12.3 Storage Advances to Watch


The maturity of storage networking solutions, from the basic hardware infrastructure to the overall management, sets the stage for incremental advancements. Striving to extract additional performance and functionality at reduced costs, future storage networking products have to dig deeper or smarter into the functionality layers . In this section, we explore a few of those options.

12.3.1 Application-Specific Storage

Application-specific storage is one method to enhance the performance and functionality characteristics. By designing storage systems that tie directly into defined application requirements, users get the benefit of a custom-fit solution. The most common examples of application-specific storage include those focused on Microsoft Exchange and Oracle databases.

In the case of these particular applications, unique requirements directly reliant upon the storage system make "application awareness" a useful feature. For example, a storage system aware of the allocated capacity assigned to a Microsoft Exchange server could automatically provision more storage once that capacity was nearing full utilization. This helps administrators avoid the trouble of manually adding capacity and, more importantly, frees them from the requirement to continually monitor the system.

Similarly, table space for an Oracle database ”or any other database, for that matter ”could be automatically increased based on communication between the database application and the storage subsystem. In doing so, the increment must be large enough to prevent fragmentation and allow adequate performance.

The interaction between applications and storage subsystems can take place via a variety of mechanisms, such as Simple Network Management Protocol (SNMP) traps or the increasingly more sophisticated XML-based exchanges, such as those using the Common Interface Model (CIM) specification developed by the Storage Networking Industry Association.

Storage professionals looking to improve overall system functionality will benefit from assessing application-specific requirements and working to match those with more intelligent storage systems.

12.3.2 Automation

As discussed throughout the book, the management component is usually the largest ongoing expense of storage systems. The plethora of hardware and software products tailored to solve specific problems can lead to an inordinate time commitment for simple administrative tasks . For example, setting up and configuring a remote replication system for offsite storage can take anywhere from weeks to months. Creating a duplicate database for testing and experimentation can take hours or days. These functions and hundreds of others often require so much manual involvement that they are viewed as too troublesome to implement.

Storage management systems continue to offload more administrative tasks from storage professionals through a variety of intelligence mechanisms. At the most basic level, management systems use pattern identification to trigger alerts or traps that can be sent to administrators. This frees IT professionals from day-to-day oversight functions, but still requires them to act on the information. An example of pattern recognition might be a database table reaching allocated capacity on a volume.

The next step in the intelligence spectrum is independent decision making. In this stage not only can the system detect patterns, but it can also act on a predefined set of rules. Taking our previous example one step further, when a database table approaches maximum capacity of a volume, the management software could expand that volume by assigning more storage from a free-space pool. In this case, a pattern was identified and a solution applied without manual intervention.

The previous examples represent relatively closed sets of functionality with defined parameters. In the case of the decision process, an administrator still had to define a free-space pool in order for the system to automatically allocate storage. Only part of the process has been offloaded.

Peering further in to the future, the only practical way for administrators to cope with the data set growth and reliability requirement increases is through centralized automation. Through these systems, agents will act on an array of input criteria to determine an action set. Let's take the previous example one step further. The database described happens to represent mission-critical e-commerce data. In expanding the volume, the system must ensure that additional storage will have the same performance characteristics as the original storage. It also needs to confirm that access to the expanded volume can support mission-critical applications. The agents will do this by exploring the storage network access paths between the database server and additional storage. Further, since this database requires synchronous replication, the system will need to confirm that the remote site has the expansion capability as well and that the data set increase and resulting replication traffic can be sustained on the allocated bandwidth of the remote link. Finally, all of this must take place without any data loss or system interruption. A graphic outline of intelligence systems is shown in Figure 12-4.

Figure 12-4. Types of intelligence systems.

graphics/12fig04.jpg

As evidenced by the numerous system requirements, automated storage systems will require complete knowledge of the entire infrastructure and the ability to act upon that knowledge. This implementation style is often classified as true utility computing where proper storage can be delivered anytime and anywhere. Most large system and storage vendors have automated management programs underway, and this area will continue to be a driving sector for storage area networking improvements in the future.

