Cost Effective DR Implementation With Virtualization
- Published on Monday, 02 May 2011 15:42
- Written by RICHARD DOLEWSKI
Why do so many traditional disaster recovery (DR) technology strategies under-perform and under-deliver? The simple answer is that a recovery plan using a physical server deployment that is equivalent in number to the production data center can be very complex. Delivering repeatable disaster recovery solutions with non-standard hardware and software platforms with already over-utilized human resources has never been more difficult.
The difficulty is exaggerated when you add additional business applications that require new underlying infrastructure. As companies adopt new business strategies, they often jump directly to a new technology initiative. While such moves often make business sense, they often don’t leave time for realizing the impact to the disaster recovery strategy. Traditional DR strategies require these changes are accompanied by planning, budgeting, and buying twice the required IT capacity. Yes, two of everything! It’s clear that continuous improvement directly impacts capital expenditures for traditional disaster recovery planning.
Traditional Barriers to DR Implementation
- Cost of solution. The biggest barrier to implementing a solid DR foundation is the prohibitive cost of matching production HW components; e.g. a one-to-one system ratio. This capital expenditure can be difficult to justify. Even when the benefits are clear, these are expensive solutions for any business.
- Infrastructure complexity. The number of systems to provision can consume valuable recovery time and funding. This is due in part to supporting numerous hardware components with dual maintenance challenges including the need for base metal restore.
- Reliability and repeatability of solution. Complex solutions are hard to test primarily due to the difficulty of provisioning sufficient equipment to re-create all servers that need testing. Because complex solutions are difficult to test, they are typically not tested frequently, which should leave the business questioning the reliability of the solution.
Can A Virtual DR Strategy Work For You?
The concept of server and storage virtualization is receiving broad support from enterprises in nearly all market segments. Most IT organizations are embracing virtual server infrastructure technologies as a means to cut costs in their production data centers. This technology is compelling in more than consolidation and server deployment activities; it also holds great promise for making DR a cost-effective and reliable deliverable. By implementing virtualization, you can minimize the number of physical servers needed for DR, which in turn allows for reduced complexity and cost by lowering power and rack space requirements. Virtualization can create significant improvements in the speed and simplicity of disaster recovery and is capable of further reducing recovery time objectives for mission critical applications.
Here are some underlying advantages that server virtualization provides when applied to meeting your disaster recovery requirements:
- Reduced direct costs. The biggest cost benefit comes from significantly reducing the quantity of physical production servers required for recovery. With virtualization, this becomes a many-to-one ratio. Consider anywhere from 10-40 virtual machines to one physical server ratio, depending on the type of workload.
- Effective hardware utilization. With virtualization, you avoid the large capital expenditures currently required to deploy under-utilized servers that lie idle in a traditional DR server setup.
- Standard hardware. All operating systems would see the same virtual hardware, allowing ease of migration between physical systems for maintenance or capacity management purposes.
- Standby operations. With physical to virtual (P2V) capabilities, virtual machines can be pre-provisioned with the operating system, application and system state ready for data to be restored.
- Easier testing/proven recovery. The simplified DR environment deployment equates to more frequent and thorough testing of the applications. Concluding a successful test, server configurations are maintained and ready for next use.
- Easier manageability. Recovery of your DR setup and delivery begins with key strokes and mouse clicks allowing for recovery from a remote location.
- Reduced indirect costs. In addition to space and power savings, there will be a meaningful reduction in needed human resources to deliver in a DR scenario.
- Handles tougher RPO, RTO. Additional capabilities exist to improve recovery objectives as business needs justify the additional expenditure.
- Learning curve to implement a new technology. Server, storage and network teams are impacted, creating a potential conflict in traditional administrative paradigms.
- Initial cost and maintenance of virtualization software and hardware. Server hardware and storage selected to meet the performance needs of the aggregate workload can be costly.
