IBM Pulse 2012: A New Storage Hypervisor

IBM is promulgating a storage management concept that it calls a “storage hypervisor” though the final product name has not been determined. The company claims the technology will offer benefits – such as better storage utilization – leading to better storage cost economics, and data mobility – leading to increased flexibility, such as non-disruptive storage refreshes – and parallels the acceptance of server hypervisors and virtualization. But there are also broad implications about how storage will be deployed and managed under IBM’s hypervisor solutions and strategy that you may find worthy of attention. Let’s see why.

Last week’s IBM Pulse 2012 conference in Las Vegas had the imposing sub-title “Optimizing the World’s Infrastructure.” And the company attacked a broad range of both physical and digital infrastructure issues under the now-familiar – at least to the IT world – integrating concepts of Smarter Planet and Smarter Computing. But Pulse attendees wanted not only an overview of the big transformational trends in IT, but also to be able to drill down in their areas of particular expertise so that they could return home with a game plan or set of action items based on what they had learned. Rather than going into breadth of coverage from Pulse 2012, I will concentrate on an area of focus for me – storage management – to illustrate that type of specialization.

Storage management was such an area of particular attention for specialists, with emphasis on the concept of the storage hypervisor that IBM is working diligently to stress. Now, all IT roads lead to storage – as without data processors and networks can do no useful work and all data not in transit has to reside on some form of storage medium – so deploying and managing storage more efficiently and effectively is critical, not only to today’s storage infrastructure operations, but also as a cornerstone to moving to the cloud that offers true IT-as-a-service.

Enter the storage hypervisor in general, and IBM’s Storage Hypervisor in particular. Now the term “storage hypervisor” is not generally accepted currently in the IT industry, as only two smaller companies – DataCore and Virsto – in addition to IBM, seem to be advocates of the term. Moreover, other terms, such as “virtual storage,” may be used instead with different approaches that yield the same essential capabilities. Still, after understanding what it does, the term should provide a good mental recall mechanism for understanding what is happening, and should happen, to the underpinnings of a hypervisor-enabled storage infrastructure.

For simplicity’s sake, I’ll focus on what IBM is offering. Note that while “storage hypervisor” may be a concept, IBM implements the concept through real products. The company believes the storage hypervisor is a combination of application software that performs the necessary storage virtualization functions, and management software that provides the centrally, automated framework for all virtualized storage resources. The “actor” software underlying the whole thing is IBM’s System Storage SAN Volume Controller (SVC), and the “director” software is the IBM Tivoli Storage Productivity Center (TPC). To this, IBM also adds the IBM Tivoli Storage FlashCopy Manager, as it feels that the special snapshot capability incorporated as part of the storage hypervisor is an essential ingredient.

Now, the first question that many might ask is: Isn’t the storage hypervisor simply a re-bundling of existing IBM products? While the answer at first blush would be yes, a little closer examination says that the synergies that this new combination brings might not have happened if the products were used individually. Moreover, putting the combination under the rubric of storage hypervisor better aids in understanding what it does, its benefits and the larger implications.

Obviously, the use of a storage hypervisor invokes the concept of the server hypervisor in the minds of CIOs and other IT professionals. The server hypervisor – and server virtualization – were once relegated to enterprise-class mainframe computing environments, but are now considered a “good thing” – albeit with some caveats, perhaps – in server systems of most every type. Although storage virtualization has been around for a long time, it has not realized the same level of attention or success that has occurred on the server side. That has to change, as explosive data growth and tight budgets can no longer keep up with just the falling cost of storage alone. Thus, IBM’s storage hypervisor may provide a mental rallying point around which a next stage in storage infrastructure evolution can take place.

A storage hypervisor creates a single pool of managed storage that can span across multiple storage arrays or even JBODs (Just a Bunch of Disks) boxes. Now, virtualized storage (even in a single array) divides up storage in a SAN array differently than the traditional method. Traditionally, shared storage in a SAN means that each application is allocated a portion of the available physical storage, based on a guess of what it will need over time.The problem is that trying to predict future storage use on an application-by-application basis is likely to be doomed to failure and leave a lot of unutilized space. For example, let’s say that Application A “rented” a two-bedroom storage “apartment” but only used one bedroom. That leaves a lot of wasted space. On the other hand, Application B is running out of space in the one-bedroom storage apartment it once thought was more than big enough. Reprovisioning in a typical storage environment is hard. Using an approach called “thin provisioning,” the storage hypervisor offers applications virtual storage “apartments” that provide as much storage as they want – within reason and as long as the overall physical storage is not exceeded. Note that an analogous process occurs with virtual machines on a physical server.

The storage hypervisor also decouples storage services from underlying physical media, i.e. disks. What does that mean? A SAN storage system derives its value from both its associated hardware and software. The hardware provides both higher availability – such as no single point of failure – and performance – such as the use of a sophisticated controller cache – to provide differentiation from, say, JBODs. The software provides added value in what are called storage services, such as remote mirroring or replication capabilities, and various forms of snapshots.

These capabilities are essential for many tasks, including such data-protection activities as local backup and remote disaster recovery. However, these capabilities are traditionally associated with physical media, such as particular disk LUNs (logical unit numbers). When an application needs to add LUNs or change existing LUNs, the process may not be easy. A storage hypervisor essentially changes the paradigm to data-centric storage services – designed to meet the often rapidly changing requirements of applications/information – rather than media-centric storage services – limited to the characteristics of physical disks, tapes and arrays.

