For IT professionals, regardless of their technology domain focus (server, operating system, storage or networking), VMware and server virtualization are garnering plenty of attention, with the successful initial public offering (IPO) of VMware, the acquisition of XenSource by Citrix and the flurry of vendor announcements timed with the recent VMworld conference (see Storage Vendors Take Aim at VMware and VMware and Symantec Virtualize Disaster Recovery).
Virtual machines, or VMs, are not new, having been around for a few decades in proprietary implementations tied to specific servers such as IBM’s VM operating system running on IBM, Hitachi, Amdahl (now Fujitsu) or NEC mainframes. These and other implementations allowed multiple operating systems (images or guests) to be consolidated from multiple physical machines (footprints) into different logical partitions (LPARs, now known as VMs) to isolate applications and workloads, enable consolidation and improve resource utilization.
Virtual servers in the open systems world are being focused primarily — initially at least — on supporting server consolidation to improve utilization, which also means consolidating storage for applications and data as well as protecting the consolidated servers. In addition to being used for server consolidation, virtual servers or VMs are being deployed to facilitate server and operating system management as well as disaster recovery (DR) and business continuance (BC) initiatives. While dedicated and shared internal storage continues to be shipped and supported by vendors with their servers and blade center servers, external networked storage using either Fibre Channel, iSCSI or NFS– and CIFS-based networked attached storage (NAS) devices allow VMs to be moved across different physical servers independent of where data and configuration information are stored to enhance resiliency and enable BC and DR.
With VMware fast becoming a popular tool for virtualizing and consolidating physical servers and their operating systems, it should come as no surprise that IT vendors want to jump on and ride the VM bandwagon. Many of the vendor announcements surrounding VM support have been tied to managing and provisioning virtual and physical resources as well as data protection, including backup and BC/DR capabilities.
Backing Up VMware
Why is backing up on VMware such a big deal? It seems everyone’s got a product for it. Does it really require a specialized tool? Issues with backing up in a VMware environment can be as simple as basic interoperability support: does your existing backup solution running on your particular operating system work when running as a guest on a virtualized machine such as VMware ESX, and is it supported by your vendors?
When looking at VMware-related products and announcements, consider what is actually being done or enabled. Since VMs are stored in files on disk, they can be backed up similar to an open file and moved (VMotion) locally or across ESX implementations. While VMs provide crash-consistent recovery for reboot, application state and changed data is not captured without using third-party advanced data protection tools. VMware provides for rapid server restart in the event of a server failure or crash, but without integration of third-party technology, including application-aware snapshots and data replication, application and data recovery is dependent on the recovery point of when the last good known backup was performed and the timely restoration to that point.
For a backup product, VMware support could mean simply supporting an agent or client running on a guest operating system in a VM similar to a traditional physical server. VMware support could also mean integrating with VMware Consolidated Backup (VCB), including application and transaction aware data protection of guest operating systems, their application and data. Another layer of support could be that the backup software’s media and backup manager can be installed in a VM on a guest operating system instead of running on an external physical server. VMware backup support could also mean integration at the VMware ESX console layer, among other variations.
Storage-related applications and functionality could include backup and snapshot software, replication and continuous data protection, as well as storage services functions installed and running in a VM. Storage servers could include virtual tape libraries (VTLs) or backup and archive target devices as well as virtual storage servers. For example, LeftHand is now enabling its SANiQ software so it can be installed in a VM to virtualize internal dedicated storage on an ESX server and co-exist next to other VMs hosting other operating system and applications.
Similarly, EMC has enabled its Avamar backup target repository software to run in a VM as a target for Avamar agents installed on operating systems running in other VMs on the same physical server or from other Avamar agents on other servers. Given the workload and demand for server CPU, memory and I/O resources, installing a backup or archive target in a VM next to other applications is not generally going to be based on a performance value proposition. Ironically, IBM which happens to be a server, storage and virtualization vendor, has had the ability to run guest applications in its DS8000 high-end storage array based on the IBM p5 PowerPC hyper visor-enabled processor for several years, but Big Blue has not actually done anything with what it refers to as partitions (VMs) other than to talk in theory of what could be done, at least from a shipping product status.
The value proposition of a consolidated server and application workload running in different VMs on the same physical server that is also hosting a storage virtualization appliance in a VM is not one of performance. Instead, the potential benefit of hosting a storage virtualization or storage services appliance in a VM is in using what might otherwise be stranded or orphaned internal storage and making it accessible to other applications, perhaps for archiving or other low-performance applications, while eliminating the cost of having a dedicated server appliance for the storage function. In other words, look beyond increasing utilization to consider the performance and stability aspects of resource virtualization.
As Norm from “Cheers” used to say, don’t ask “What’s going on, Norm?” Ask instead: “What’s going in Norm?” The same could be said of IT vendors: ask them not just what’s going on VMware in the form of agents in guest operating systems and VMs; ask them what’s going in VMware in the form of tighter integration with ESX’s underlying capabilities. In other words, sift through the virtual announcements to find out what the reality is for virtual server and storage support capabilities.
Greg Schulz is founder and senior analyst at the StorageIO group and author of “Resilient Storage Networks” (Elsevier).