Moving to a virtual infrastructure brings a whole new set of considerations and challenges around managing storage....
These issues range from basic questions on configuring and provisioning storage to storage management and security issues, to backup strategies and disaster recovery.
This tip discusses some management pain points and challenges related to storage for virtualization
Input/Output (I/O) performance
One of the side effects of putting multiple virtual servers onto one physical server is the increase in I/O requirements per server. In addition, increased memory requirements -- and contention for RAM -- can contribute to disk I/O issues by increasing paging and thrashing. While many people consider the need for increased memory as the key to performance in virtualized environments, many initially overlook to connection to storage and I/O performance. Without good planning and ongoing monitoring and adjusting, I/O performance can drag down the overall system performance.
Optimizing your storage layout can help significantly in this area. Using faster storage -- connections and disks where it counts and lower cost storage where it doesn't -- can improve performance while managing costs.
For VMware users, the new VI3.5 release of ESX adds support for several options that can help. For instance, ESX 3.5 adds InfiniBand as a high-speed interconnect and adds 10 GB Ethernet, which can help performance for those running iSCSI SANs -- such as EqualLogic or LeftHand Networks -- as they support 10 GB Ethernet (gigE).
High speed interconnects will help performance both for running more VMs (virtual machines) per physical server or running VMs with more I/O intensive workloads. XenSource has demonstrated 10 gigE support and announced release early 2008. On the lower end, VMware also announced support for SATA drives, giving a lower cost option for local disks, as opposed to SCSI and Serial-Attached SCSI (SAS) drives or SAN attached storage.
Managing storage placement across heterogeneous storage devices and vendors also becomes a bigger challenge as servers go virtual. Moving towards an entire virtual infrastructure can bring many of the same advantages of server virtualization to the storage world, abstracting the logical management from the physical devices. This makes storage virtualization and its services an excellent partner to server virtualization.
Storage virtualization allows the aggregation of storage -- potentially from multiple storage vendors -- into a pool of storage resources that can then be managed centrally. Storage virtualization enables capabilities like migration and replication across heterogeneous types of storage. There are three basic types of storage virtualization: host-based (e.g. Symantec/Veritas Storage Foundation); fabric appliance or switch-based (e.g., DataCore SANsymphony, FalconStor IPStor, IBM's SAN Volume Controller (SVC)); and device or array-based (e.g., Hitachi Data SystemsTagmaStore).
As an example of the move towards marrying server and storage virtualization, XenSource recently announced a relationship with Symantec, bundling their Storage Foundation into the latest release of XenEnterprise.
Once the virtual environment is in place, and provisioning new servers becomes easier than it used to be, virtual server sprawl can kick in, creating virtual server storage sprawl. Provisioning gets easier with virtualization, but managing and troubleshooting get harder. Managing the growth of storage for VMs can become a big issue.
Keep in mind two points here: thin provisioning and VM life cycle management.
- Thin provisioning is the ability to provision a specified amount of storage, but only actually use physical storage as its needed. Thin provisioning capability is offered by 3PAR, Compellent, NetApp, and HDS in their virtualized arrays, and by DataCore across most hardware.
- The second point is to implement some type of VM Life cycle management, creating policies or automated mechanisms to de-provision servers so they don't continue to add to virtual server storage sprawl. This will continue to be an emerging area of management, exemplified by VMware's recent acquisition of Dunes.
Managing by VMs
Managing storage and having visibility down to the VM level has been a challenge for Fibre Channel (FC) SAN storage. Because virtual servers on one physical server share a physical HBA, until recently the virtual servers have had a shared view of the physical HBA. This issue created security exposures and made standard FC capabilities unusable for managing down to the VMs.
Now with N_Port ID Virtualization (NPIV) support and Emulex Virtual HBA technology, there is granularity and the ability to define relationships between VMs and FC WWNs. NPIV assigns a virtual WWN to each VM connecting through a physical HBA. This allows the use of FC zoning and LUN masking down to a VM level, rather than all VMs on a host having a shared view of the physical HBA, making everything work the way you'd expect it to work. There is still one challenge here, in that NPIV currently doesn't support VMFS (Virtual Machine File System), and so can only be used for VMs with storage on raw devices -- Raw Device Mapping or RDM -- and not on VMFS disks. By the way, Emulex representatives say they are working on this.
VMware has just announced another storage management tool which will make storage and virtualization life easier. Storage VMotion gives the same type of live migration capability for VM storage that VMotion provides for moving VMs. For example, you can VMotion all VMs off of a server during planned maintenance for that server, and then VMotion them back when you bring the server back up. Likewise, Storage VMotion allows you to migrate all the storage off of a storage array during its planned maintenance, and then VMotion it back. This capability also offers a potential platform for more advanced storage migration capability, perhaps coupled with VMware Distributed Resource Scheduler (DRS), where storage/disks are moved around to different tiers of storage based on policies, as in Information Lifecycle Management (ILM) or Hierarchical Storage Management (HSM).
Backup and disaster recovery
Just like the non-virtual world, backups are still a challenge, only maybe even more so in the virtual world. Options include VM-based backup with an agent on each host, which brings additional overhead but gives file restore capability; or VMFS snapshots or cloning; or a combination with agent based software for data on VMs and snapshots for server images. Another option is to use consolidated backup like VMware Virtual Consolidation Backup (VCB), which offloads the backup processing to a separate backup server. Also check out Vizioncore, which offers both backup and replication software for VMware VI3 and the new VMware Site Recovery Manager for Disaster Recovery.
About the author: Barb Goldworm is president and chief analyst of Focus Consulting a research and consulting firm focused on systems, software, and storage. Goldworm has spent 30 years in various technical, marketing, senior management, and industry analyst positions with IBM, Novell, StorageTek, Enterprise Management Associates and several successful startups. Goldworm is a frequent speaker, columnist and author of numerous white papers and research studies. She also recently released the book, Blade Servers and Virtualization. Goldworm can be reached at email@example.com.