Published: 01 Jun 2010
Almost every IT shop has virtualized at least some of its servers, putting much of the burden of server consolidation on data storage systems.
By Jeff Byrne
Managing data storage resources has proven to be a challenge in virtual server environments since server virtualization first took root in data centers approximately seven years ago. By consolidating multiple virtual machines (VMs) on a single physical system, server virtualization imposes additional pressure on storage and I/O resources, and complicates functions such as data protection and storage management. In a recent Taneja Group end-user research study, we learned how your storage management peers are coping with the impact of server virtualization, and how they're attempting to employ the technology to their advantage. A total of 360 IT managers from a mix of North American enterprise and midsized organizations gave us their perspectives via an online survey, supplemented by follow-up phone discussions with selected respondents.
Moving beyond Fibre Channel
A majority of respondents use Fibre Channel SANs (FC SANs) as their primary networked storage platform, and it's still the leading virtual server storage protocol, used by 75% of respondents. But iSCSI (used by 36% of those surveyed) is the fastest growing storage choice for virtual servers. While most initially considered iSCSI because of its lower cost, they tended to be pleasantly surprised by the functional strengths and performance of current iSCSI offerings from suppliers such as Dell EqualLogic and Hewlett-Packard (HP) Co. (with its LeftHand Networks acquisition).
NFS is becoming more of a factor on the NAS side: While just 19% deploy it overall, NFS is used by three out of 10 enterprises (i.e., organizations with 10,000 or more employees) and four out of 10 companies with at least 1 PB of total storage capacity. Large firms are adopting NFS for two primary reasons: ease of management and its virtual server data protection capabilities. We believe enhanced functional support by vendors such as VMware and NetApp is also playing a role. In addition, the growing interest in NAS is driving significant innovation from scale-out NAS vendors such as Isilon Systems and HP with its StorageWorks X9000 Network Storage Systems.
Traditional backup for VMs still preferred
If you're using a physical system backup approach to protect your virtual server data, you're not alone. More than 75% of our survey respondents are using one or more physical server data protection processes to back up their virtual servers. Most users said they opted for physical system backup because they wanted a proven and familiar backup approach; many also cited the cost and convenience of sticking with backup software from vendors they already work with. The most popular method is agent-based backup inside a guest OS (53%). File-level backup is the leading approach, which is consistent with the requirement (stated by 57%) that virtual server files be directly recoverable in a single step, rather than first having to restore a virtual machine image.
While current virtual server data protection practices suggest that many storage managers are averse to change, we found 30% of respondents are planning or strongly considering new backup approaches tailored to virtual servers, such as the use of hypervisor-based VM snapshots (currently used by 25%). Array-based snapshots and VCB-style proxy server backup (with nearly 30% and 27%, respectively) were also mentioned as data protection practices likely to be deployed in the next 12 months for at least some virtual server workloads.
Virtual servers not yet key for disaster recovery
Improving disaster recovery (DR) capabilities is often cited as one of the top reasons for virtualizing servers, but our survey revealed that virtual server-enabled DR is still in the early stages of adoption. Three out of four users have overall DR or business continuity (BC) plans in place; but despite virtualization vendors' best marketing efforts, only a minority of users have committed to DR strategies built around virtual server technology. For example, 40% of those surveyed are using capabilities like VMware VMotion to automate the movement of encapsulated workloads, and only a fraction of those are employing the technology to migrate workloads from a data center to DR site. Of course, technology constraints currently limit the effective range of such tools, and we expect their use for disaster recovery to increase significantly once those limitations are removed.
A large majority of end users also indicated that they're not yet able to rely on their virtual infrastructure to help them rapidly recover from a disaster, even though the frequency of outages in their virtual server environments is relatively high. Case in point: 33% have experienced at least one disk-related outage or failure in the past three years. Interestingly, most enterprise IT managers we interviewed still tend to think about disaster recovery from the storage side, focusing, for example, on remote replication capabilities in the array. It appears that server virtualization providers still have a lot of work to do to persuade and fully equip their customers to invest in virtual server-driven DR practices.
Storage management remains a challenge
Nearly 80% of respondents point to specific storage barriers that have slowed the progress of their virtual server deployments. The leading impediment is the increased complexity of data storage capacity and performance management (cited by 37%), followed by challenges in application-to-storage mapping and data protection (each noted by 28%).
To address the first issue, storage managers are increasingly using data reduction and related technologies to improve virtual server storage utilization and performance. For instance, in the next 12 months, respondents plan to deploy data deduplication (44% will deploy vs. 27% using it today); virtual server or array-based thin provisioning (46% will deploy vs. 33% today); and data compression for primary storage (9% will deploy vs. a little more than 5% using now).
VMs and storage: A work in progress
Storage managers have made great strides in defining and enhancing their virtual server storage practices during the past few years, but a lot of work remains to be done. To gain the upper hand, we strongly encourage IT practitioners to view storage as a strategic and integral component of their virtual infrastructures, and to give storage equal weight with servers in planning and deployment. We'll continue to keep tabs on how users address these issues as the book on virtual server storage best practices continues to be written.
BIO: Jeff Byrne is a senior analyst and consultant at Taneja Group. He can be reached at email@example.com.