Page 2 of 3   « previous page   next page »

iSCSI and Virtualization
iSCSI systems like Dell’s EqualLogic storage arrays broaden the potential for full-fledged virtualization deployments in small and medium businesses by helping remove the need for a costly and complex Fibre Channel-based SAN deployment. The acquisition, implementation, and operational costs of an iSCSI-based SAN are intrinsically lower than those of a Fibre Channel-based SAN, helping remove an economic barrier to the adoption of server virtualization technologies. Advanced iSCSI-based system architectures can help change enterprise expectations for large deployments as well. A truly flexible iSCSI-based virtualized environment can help not only reduce costs, but also simplify deployment, provide comprehensive storage management and data protection functionality, and enable seamless VM mobility between hosts.

The addition of iSCSI protocol interfaces to classic storage system designs does not fully exploit the potential of the iSCSI protocol and, in turn, does not adequately solve the storage management challenges faced today by IT managers. Even with the introduction of iSCSI, storage administrators using classic storage systems typically require a rarified level of knowledge for configuring and tuning storage arrays, RAID geometries and data layout considerations, application workload analysis, forced data migration, and complex system upgrades.

Historically, the adoption of storage virtualization technologies in SAN environments has been limited to enterprise data centers needing a tool for online data migration between heterogeneous storage environments. This add-on design allows for heterogeneous storage and data movement, but does not simplify managing the configuration. Most of these designs have been implemented as external appliances sitting in the SAN. These devices generally add complexity to an already complex environment – introducing additional points of management in the SAN, masking value-added features of the attached arrays, and limiting the performance and scalability of the consolidated SAN storage pool to a single gateway hosting the back-end storage.

Advanced storage virtualization technologies offer the ability to virtualize storage at the storage device level. This approach helps achieve scalable performance and capacity and reduce overall management by aggregating multiple controllers into a cooperating set of resources – that is, by virtualizing volumes not only across disks in a controller but also across storage systems in a SAN.

iSCSI storage arrays are an excellent example of such a “scale-out” architectural design. Despite their powerful simplicity, such designs are virtually non-existent in the Fibre Channel-based SAN world, primarily because of architectural constraints inherent in Fibre Channel network deployments.

A flexible storage architecture is particularly appealing in a virtualized server environment because the scale-out architectures of both server and storage infrastructures complement each other, providing an end-to-end virtualized infrastructure (see Figure 1). By deploying both virtualized servers and virtualized storage together, IT managers achieve an infrastructure that’s particularly adaptable to changing business requirements by providing non-disruptive workload migration and balancing across all physical resources (both storage and server resources) as well as online resource expansion that can immediately apply to pre-existing server and storage workloads without intervention and tuning.

Virtualized iSCSI-based SANs
A virtualized iSCSI SAN complements a server farm (see Figure 2). The SAN needs to be designed to be fully redundant, containing disks, multiple high-performance network interfaces, redundant controllers with mirrored battery-backed caches, and other advanced features. The disks can be automatically protected with RAID (RAID-5, RAID-10, or RAID-50) and hot spares. Multiple models of arrays exist, leveraging high-performance 10,000 RPM and 15,000 RPM serial attached SCSI (SAS) drives or cost-effective, high-density Serial ATA (SATA) II drives.

A storage group comprises a single array or multiple arrays. A group is a virtualized resource, appearing to server systems as a single entity that offers network storage access to a single large pool of storage – a SAN composed of a single virtually scalable, high-performance storage system. Unlike a conventional SAN, in which multiple controllers appear as independently managed islands of storage, each group member can cooperate with other members to automate resource provisioning and performance optimization.

Administrators can create data volumes from within the storage pool. The group exports its volumes as iSCSI targets protected with security measures that include authentication and authorization. Upon connection, server systems work with volumes as VMFS data stores, which contain multiple VMs and virtual disks. A variation on this is a raw device map in which the SAN-resident volume is mapped directly to a corresponding virtual disk in the server environment, circumventing the VMFS data store. A third alternative is to employ the native iSCSI software initiator of the individual guest OS to connect directly to the SAN-resident volume. This last option allows administrators to use all array-based functionality and allows seamless integration with high-level applications, particularly backup applications that use Microsoft Volume Shadow Copy Service (VSS) or Virtual Disk Service (VDS) as an underlying protocol.

Volumes are distributed among the arrays, with data placement and access continually adjusted for optimal performance as resources are added or workloads change. When an array is added as a group member, its disk space is added to the group’s storage pool. Volumes can be automatically re-striped and distributed across the members of the storage pool. Controller resources are also dynamically adjusted based on the workloads being generated by the server farm.

Data and network I/O to the group can be automatically load balanced across the group members’ resources. As capacity and performance requirements increase, a group can be scaled linearly in both capacity and performance – all while online. New members “learn” configuration and performance information from the group – with no manual intervention. Data and client connection load balancing can occur automatically as the group scales. I/O activity is monitored, and data and network connections are adjusted as needed.

Page 2 of 3   « previous page   next page »