Does the IBM zEnterprise or even the System z need a storage hypervisor? I would have thought no, but a recent briefing by IBM’s Ron Riffe has changed my thinking. Riffe also has posted a very informative three-part blog on the topic of storage virtualization here.
Mainframes running z/OS already are thoroughly virtualized and don’t need a storage hypervisor. Many mainframe shops today, however, run other operating environments beyond z/OS, such as z/VM and Linux, and for those environments a storage hypervisor can add value. As mainframe shops move into the hybrid zEnterprise environment a storage hypervisor can play an even bigger role.
In general storage hypervisors serve the same function as server hypervisors in the distributed computing environment. They facilitate the pooling of shared physical resources among virtual machines by playing traffic cop to resolve contention for resources. The hypervisor also provides management capabilities. The use of a hypervisor usually improves resource utilization, saving money and increasing flexibility in the process. It is the server hypervisor that allows distributed shops to scale out their servers in an attempt to match the high levels of utilization and management efficiency that are taken for granted with the mainframe.
The storage hypervisor performs a similar job, enabling the sharing of storage resources and streamlining the management of diverse storage resources. IBM has long had the key components of a storage hypervisor in the IBM SAN Volume Controller (SVC), a storage virtualization platform, and in the Tivoli Storage Productivity Center, which provides the storage virtualization management capabilities that enable non-disruptive data mobility and management across heterogeneous storage tiers. The combination is an IBM storage hypervisor that allows data access and mobility between two physical data centers as a stretched cluster, up to 300km apart for synchronous data movement. If used with VMware vMotion or PowerVM Live Partition Mobility, Riffe noted, you can achieve transparent migration of virtual machines and their corresponding applications and data, a necessity for disaster avoidance.
The real interest in the storage hypervisor for the mainframe is being driven by private clouds. The System z and, better yet, the zEnterprise can be a prime candidate for the core of a private cloud. In today’s heterogeneous environment platforms, systems, and resources beyond the reach of the inherent virtualization capabilities of z/OS also must be addressed in any private cloud scenario. IBM’s storage hypervisor would be a useful addition to the private cloud, especially where mixed storage is sure to be involved.
With SVC as a component of the storage hypervisor, the organization can pool physical storage resources from almost any disk array vendor and move virtual volumes between any of the resources, either as a snapshot or a mirror copy. From there, it is straightforward to add things like I/O caching, thin provisioning, automated tiering, snapshots and data mirrors, and mobility for disaster avoidance. With the Tivoli Storage Productivity Center you can then add centralized management, health visualization, capacity and performance management, and a storage services catalog combined with automated provisioning and pay-per-use chargeback.
Other vendors are starting to roll out technology to facilitate private clouds. EMC, HP, and Hitachi all have private cloud offerings that include some capabilities of a storage hypervisor. Most recently, the latest versions of Symantec’s Veritas Storage Foundation, Cluster Server, and Operations Manager promise to get organizations to a private cloud and do so without needing to rip and replace existing resources. That’s nice, but the Symantec technology is focused on distributed systems and platforms.
If the goal is an efficient, transparent private cloud, virtualizing the servers is the easy part. It is storage that complicates things because data mobility is essential. Without highly virtualized storage across the wide range of storage devices and tiers the organization can’t move and manage the data as quickly or efficiently as it moves virtual servers. That’s why it needs a storage hypervisor.
For mainframe shops the idea of a storage hypervisor seems strange. As soon as you start to think of the mainframe as a core component of a private cloud that combines distributed and multi-vendor resources it makes more sense. At that point, the IBM storage hypervisor is the way to go.