Storage virtualization
Storage virtualization is a technology from computer science to apparently change the physical properties of existing storage space towards users or to abstract them from them. The technology is used so that users do not necessarily have to manage the available storage space along the physical boundaries, for example per hard drive or per storage unit. Storage virtualization means that the storage space appears virtual to users : The storage space can be divided into storage systems or hard drives, but these media do not have to be physically available. A software ensures that the virtual memory allocation fits a suitable manner to the existing physical space.
Users benefit from storage virtualization by not being tied to physical boundaries. Restructuring or expanding the physical storage facility affects users of the storage facility less or not at all if the storage facility is available in a virtualized manner. For system administrators , on the other hand, the advantage is that the available physical storage space can be distributed more effectively to the existing users. The degree of utilization improves.
There are different technological approaches to performing storage virtualization.
Virtualization method: Out of Band or Switch Based
Considering out-of-band storage virtualization is now only necessary academically. The products that followed these approaches have been pushed out of the market. The main reason for this is the very complex setup that was necessary for this type of storage virtualization.
This approach promises that the actual virtualization systems are located outside the data path. The actual virtualization is carried out in switch systems. This is where storage areas are logically linked to form virtual drives.
Advantages:
- No obstruction in the data path. Since the virtualization is carried out very close to the switch, we speak of IO delays of around 30 microseconds per IO. This is comparatively little if you assume that disk access takes place in the millisecond range.
Disadvantage:
- No way to establish caching in the data path. The performance of virtualization therefore depends on the individual performance and thus also on the costs of the underlying storage systems.
Examples: EMC Invista, LSI SVM and HP SVSP
Virtualization method: virtualization via a storage system
The virtualization is carried out directly in a storage system that now has the option of integrating external storage into its own processing.
Advantages:
- A storage system that can be individually expanded with external storage. All functions can now be managed from a central point.
- The concept is very suitable for the consolidation of different storage systems.
Disadvantage:
- Due to the high consolidation value, there is a risk that it is much cheaper to expand the storage system yourself than to connect external storage to the system. This in turn contradicts the virtualization approach, especially when storage virtualization is viewed as a future concept.
Example: HDS USP-V, HP XP P9000, IBM Storwize V7000 and OEM Derivate SUN
Virtualization method: In band
Special virtualization systems are established in the data path in the SAN. Each IO is routed through these systems and mapped accordingly to the disk blocks.
Advantages:
- The virtualization is completely transparent for the server.
- The virtualization system becomes a storage instance: the servers do not need any special drivers for the storage hardware, as they can no longer see them.
- Integrated caches can improve the performance of the system.
- As a result, cheaper storage systems can provide significantly higher performance.
Disadvantage:
- Whenever the virtualization system has weak cache properties, the performance of the memory deteriorates.
- Increased latency due to the use of a layer is generally compensated for by cache.
Examples: FalconStor IPStor, IBM San Volume Controller SVC, DataCore SANsymphony-V storage hypervisor
Virtualization method: host-based
A driver on the host system takes over all virtualization functions.
Advantages:
- No additional hardware is required.
- Best performance of all storage virtualization methods, since all storage devices are always direct, i. H. can be addressed without intervening instances (see in volume) and thus without additional latencies. Caching processes on the host enable additional acceleration and are inherently faster than caches in the SAN.
- Freely scalable, the limiting factor is primarily the number of possible I / O controllers.
- Easy access to the storage virtualization functions from the host's operating system or even from applications.
- This makes it particularly suitable for combination with virtual machines .
- Easy interchangeability of storage systems, host-controlled live migration of data between storage systems.
Disadvantage:
- The host-based storage virtualization must be ported for each desired platform and is therefore usually only available for selected operating systems.
- In the SAN u. Several hosts may have simultaneous access to the storage resources. Only a cluster-capable host-based storage virtualization can solve the problems that arise, but then offers more convenience, e.g. B. for high availability solutions, live migrations and resource sharing.
Examples: Veritas / Symantec Storage Foundation, OSL Storage Cluster, Linux Logical Volume Manager