There are many ways to optimize the data center. But as with all things good, it is necessary to take each step in moderation. Consolidation and virtualization is one way to achieve higher utilization of IT assets. But shrinking the IT assets by more than 30% overnight will create a sudden over-capacity at the physical facility layer. The way to achieve a sustainable overall efficiency in the data center is through finding the right balance between optimization the physical facility and IT consolidation & virtualization.
Virtualization technology coupled with the right management tool is a powerful method to help reduce the total number of IT assets by consolidating multiple workloads into each IT asset, hence, maximizing its utilization, reducing the complexity and management of disparate s
erver hardware and OS platforms. Nonetheless, server virtualization is only one component in a truly virtualized enterprise infrastructure. The other critical component is storage virtualization. Both of these works in tandem – one addresses the compute side of the equation, while the other addresses the data side of it.
What does storage virtualization actually mean? Similar to server virtualization, storage virtualization is the abstraction of logical storage from physical storage devices. With storage virtualization, the growth and movement of data is no longer restricted within a physical storage array. Virtualization of storage provides the user with logical storage space that is location independent. Virtualization of the storage infrastructure is made possible as the virtualization engine which presents to the user, a logical storage space. The storage virtualization engine handles the process of mapping the local storage with the physical storage (transparent to the servers utilizing the logical storage).
When we design a virtualization architecture for a heterogeneous environment, it is likely that we will find DAS (Direct Attached Storage), JBODs, NAS (Network Attached Storage), Enterprise and Modular SAN arrays, and a convoluted SAN fabric configuration. Storage virtualization allows the grouping of these storage assets into a consolidated and centralized management and allow LUNs to be provisioned across heterogenous devices. So, the benefits are better re-usability of heterogeneous storage, optimized storage utilization, establishing an excellent foundation for charge-backs, as well as to better support server virtualization functionalities.
There are a couple of common ways to accomplish storage virtualization – array-based, or network-based. Array-based provides the benefit of an integrated solution from a dedicated storage vendor. Example of this is the Universal Storage Platform (USP) from HDS. The drawback of this approach is that you get locked in to a specific vendor and sometimes the storage vendors may have compatibility restrictions within their own storage array product line, making technology upgrades and data migration potentially expensive. However, the advantage is that array-based is likely to have less overhead than a network-based approach, and there is no need to acquire expensive licenses for replication function across heterogeneous devices. A network-based approach, on the other hand, uses intelligent appliances or switches (e.g. Invista) with varying degree of embedded functionality. It is usually prudent to validate with the vendor if the appliance provides data migration as part of the standard product.
So, why is storage virtualization important in a virtualized environment? Well, the full benefits of server virtualization can only be realized in combination with a virtualized storage layer that works in conjunction with and complements the virtual server layer. For example, dynamic virtual machine failover or load balancing feature such as VMware Distributed Resource Scheduler (DRS), helps to improve resource allocation across a resource pool in a cluster. When you enable DRS for a cluster, VirtualCenter will then monitor the distribution of CPU and memory resource usage for all the hosts and VMs in that cluster with their entitlement (based on resource policies) and may dynamically distribute the VMs across that cluster for load balancing when appropriate. This is a key benefit of leading server virtualization solutions and can facilitate not just load balancing but disaster recovery too. It is incomplete, however, without dynamic storage failover, which is a storage virtualization feature. Implementing storage virtualization, in other words, extends the benefits of an investment in server virtualization and builds upon them, providing a simplified storage management, improved storage utilization and a diminished need for proprietary vendor solutions.
It is possible to perform snapshots of VMs data at the server virtualization layer using the functionality of the service virtualization technology. The main advantage is the ability to quiescing the VMs for such snapshots, which assures that the data in the snapshot is crash consistent. However, this method has some drawbacks. For example, server virtualization snapshots do not support all the data accessible to a VM, and if you’re running VMware and your VMs are accessing raw disks or RDM physical mode drives, you cannot take a snapshot of their data via the VirtualCenter. The other disadvantage is that doing a snapshot at the server virtualization layer is very CPU and IO intensive. On the other hand, doing snapshots within a storage array is much more efficient, although the storage array snapshots may not guarantee a state consistent snapshot.
However, for a storage virtualization implementation to be effective and successful, it is necessary to have a matured policy-based deployment for abstraction of the physical storage layers coupled with matured IT service management processes and standards, and IT governance. Storage virtualization is typically implemented in conjunction with data classifications and grouping of the data storage environment into tiered storage. Tiered storage will help pave the way to support a policy-based deployment.
Tags: Consolidation, Cost, ITIL, Policy, Service Management, Storage, Virtualization