceleration, VPNs and load balancing, have been baked into enterprise networking equipment.
As you go up the food chain with additional software services in the storage network, you move up the layers. Today, most of the early platforms are really more like Layer 4 storage routers, but longer term, they will look more like Layer 7 storage routers. This is because of the software stack and the robust hardware that is built into the platforms for running applications in the network.
Get a sense of the software tools these intelligent switches will support and prioritize where you want to start prototyping in 2004.
What software tools will show up here? At first, you will see core infrastructure services such as storage virtualization, volume management, LUN masking, file aggregation and element management. In fact, a small number of small vendors are already offering such services.
What will be really exciting is the emergence of more advanced data management functionality in 2004. This will include advanced policy-based backup (rapid recovery services), data replication and a new class of applications I call "data mobility" that will be essentially data movers between storage devices.
Data mobility will become a core part of vendors' information lifecycle management strategies. These strategies are based on policies that shift data of a certain age and application from one storage system to another as the data grows older and migrates in value from operational to referential.
You might also see more advanced policy-based storage resource management on these platforms later on, but short-term success really depends on how well storage networking and software vendors work together to gain customer trust and momentum using this approach. Which brings me to my next point...
Evaluate the differences between the various storage networking platforms that come to market in support of these services.
I firmly believe some storage networking vendors are engineering their platforms the right way to support these services, while others have rushed to market without thinking the problem through.
I wouldn't want to name which vendors are not on the right track at this point, because we are so early in the market, and customers haven't had a chance to look at all the various options. But there are a number of factors you need to consider about the architectural approaches storage networking vendors have relied on. For this reason, it is a good idea to make a checklist that includes the following points of comparison:
Hardware platform architectures: There are a number of different approaches that use line cards that insert into the system, appliances that have networking capability and hybrid switches with advanced software services running across the system. The jury is out on which is the best design, but clearly one that provides serviceability, upgrade capability and robustness will be important. I believe that line cards and blades that fit into a switch chassis will ultimately win out.
Performance: Wire-speed performance running storage management services on these platforms is a requirement, but in a lot of cases, vendors still need to provide it with real-life workloads. Some have started to release benchmarks, which is encouraging.
Deployment scenarios: Customers need to understand that some of these platforms will require varying levels of changes within their SAN fabrics, so ask vendors to give you detailed deployment scenarios before committing to a specific platform.
Application services supported: Storage networking vendors, especially larger ones, are actively working with numerous system and storage management vendors to design support for their routers and switches. The challenge for customers will be deciding which service should be prototyped first and then evaluating vendors that provide those services. Volume management, data replication, rapid recovery and provisioning are here now, with other management services to follow. A number of customers I have talked to suggest provisioning and data replication (heterogeneous!) are the most appealing services on these platforms.
Cost: It is really unclear how vendors will charge for software services on the switching platforms in the future, but in the short term, the cost per port generally comes in at a director-class pricing level or slightly higher. This expense could offset software and hardware costs associated with deploying the same management services on a host or array. An example would be data replication.
The important take-away is to consider where this new generation of intelligent switches will fit best in your environment. Be cautious about deployments, but take heart, because I've already seen a number of Fortune 500 companies that have started to prototype first-generation products.
- Jamie Gruener is senior analyst, Enterprise Computing and Networking at the Yankee Group in Boston.