That means that in hypervisor-enabled environments, storage services accompany the data and data can easily be moved virtually from one physical instantiation to another. Moreover, it is easier to apply different storage services to different sets of application data; for example, mission-critical data requires greater high availability (HA) requirements for processes such as remote mirroring, as contrasted to year-old e-mails that have to be safely preserved for e-Discovery purposes, but do not have to be instantly retrievable.

It should be obvious that life for an administrator utilizing the storage hypervisor schema can be devoted to more value-added tasks since a combination of IBM SVC makes storage virtualization happen, and IBM TPC makes not only managing changes themselves, but also managing at a more granular level, significantly easier. That IBM’s management process can address data on virtual volumes across multiple tiers of storage – including tier 0 SSD flash memory, tier 1 FC/SAS, and tier 2 SATA – across disparate storage systems – such as IBM XIV and DS8300 systems, but also with storage arrays from another vendor – and from site-to-site – with probably some reasonable limitation on distance – is the icing on the cake. Although no cloud is required, a storage hypervisor sounds like an essential good mix-and-stir ingredient for a private, public or hybrid cloud.

At IBM Pulse 2012, IBM customer Ricoh testified to benefits, such as cost savings, it had garnered by using IBM-originated storage hypervisor products. That is a type of benefit that most customers would strive to achieve, but IBM likes to use an example that makes data mobility without disruption – i.e., no downtime of applications and their data – even more dramatic than migrating data from one array to another during a technical product refresh. Let’s say that you have one site in the impending path of a major hurricane and another site situated safely outside the potential path of destruction. Let’s say that you have a server hypervisor (such as VMware vSphere for Intel servers and IBM PowerVM for Power servers) and the IBM storage hypervisor platform. With IBM SVC Stretched-cluster – part of the IBM Storage Hypervisor where SVC supports servers and storage at two geographically separate sites – the same data can be accessed at each site, and it also supports the ability to perform a VMware vMotion and IBM PowerVM Live Partition Mobility (LPM) move non-disruptively to end users. Can your sites do that?

But wait, there’s more. In moving to a cloud, a services catalog is essential so that users can easily select the services they need, what IT-as-a service is all about. The implementation of a storage hypervisor enables the development of a storage services catalog. IBM believes that each company has roughly 15 different data types – such as e-mail, database, word processing documents and video – that each requires distinctive service levels across four dimensions; capacity efficiency, I/O performance, data access resilience and disaster protection.For example, a revenue-generating online transaction program (OLTP) needs a lot more I/O performance than years-old sales histories that are used occasionally for analytical purposes. A one-size-fits-all-data approach can be very costly, such as supporting unnecessary copies and swift recovery capabilities from a remote site in the case of a disaster. In other words, a storage hypervisor can save MONEY through capturing efficiencies that not only improve storage utilization, but also use I/O resources more carefully and avoid the need for unnecessary copies or replication capabilities, such as providing excessive, unnecessary networking bandwidth).

Still, a little rain must fall.IBM’s storage hypervisor approach is currently limited to block data, but a unified storage approach that also embraces NAS file storage could very well appear in the future. No, that is not the caveat. The issue is that one of the true benefits of IBM’s strategy is to use a storage hypervisor to virtualize heterogeneous storage. That is, IBM wants to virtualize the storage of one or more of its competitors’ solutions. Now, naturally enough, competitors want to do the same to IBM as the vendor that controls the storage hypervisor – or equivalent set of capabilities under another name – is essentially king of the storage hill. No one buys, say, enterprise-class storage, for the hardware characteristics alone, but rather also for the storage services that go with them. And the IBM storage hypervisor would replace those services with services of its own. Hence IBM would be in the catbird seat when storage technology refreshes occur.

Fortunately for IBM’s storage competitors, IT is unlikely to want to get rid of existing storage services in which they have sunk investment for both equipment and personnel training, and for which the switchover could be difficult, as it is not just a matter of turning off one service and turning on another, such as a home Internet service provider. In terms of economics, switching costs often provide a significant barrier to change.

Still, what is likely to happen is that there is likely be more and more pressure on single-vendor storage shops to make necessary changes to achieve or adopt a single-vendor storage cloud. Heterogeneous server vendors may very well co-exist in the cloud, but the tendency would seem to be towards “standardization”, namely, one vendor.

Mesabi Musings: IBM Pulse 2012 covered a sweeping panorama of the world’s IT infrastructures and what IBM and its partners are doing to “optimize” those environments. Storage management is a key component of the overall infrastructure mosaic and IBM directed attendees’ attention to the need for a storage hypervisor. Now, a storage hypervisor can deliver storage cost efficiency and data mobility as standalone benefits, but IBM’s approach can also provide the underpinning for broader service level benefits – such as more effective use of I/O capabilities – that are necessary for providing storage-as-a-service as part of an overall IT-as-a-service strategy in a cloud.

The rain cloud inhibiting the adoption of “true” cloud computing is that IT may resist replacing existing storage services from vendors that compete with IBM because of perceived switching costs. Still, the argument about standardization will set the tone for discussion in future years as the cloud would seem to demand standardization. How IT will react will be very important in whether or not the cloud in a true IT-as-a-service sense, rather than partial implementations that still leave some non-integrated information silos –with their implied cost inefficiencies – in place. So IBM’s storage hypervisor is really stirring up the IT storage cloud pot.

IBM is currently a client of David Hill and the Mesabi Group.