IP Storage Networking: Straight to the Core

Effective storage strategies ride on the recognition that networked storage goes well beyond a set of required functions. Transitioning from the mindset of "We need this to protect our business" to "We have an opportunity to exploit a competitive advantage within our industry" means that technology deployment and business strategy meet to provide tangible , cost-saving benefits to the bottom line. Adopting an offensive mindset is the first step to realizing these gains.

Recent developments in the rapid progression of storage networking have made competitive advantage openings wider than ever before. Specifically, the merging of traditional network architectures, such as those based primarily on Ethernet and IP, with traditional storage architectures, such as those based on Fibre Channel and SCSI, has created enormous opportunities between requirements, capabilities, and solutions. The ability to connect traditional storage devices with common IP networks creates an entirely new distribution mechanism for storage. The power of that distribution mechanism cannot be underestimated, especially since IP networking represents the single greatest electronic distribution network in the world today. Granting ever-increasing data storage capacity access to that new distribution system fosters entirely new methods of handling data storage and management.

Companies that recognize IP networking as a new distribution mechanism for storage can rapidly transform their infrastructure and business processes to take advantage of it. The equipment exists today to seamlessly migrate from conventional storage architectures to those that embrace IP as a distribution mechanism. The resulting confluence of storage and networking, and the consolidation between the two, serves as the starting point for offensive storage strategies.

5.3.1 Storage Agility through Scalable, Flexible SANs

5.3.1.1 Meeting Unforeseen Demand

SANs provide operational agility for storage through scalable, flexible architectures. Once servers and storage are networked, that network can grow and change with varying business and technical needs. One example of SAN scalability is the rapid expansion of new capacity. In the previous section on budget protection, we discussed defensive approaches to containing storage capacity costs through SANs. In the offensive view, rapid expansion, such as adding near-instant capacity growth to a successful e-commerce application, can be accomplished through a storage network. In a direct-attached approach, where each server has its own captive storage, adding new storage might require adding new server HBAs and new storage devices. Both of these time-consuming processes might require that the server be offline for the reconfiguration period. With a SAN, available capacity from underutilized storage devices could be reassigned as needed. Alternatively, a new storage device could be added to the SAN nondisruptively and quickly assigned to the appropriate server.

The ability to deal with unforeseen business changes mandated by underlying technology shifts makes storage competence and agility an integral part of any IT organization. Not only will networked storage solutions allow for added capacity, as described in the previous example, but they also allow for rapid deployment of applications on new server platforms. Either side of the network (servers or storage) can be tailored to fit the appropriate business needs. This growth and realignment capability ensures that unpredictable business requirements are met.

5.3.1.2 Providing Equipment Choice

SANs offer storage flexibility by being able to assign any storage to any server. Many administrators use this flexibility for storage consolidation applications, such as connecting a variety of disparate servers to uniform, centralized storage or creating pools of heterogeneous storage that can be assigned to individual applications. In either case, the underlying network architecture allows administrators to substitute one storage device for another, providing freedom to choose among equipment providers.

This is not to say that all storage devices are created equal. The internal design of a storage array or tape library can vary dramatically from one vendor to the next . For disk arrays, this might include the type of drives used, the size of the disk cache, or the capacity of an internal switching backplane. These characteristics directly impact the storage speed and reliability, resulting in some devices that can't easily be swapped for another. Yet the storage fabric linking servers to devices provides flexibility to substitute storage devices among each other, especially if they are within the same category of performance and feature characteristics.

This device flexibility provides an offensive approach to vendor management, allowing heterogeneous storage within a single configuration and providing more customer control. If one type of disk array becomes too expensive or can't provide the feature set required, another can be added without having to reconstruct the entire configuration. So, in addition to solving short-term storage consolidation needs that may be defensive in nature, customers build in an offensive approach to long- term vendor management with the installation of a storage network.

5.3.2 Network Centralization

Network centralization is the epicenter of offensive approaches to storage deployment. The ability to utilize the world's largest electronic distribution data network for storage implies that enormous discontinuities in current practices can be realigned accordingly . Specifically, companies that have DAS or Fibre Channel SANs can begin to focus on methods to handle data storage that leverage the global IP network.

In the early part of this decade , the storage networking industry debated the merits of different interconnect technologies, primarily IP and Ethernet as compared to Fibre Channel. While Fibre Channel has proven an effective edge interconnect for storage devices and SANs, it simply cannot compete with IP and Ethernet on a global scale. This discussion was covered in more detail in Chapter 2, "The Storage Architectural Landscape." More important than the technology is the size and scope of the industries at hand. IP and Ethernet have dominated corporate networking for years and served as an effective mechanism for LAN clients to talk to servers and NAS clients to talk to NAS servers. These applications have been well served with IP and Ethernet, and the networking component has provided more than adequate performance, reliability, security, and flexibility.