- Potential network upgrade. Network components will need to be reviewed to ensure sufficient bandwidth and connectivity are achieved for the virtual infrastructure. Upgrades may be necessary to accommodate aggregate workload performance demands.
- Larger impact with a server failure. Consolidation ratios of 10, 20, or even 40 virtual to physical servers increases the impact of a physical server failure. Careful planning should ensure workloads are distributed among physical servers, keeping application availability in mind; e.g. make sure not all domain controllers or web servers are on the same physical server.
- Requires production to be virtualized to realize full benefits. Full automation of system maintenance and recovery require production systems to be virtualized as well. Not having VMs in production requires mature and strict change control capabilities to ensure P2Vs are captured and provisioned accordingly. In large environments this can be highly burdensome on staff.
In the past, physical server recovery required a complex and lengthy setup process (ghost, tape-based, bare metal recovery at the DR site), and separate hardware. With tape recovery, the DR process can be long, complicated, and error-prone. Finding the correct tape can also be a nightmare if cataloging is not up-to-date.
Virtualization dramatically simplifies the entire approach to managing your DR systems. You can now manage the OS and applications as a single unit. Traditional DR required changing over to new hardware, re-installing the operating system, followed by a process for installing and configuring the application. With virtualization, you can restore the OS and application as a single unit, more simply and much more efficiently.
Server virtualization breaks the dependencies between the operating system and the server hardware. The OS communicates with the “virtual hardware” provided by the software virtualization layer. The virtualization layer effectively makes the virtual machine hardware-independent. The virtual hardware seen by the operating system looks the same regardless of what x86 hardware is actually underneath the virtual machine.
Utilizing a physical-to-virtual (P2V) tool will help in creating virtual machines that are identical to their physical server equivalents. Virtualization software encapsulates an entire server into a set of files containing OS and data volumes as well as configuration files simplifying the components needed to recover a server. This set of files is the virtual machine (VM).
Virtual machines intrinsically possess two qualities that are beneficial to DR:
- Hardware independence. Virtual machines are isolated from the underlying physical hardware. This characteristic of the virtualization makes it simpler to move VMs between physical systems.
- Encapsulation. Virtual machines are encapsulated into a set of files.
In this example, we can use one single physical ESX server on the right to recover seven physical servers on the left. The back-up and recovery process remains the same as in a physical-to-physical scenario, and it still uses the same backup server and storage media that were used in a physical-to-physical server back-up scenario. When there’s a need to recover a system, you can use this archived virtual machine template and have it up and running in minutes, with applications and backup/recovery agent already installed. The great value of server virtualization is seen in how much faster and easier it is to stand up a recovered server. Once the virtual server is powered on, the data restore process will be identical to the restore process in the physical-to-physical scenario.
An increased benefit can be achieved by implementing the virtualization in production. The added benefits are:
- Production servers are already virtual (i.e.the VMDK file already exists, alleviating the need to perform a P2V every time a system change is made).
- Restoring a VM is equivalent to a complete system restore including data reducing RTO.
Traditional disaster recovery plans require many manual, complex steps to perform bare-metal hardware recovery. Virtualization simplifies this environment. Hardware configuration, firmware, operating system, and application installation become data stored in just a few files (VMDks) on your SAN. You can simply protect these files using your back-up or replication software, and you’ve protected the entire system. These files can then be recovered to any hardware without requiring any changes because virtual machines are hardware-independent. With hardware independence, you can repurpose existing servers for disaster recovery rather than needing to buy duplicate servers for DR. In this approach to recovering IT infrastructure, the focus moves from machines to a more holistic view that focuses on capacity, can be dynamically added or removed, and can be turned on instantly in the event of a disaster. So now you can have a “real” recovery rather than a “virtual” nightmare.
Richard Dolewski is chief technology officer and vice president of business continuity services for WTS. Dolewski is a certified systems integration specialist, disaster recovery planner, and is globally recognized as a subject matter expert for business continuity for IBM iSeries and i5 environments.