Employing IP SANs to address business needs. (Storage Networking).

In today's economic climate, enterprise IT organizations are challenged to minimize capital acquisition expenses (CAPEX) and operational expenses (OPEX) while meeting overall business objectives. Data storage needs continue to grow despite suppressed IT spending. This forces enterprise IT organizations to look for ways to maximize the productivity of their limited human and financial resources.

As part of their efforts to increase IT staff productivity and reduce software license costs, many IT organizations are consolidating their application servers and storage. They are migrating from direct attached storage (DAS) to networked storage to increase storage utilization and fulfill disaster recovery requirements.

However, the high cost and complexity of most Fibre Channel storage area networks (SANs) and components have hampered the adoption of SANs for the fastest growing segment of data applications, commonly referred to as functional or reference data. This type of data includes email, document and medical images, engineering designs, scientific data, multimedia content, and the like. While this type of data is mission important and needs to remain online and readily available for users, it typically does not justify the cost of implementing a Fibre Channel SAN solution.

The Internet and Engineering Task Force (IETF) recently ratified the draft iSCSI standard, which promises to provide a lower-cost option for users requiring high levels of performance and reliability for their SAN implementations. This has paved the way for a new class of smarter, more intuitive, yet much more affordable networked storage solutions that are a result of a fundamental shift in the way network storage is designed--from making storage an extension of the network to designing networked storage to be a native part of the network infrastructure. By leveraging established and broadly deployed industry standards, such as Ethernet, IP, and the vast selection of IP-based services, these new intelligent IP SANs will enable IT organizations to deliver cost-effective, easily managed, flexible networked storage as part of their strategic IT infrastructure.

The Road to IP SAN

The initial application of iSCSI for storage has been solutions that act as a bridge between IP and Fibre Channel. The resulting "simple" P SANs, however, addressed only a small portion of today's needs by merely tying together Fibre Channel SAN islands. They fall well short of leveraging the maturity and power of existing IP networks and the lower connectivity costs associated with Ethernet to not only address the "stranded" application servers, but also enable distributed, networked storage services (see Figure 1).

Drawing a parallel to iSCSI development is the deployment of low-cost ATA-based storage arrays as a solution for nearline storage, enabling IT administrators to regain control over the ever-increasing backup/restore window. As a staging area for tape backup and archive, ATA-based nearline storage is a natural networked resource for disk-to-disk backup of all of the application servers' Direct Attached Storage. Similar to Fibre Channel SANs first being adopted for LAN-free backup, iSCSI-enabled ATA disk arrays and tape arrays will form one of the early native applications of IP SANs.

Another emerging native application for IP SANs is server consolidation. Many IT organizations are looking to consolidate their numerous legacy PC-class servers within their data centers and among their enterprise workgroups, which will enable them to reduce management and application software license costs (i.e., OPEX). The most notable applications are Web, email, and CIFS/NFS file services (i.e., the NAS "heads").

With the introduction of iSCSI storage arrays, IP SANs provide cost-effective, shared networked storage to application servers. Where server blades are being deployed for server consolidation, an P SAN offers an ideal complementary solution since, by design, the server blade has very limited local storage. Such P SANs can also incorporate nearline storage functionality and provide iSCSI tape backup to form a complete SAN that delivers the functional benefits of an equivalent Fibre Channel SAN (see Figure 2).

Since iSCSI is defined on top of TCP/IP, an application server host can become iSCSI-enabled simply by installing iSCSI initiator software. The iSCSI initiator software operates over the native TCP/IP stack and utilizes the Ethernet network interfaces that are already installed. The application host does not need to be shut down to install an iSCSI host bus adaptor (HBA).

Furthermore, since TCP/IP/Ethernet already provides the basic transport interoperability, iSCSI interoperability is realized much faster than with Fibre Channel. As native TCP/IP support within various operating systems (particularly Microsoft Windows) paved the way for the explosive growth of the Internet, the availability of native support for iSCSI initiator software (also within various Microsoft Windows operating systems) will dramatically accelerate the deployment of IP SANs.

For enhanced P SAN performance, IT organizations have the option of installing TCP offload engines (TOE) or iSCSI HBAs in selective application server hosts to offload protocol processing. Similarly, the increasing availability of jumbo packet support in commodity Ethernet switches also serves to mitigate the performance gap between IP and Fibre Channel SANs.