12.3.3 New Platforms

The increasing digitization of corporate assets now drives enterprise storage growth across a number of categories. While previously storage growth may have been limited to databases residing on mainframe computers ”a fairly localized and contained infrastructure ”distributed systems, desktop PCs, and the use of Internet-oriented applications precipitate a wider spectrum of storage categories. Key storage category metrics include the following:

  • Where is the data stored?

  • How frequently is the data accessed?

  • How frequently is the data changed?

  • Who needs to access the data?

  • What are the longevity requirements?

  • What is the underlying data structure?

  • What are the security requirements for the data and infrastructure?

Since answers to these questions vary, each category of data emerges with its own requirements. Block-accessed SANs and file-accessed NAS have served multiple storage categories respectively. More recently, the emergence of object-oriented storage systems, such as content-addressed storage, has helped satisfy requirements for the fixed-content storage category largely represented by digitized corporate assets that are created once, do not change, must be accessed by numerous users, and have extensive longevity and data protection requirements.

But even within these primary categories of block, file, and object-oriented approaches, distinct variations exists. Consider the difference between managing millions of short, text-based emails and managing hundreds of thousands of large, image-based medical x-rays, cat scans , or MRIs.

As subcategories expand, we are likely to see a drill-down to more specific storage platform subsets or entirely new approaches. Storage administrators can benefit from storage platform customization by tracking storage categories, growth, and usage patterns.

12.3.4 Declassifying Storage Networking

During the first few decades of computing, storage was an integrated part of the overall solution. The shift from DAS to NAS mandated a separate technology classification and launched an entire industry specifically focused on the storage component. Soon enough, many organizations realized that the storage piece often outweighed the server piece in terms of dollars spent on acquisition and maintenance. More vendors entered the market, more products were introduced, and the multibillion-dollar storage industry took on a life of its own.

Jump to today's environment ”a collection of disparate storage resources that serve different purposes within the organization, require an inordinate amount of manual administration, and cannot be easily integrated with one another. To some degree, the classification of storage as a separate system component has caused considerable headaches .

IP storage networking is not the panacea to solve this problem. Integration of storage into mainstream computer communications systems is. It just happens that the means to accomplish that task is through the use of a communications protocol more widely used than any other the technology industry knows ”the Internet Protocol.

Sometimes the storage industry (and many other industries, for that matter) get swept up in the debates over lower-level technologies and user -level insights. The very introduction of IP storage networking as a term initially incited a riotous uproar from the Fibre Channel community, and an ensuing debate continued to ricochet for years in the media. The proponents of IP storage networking, your author included, would have been better served by calling the technical advancements integrated storage networking.

Providing a means for storage components to interact seamlessly with other computing components is the first step towards declassifying storage networking. This does not imply declassification from the security sense of the word, but rather the elimination of a separate category and a more complete integration with the overall computing system.

Open, interoperable communications serves as the first step. IP storage, in its many permutations , will pilot distinct and diverse storage resources towards a cohesive communications model. This is not to say that other communications methods will disappear. Fibre Channel and yet-to-come technologies will prosper and serve respective purposes. But the glue from a global to local perspective will be the ubiquitous IP network.

Abstraction layers between storage consumers and storage repositories can then facilitate a more accurate matchmaking scheme. Underlying virtualization layers will serve storage management applications in providing the right storage to the right user at the right time. These matchmaking skills are still in their infancy, yet promise to radically improve gaps that exist today.

Finally, the incremental addition of intelligent services and the use of automated systems will hone matchmaking and maintenance skills to a level of specificity likely unattainable through manual administration. Whether through pattern recognition, decision process, or self-directed intelligent agents, incorporation of artificial intelligence to storage management will drive complete system and storage integration, and ultimately the declassification of storage networking.



IP Storage Networking Straight to the Core
IP Storage Networking: Straight to the Core
ISBN: 0321159608
EAN: 2147483647
Year: 2003
Pages: 108

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net