Storage Virtualization

The term “Storage Virtualization” is used to indicate the operation of dividing the available storage space into “Virtual Volumes” without regard to the physical layout or topology of the actual storage elements. Typically virtual volumes are presented to Operating Systems as an abstraction of physical disks and are used by these OS as if they were such disk drives.

Storage Area Networks impose the need of managing large amounts of storage in a uniform way and from a central location. The fast growth in storage capacity and processing power in many enterprise installations, coupled with the need for high availability and 24x7 operations, requires from the SAN architecture to enable seamless addition of storage and performance elements without downtime.

Nexsan iSeries is the perfect solution for server virtualization, databases, file storage, data consolidation and multiple data storage requirements when performance and scalability are a must and simple IP deployment is desired. With complete storage services, Nexsan iSeries gives users the ability to create and control thousands of virtual volumes supporting multiple applications and storage pools simultaneously.


By reliably incorporating SAS and SATAS in the same storage system, users no longer have to choose between a performance or capacity driven iSCSI SAN system. Separate storage pools can be created for the needs of different applications from databases to backup.

Virtualization at the Storage Sub-System Level
One method of virtualization is via storage management software that runs at the server level. The main advantage of this method is that it enables multiple storage subsystems to work in parallel with multiple servers. A key difficulty with this method is that it assumes a prior partitioning of the entire SAN resources (disks or LUNs) to the various servers. Virtualization is only performed on pre-assigned storage, losing a key advantage of SANs as well as the independence of volumes from servers. Typically, entire LUNs are assigned to specific servers, thus limiting the amount of servers that can use the same storage subsystem. Further, a virtual volume created over the storage space of two LUNs will not be easily moved to another server, especially if there are other volumes created over the same LUNs. Virtualization at the host level typically also require augmenting the management function with some parallel mechanism of zoning and LUN masking and relying on LAN connectivity for synchronization between servers, which may affect the reliability of the entire SAN.

Network Based Virtualization
A key requirement from Storage Virtualization is “making disparate storage look and behave like a common storage resource.” Network base virtualization assures vendor neutrality.

Network-based virtualization is being implemented in two major architectures:

• Symmetric approach – separate appliances in the data path of the storage network infrastructure
• Asymmetric approach – separate appliances installed out of the data path of the storage network infrastructure.

Symmetric (In the Data Path) Virtualization
The symmetric approach (sometimes referred to as In-the-Data-Path) calls for the appliance to be installed between the storage users and the storage resources. The key drawback of the symmetric approach to virtualization is that it creates a SAN bottleneck and thus limits the SAN performance and scalability and significantly complicates the design of large-scale, highly available configurations.

The symmetric virtualization concept requires that all data from all application servers pass through one single computer. To avoid serious performance difficulties, this computing platform needs to be capable of sustaining the throughput of the entire SAN, typically resulting in expensive hardware configuration. Even with high performance hardware, scalability remains an issue as the symmetric appliance has a fixed maximum bandwidth.

As an example, in a SAN with 50 application servers, the throughput required for each can reach 100 MB/sec. This value*50 servers = 5 Gbytes/sec. While it is fairly easy to install multiple RAID subsystems with aggregated performance exceeding this requirement, a typical symmetric appliance based on a standard Pentium III processor with a 64/66 PCI bus can deliver no more than 520 MB/sec (that is 260 MB/sec sustained rate as data needs to enter and exit through the PCI bus on the appliance) – a far cry from the specific requirement. Needless to say, as the SAN becomes larger, this problem becomes more acute.

The hardware cost to construct a symmetric appliance that will support 20 servers with reasonable performance could reach $25,000 and more; this cost is at least doubled if a High Availability configuration is required.

Some symmetric designs are attempting to circumvent the performance problem through adding a cache facility to the appliance. This approach further loads the appliance’s memory subsystem. For many reasons cache is much more effective when it is distributed (i.e. part of the RAID subsystem) rather than centralized (i.e. part of the storage management appliance).

The use of caching also complicates High Availability (HA) configurations, which try to avoid single points of failure, as well as Scalable configurations, where additional appliances are added to respond to increased load. If caching is used with multiple appliances, some kind of cache coherency strategy is required. Experience with RAID controllers shows that this can be very complicated (and expensive).

In order to achieve high performance on the server side, while using two (or more)HBAs, a software driver should be installed on each server. That feature is common to the symmetric and asymmetric solutions. Both need software to be installed on the server in order to create redundant and high performance (multiple paths) solutions for increasing throughput.

Asymmetric (Out-of-The-Data-Path) Virtualization
This method uses a combination of an Appliance and Agents to create and manage virtual volumes while enabling direct data transfer between server and storage subsystems. By having multiple storage subsystems working in parallel with multiple servers, total performance is increased up to the maximum of the FC fabric bandwidth or even more.
In the Asymmetric approach (sometimes referred to as Out-of-The-Data-Path), the handling of metadata is done separately from the data path. The appliance serves as a Metadata Center that “sees” the physical storage and allocates virtual volumes, while an Agent at each of the servers on the SAN is responsible for the actual virtual volume mapping. The Agent retrieves the volume configuration from the appliance and presents virtual volumes to the operating system as if they were disk drives. When the operating system send an I/O to the virtual volume, the Agent intercepts the I/O, translates the volume’s logical address to the physical address, and sends the I/Os directly to the storage devices.

This architecture maintains the flexibility of symmetric virtualization without incurring the degradation of performance or the high cost of hardware. The appliance can be a small and inexpensive unit, because it does not have the handle actual data transfers.

High Availability (HA) SAN configurations can be implemented through simple redundancy of the appliance. Very high degrees of scalability can be achieved through loosely connected storage domains and similar designs, benefiting from the key advantage of the asymmetric concept, i.e. that there are never pending I/Os stored in the appliance nor there is a need for clustering, cache coherency algorithms and other complex mechanisms which limit the practical expansion of real life enterprise SANs. HA server configurations are achieved (as in the symmetric approach) by installing multiple HBAs in the server and having a S/W Agent handle the Load Balancing and Fail Over functions.