For security, ISCSI employs Challenge Handshake Authentication Protocol (CHAP) for basic authentication to protect the storage from unauthorized access. This can be used in conjunction with the widely deployed RADIUS authentication servers for consistent, centrally managed access control to shared storage resources. For highly sensitive or secure data, iSCSI has the option of utilizing IPsec not only for stronger authentication and access control protection, but also for data encryption.

For small deployments, iSCSI includes rudimentary discovery capability between the iSCSI initiator and targets. The iSCSI standard also specifies the use of existing Service Location Protocol (SLP), the enhanced dynamic host configuration protocol (DHCP) and the new Internet Storage Name Service (iSNS) for service discovery and name resolution to facilitate deployment and enable the management of large IP SANs.

With lower equipment acquisition costs and much quicker planning and deployment than with Fibre Channel SANs, IP SANs also offer lower overall CAPEX. When the additional cost of maintaining a separate Fibre Channel SAN is obviated, further OPEX savings are realized.

Bringing the Promised Benefits of SANs to the Masses

A significant part of the high costs associated with Fibre Channel SANs result from the complexity required to design, deploy, manage, reconfigure and expand them. The high Total Cost of Ownership applies to both the connectivity (i.e., the FC fabric) and the storage subsystems. As such, the benefits that SANs promised (e.g., simplified shared storage provisioning and management) have remained beyond reach for the vast majority of IT organizations.

IP SANs, however, will deliver the promised benefits of SANs to those enterprises, and for those applications, that simply cannot afford a Fibre Channel solution. By exploiting the robust set of broadly deployed IP network services and the low cost, standardized Ethernet connectivity, intelligent IP SANs directly address the dominant components of storage costs; specifically management, connectivity and access.

Intelligent IP SANs start with the deconstruction of the conventional peripheral-bus and loop interconnects-based RAID subsystem structure. By replacing the conventional UO within the storage subsystem with a Gigabit Ethernet network, a fully switched IP SAN architecture--from the application hosts, through to the storage array controller and IP-addressable disks--is created.

Decoupling the storage array controller from the storage elements allows both to become part of the IP network. Keeping the storage array controller and the storage. elements separate enables both to be scaled and/or repurposed with the network, either independently or together, to meet application storage requirements. With a cluster of storage array controllers, a pool of IP-addressable disks, and a flexible network interconnect, the resulting IP SAN dynamically delivers fully virtualized storage arrays within the network infrastructure (see Figure 3).

In this 3-tier IP SAN, the P/Ethernet networks that connect the storage array controllers to the application hosts and to the IP disks are logically separate, but can be physically constructed as a single infrastructure that is partitioned into virtual LANs (VLANs). DHCP is employed to manage the IF addresses of the IP disks and the storage array controllers.

Any IF disk in the free storage pool can act as a spare for any of the virtual storage arrays in the IP SAN. The loosely coupled active-active storage-array controllers are specialized IF servers executing policy-based volume management and intelligent resource management. They form a cooperative cluster that provides both failover and load balancing. As a result, the fault tolerance, the high availability and the centralized management of the system--traditionally the domain of high-end Fibre Channel storage subsystems--are easily affordable for the enterprise's mission-important functional or reference data.

Figure 4 illustrates the application of an intelligent IF SAN that not only facilitates tactical server consolidation, but also sets the stage for deployment of strategic IT capabilities. Such intelligent IF SANs are imminently available.

Exploiting the Full Potential of IP SANs

The full potential of IP SANs will be realized when intelligent storage array controllers evolve into full-function IP storage servers that operate on networked storage OS, deliver distributed storage services, and are managed as an inherent part of the network infrastructure.

These full-function IP storage servers will be endowed with smart data-movers that extend long-distance mirroring and remote replication capabilities into integrated and automated caching and content distribution services. Bladed application servers will offer both remote-boot off the P SAN, and enable access to the application data stored in the P SAN. IP SANs will manage the resources intelligently and adapt to failures as part of the IF network. Disaster recovery will be a simple policy specification away.

In essence, all of the benefits originally envisioned for a storage area network, and much more, are much closer to reality in the form of IP SANs. Intelligent IP SANs are simply smarter solutions that ride the cost curve of standardized commodity networking (IP and Ethernet) and disk technologies (ATA and serial ATA), exploit mature Internet technologies and services, and leverage the wealth of knowledge that has developed for distributed processing, to deliver centrally managed, distributed storage and services. With full-function, intelligent IF SANs, these storage services will be easily deployed and quickly provisioned to meet the critical business objectives of the enterprise, without breaking the bank.

Peter Wang is CTO and founder of Intransa (San Jose, Calif.)

www.instransa.com