With today's technology, customers have the opportunity to add storage networking to the larger network landscape of IP and Ethernet. From a storage perspective, this is likely to begin by linking Fibre Channel SANs across IP networks. Whether they are local, campus, metropolitan, or wide area networks, Fibre Channel SANs can be connected across IP networks to provide greater distance and scalability than pure Fibre Channel scenarios. Since almost every organization has an IP network in place, the extension of Fibre Channel SANs across this existing infrastructure makes business sense compared to deploying additional, separate Fibre Channel storage networking infrastructure for SAN connections.

New IP Storage switches allow storage networks to be built with IP cores, yet still retain compatibility with Fibre Channel devices. These configurations also have a natural extension to IP networks, further enhancing the value of that common technology infrastructure across a variety of networking and storage applications. Finally, the deployment of iSCSI servers and storage devices with native IP connections allows direct access to the IP network and the ability to exploit that rich and expansive resource.

As shown in Figure 5-87, IP networks provide flexibility and operational agility across all types of data networking and storage networking applications. The benefits of this uniformity can be categorized as follows :

Figure 5-8. IP networking extends across traditional data and new IP storage applications.

Familiar and Ubiquitous IP Networking Technology . IP networks provide low, intangible costs in terms of management, training, and deployment. With years of industry use and the largest installed base in the world, IP networks are more familiar to and easily managed by people than are any other type of network. As such, the tools and training for IP networks are unparalleled in simplicity of use and sophistication, leading to more rapid and cost-effective deployment. The centralization of network technology greatly reduces management and administration cost

Enhanced Functionality . IP networks provide unrestricted topologies that guarantee interoperability across cross-vendor platforms. The industry has delivered advanced routing, management, and security features that power the most demanding of network environments.

Scalability . IP networks provide the utmost in scalability in terms of network size, speed, and distance. Network size is defined as the number of network nodes. Of course, no example comes close to the size of the Internet, the world's largest IP network. Ten Gigabit Ethernet provides the highest speeds available for common networking platforms and is delivered today both as interswitch links and switch backplanes . IP networks reach distances that span the globe, making any geographically point-to-point communication possible.

By adopting the network centralization approach, IT professionals can design architectures that leverage existing staff and technology to dramatically reduce overall cost. This approach embraces the open distribution channels of IP and places storage on the most pervasive networking platform available.

5.3.3 Platform Consolidation

The introduction of IP networking to storage enables consolidation among previously separate architectures of NAS and SANs. Historically, these two storage platforms have operated on completely different networks, with NAS using Ethernet and IP, and SANs primarily using Fibre Channel. As covered in earlier chapters, NAS operates using file-level commands, while SANs use block-level commands. Since both NAS and SAN architectures serve distinct and useful purposes, most companies support both platforms and, in turn , two networking infrastructures .

Stepping back, one can easily see how the build-out, management, and ongoing maintenance of two networks could result in excess equipment purchases and additional overhead for staff, training, maintenance, and support. Without the introduction of block-based IP storage, there was no other choice. To better understand current solutions and the path to platform consolidation, we'll walk through the example beginning with a typical configuration.

In Figure 5-9, the dual fabric of an IP/Ethernet LAN for NAS and a Fibre Channel SAN requires dual management systems, additional hardware, and potentially duplicate staff and training. These two network systems essentially perform the same tasks . Both run point-to-point, full-duplex , switched gigabit traffic, yet they cannot be shared. Additionally, a pure Fibre Channel fabric does not provide a transition path to iSCSI or other IP end systems.

Figure 5-9. Conventional Fibre Channel SAN and Ethernet/IP LAN.

The platform consolidation migration begins by adopting an IP storage fabric, as shown in Figure 5-10. Using IP storage switches or routers, the SAN core transforms to IP and Ethernet while retaining full compatibility with installed Fibre Channel devices, such as Fibre Channel HBAs or storage devices, or with Fibre Channel SANs. This allows existing applications to run without modification, yet prepares for the introduction of iSCSI end systems. These end systems, such as iSCSI HBAs and iSCSI storage devices, now integrate seamlessly into the IP storage fabric, driving more of the end-to-end storage connections to IP.

Figure 5-10. Introduction of IP storage fabric and iSCSI.

At this stage in the migration, the customer benefits are numerous . First, a common networking technology platform exists between the LAN and SAN. Even though these two networks might not be fully integrated into a single network, they can be managed by common IP-trained staff members and share common IP and Ethernet management software, dramatically reducing overhead costs. Additionally, the introduction of the IP storage fabric provides access to extended metropolitan and wide area networks for a complete range of remote or distributed storage applications.

Carrying the platform consolidation migration through to completion, the Ethernet and IP networks can be combined to form a single, integrated NAS and SAN IP storage fabric. Integrated storage NICs (IS-NICs) running both block-level storage protocols like iSCSI and traditional TCP/IP protocols such as NFS or CIFS make this consolidation possible. Now, using a single fabric, servers have access to SAN and NAS subsystems, either Fibre Channel-based or IP-based. IP storage switches or routers within the fabric assist with iSCSI-to-Fibre Channel conversion when required.

Figure 5-11 demonstrates the complete platform consolidation configuration. Note that the NAS subsystems are now connected directly to the IP storage fabric and can be accessed by servers with IS-NICs or even end- user clients. Similar to the server-based IS-NICs that can access both block-based and file-based systems, the introduction of a consolidated IP storage fabric leaves room for similar storage devices. Hybrid subsystems that can handle both block and file protocols could easily attach to an IP storage fabric and provide a complete range of NAS and SAN services to applications. This greatly alleviates decision turning points for storage administrators considering the benefits of NAS and SAN solutions, eliminating the uncertainty and providing complete flexibility to tune the configuration to the appropriate platforms.

Figure 5-11. Integrated SAN and NAS.

Server redundancy costs can be cut in half with IS-NICs. Since high-end servers require redundant storage and network adapters, a conventional configuration has two Gigabit Ethernet NICs and two Fibre Channel HBAs. With IS-NICs, full redundancy for LAN messaging and storage traffic can be provided with only two cards. One IS-NIC can run in LAN mode and the other in storage mode. In the event of a failure, a single NIC could cover both functions. Though performance is impacted temporarily during the failure, those costs may be significantly less than providing twice the hardware infrastructure for redundancy.

Many customers wonder about the benefits of NAS compared to SAN and where to invest precious budget allocations . The IP storage fabric coupled with IS-NICs and hybrid subsystems allays those fears through a single investment in IP networks. Let's take an extreme example and say that SANs disappear within 10 years. With an investment in IP storage networking, IP storage fabrics , IS-NICs, and hybrid subsystems, all of that storage equipment could be redeployed to serve NAS-oriented applications. The reverse situation would also hold. In either case, the investment in IP networking infrastructure serves as a flexible, redeployable resource both within an integrated IP storage fabric and within the entire global corporate network. The power of that asset redeployment cannot be underestimated. Companies stand to benefit from extended use of the network assets and the flexibility to redeploy staff as needed to build, operate , and maintain IP networks, whether those networks are used for storage or traditional data applications.

5.3.4 Common IP Network or Common Technology?

Questions are often raised about IP storage networking and sharing storage traffic and data traffic within a common IP network backbone. Some people state that this implementation could lead to congestion bottlenecks. While combining both messaging and storage traffic on a single network is possible, initially a more pragmatic implementation is to segment the IP network infrastructure and move storage and data traffic via different paths. This approach enables customers to protect their investment in IP networking and maximize the efficiencies of moving both types of traffic over a common infrastructure. For example, familiar, standard technologies such as Virtual LANs (VLANs) permit the use of the same network equipment, yet partition the Ethernet network into separate entities.

Common network technology indicates that while deployment may be many physical networks from SAN to LAN to MAN to WAN, the underlying infrastructure is the same, providing benefits such as the use of existing IP networking staff, uniform network management, and common sourcing and maintenance of networking equipment.

Figure 5-12 shows an IP storage fabric in conjunction with a Fibre Channel fabric and an IP network backbone. In this example, the storage fabric remains independent from the network backbone that provides server farm, end user, and voice access. This segmentation allows for simplified traffic administration while still retaining the benefits of more IP technology throughout the enterprise.

Figure 5-12. Segmenting the IP storage fabric from the IP network backbone.

5.3.5 Multilayered Storage Fabrics

Chapter 2 introduced the concept of a multilayered storage fabric centered on an IP core. Now, incorporating new iSCSI devices and multifunction devices such as NAS/SAN hybrid targets, we expand the model, as shown in Figure 5-13. This highlights the tremendous flexibility of centralizing all storage devices, whether Fibre Channel-, iSCSI-, or NAS-based, on a central IP storage fabric. Storage administrators have a flexible, shared resource that can be partitioned and segmented for each storage platform, yet can also be reallocated as needed. The degree of performance and fine-tuning within the IP core far exceeds the requirements of the average large enterprise customer, leaving ample network capabilities for each platform.

Figure 5-13. Framework for multilayered storage fabrics.

The flexibility of the IP storage fabric shared resource is shown in Figure 5-14. The primary IP storage transports include Fibre Channel devices or SANs connected across an IP fabric, iSCSI devices connected directly to an IP fabric, and multifunction devices, including NAS directly connected to an IP fabric. All three platforms take advantage of robust IP and Ethernet networking features. If one platform emerges as a more economical means to conduct storage operations, administrators can easily reallocate the core network resource as needed to accommodate changes in the underlying storage choices. This flexibility protects significant network spending and ensures platform choice for storage professionals.

Figure 5-14. Allocating the IP storage fabric among storage platforms.

Категории