Evaluate backup needs Virtualization Server New Policy

Virtualization refocuses people's focus on managing efficiency and efficient use of valuable computer resources, but as physical servers host more and more virtual machines (VMs), IT administrators face a new set of data protection And backup challenges, now no longer simply copy important data files even if they are backed up, the state of each VM needs to be protected and kept accessible at all times.

After implementing a virtualized infrastructure, each organization should re-evaluate backup needs and choose the most appropriate technology to provide optimal data protection. The backup requirements for virtual servers are basically the same as for traditional non-virtual servers. Backups are designed to protect against data loss caused by hardware, operating system, software, or human error. However, the backup requirements in the virtual environment also have their own particularities, and the backup strategy cannot completely copy the previous method.

Traditional Backup and Virtual Environment Backup

First, virtualization infrastructure often needs to compete for more data. The more applications generate more files, the larger each, the VM itself. Also a big file like VMware. Vmdk files, all of which need to be protected. Since VM files are essentially a server memory image that reflects the state of the VM operating system at a certain point in time, VM backups are very frequent.

The traditional non-virtual server backup process usually consumes 100% of the server processor and I/O resources. It is very difficult for other processes to run normally during the backup. After virtualization, there may be 10 20 or more VMs share a part of the resources of the server, but the backup process is performed within the resources allocated to each VM, so it does not affect other processes.

If backing up a single VM causes excessive resource consumption on the physical server, other running VMs will be adversely affected. This problem should also be considered when recovering from VM backup. Each VM will be used again when it comes back online. Server resources, leaving less resources to restore VMs in the future, will be longer and longer, so the time to restore VMs will be longer and longer. This is often overlooked when setting up backup policies in virtual environments.

Server virtualization also places higher demands on storage and networking. Traditional servers typically use local storage, but most virtual data centers store virtual servers centrally on shared storage, such as SANs. , including the data generated by the application, the latest image of the VM and some backups, therefore, the SAN must be able to provide data exchange capabilities for dozens or even hundreds of VMs.

The network connection between the storage and physical servers must also support data communication generated by a myriad of independent VMs, which is why IP-based SANs such as iSCSI (or NAS) can satisfy many application scenarios, but data. The center still needs to deploy a standalone, high-performance Fibre Channel SAN.

The ever-increasing volume of data, increasing server computing resources, storage and network requirements make backup strategies more complex, and previously created backup strategies may no longer meet today's needs. It's time to consider re-engineering or modifying your backup strategy.

Transforming existing backup strategies to meet virtual server needs

Virtualization does not exclude tape, and companies that now use tape backup can continue to use these products, for example, each VM is independent The server exists, so adding a tape agent to each VM is as simple as using the tape as a VM backup target.

Tape is relatively cheaper, and many backup tools support tape backup. In fact, organizations that currently use tape for backup continue to use existing tape media when implementing server virtualization.

For small or non-critical virtualization deployments, tape is sufficient, but server computing resources and network performance can quickly become overwhelmed, especially if multiple VMs are backed up at the same time. The main problem now is time.

Before the backup, you must shut down the VM, so the user cannot access the VM during the backup. Due to the competition between computing and network resources, the VM backup cycle may become very long, so the backup VM is traditional. Tapes can cause unacceptably long backup windows.

RayLucchesi, president and founder of Silverton Consulting, Colorado, said: "There will be a lot of network bandwidth, storage bandwidth and throughput during backup, so you should not be between 8 am and 5 pm This service performs backups during peak hours, otherwise the performance of the business system will be drastically reduced."

Unfortunately, there is currently no effective way to mitigate the disadvantages of the direct tape backup architecture, the most common ones used to improve The performance approach is to use disk storage. Virtual Tape Library (VTL) technology is also a way for backup software to back up data to disk storage by simulating a tape system. The disk storage target can be a SAN or NAS storage subsystem. The tape drive usually creates a second copy of the VTL backup. Most organizations that use virtualization technology are finally welcoming a new disk-based data protection solution that reduces the importance of tape as a long-term or archive storage role.

The main challenge of virtual server backup is to shorten the backup window and keep the network traffic at a reasonable level. Virtualization technology can provide a powerful backup solution in combination with SAN.

For virtual servers, the most common disk-based data protection technology is snapshots. As the name suggests, snapshots are copies of VM files that are saved to high-performance storage (such as Fibre Channel SANs) at specific points in time because Snapshots can be completed in a matter of minutes or less, and the application is interrupted for a short time.

In many cases, the user doesn't even realize that a snapshot has taken place. The snapshot can capture the entire virtual machine, or it can be incremental, that is, only the changed part is collected, and the virtual machine snapshot is saved to the storage. Once you've got it, you can use it for multiple purposes, copy or mirror it to a remote disaster recovery facility, clone it to another server, or even copy it to a dedicated backup server, and finally transfer it to tape for storage, such as VMware Unified Backup ( VMwareConsolidatedBackup).

Snapshots also need to be managed. Each snapshot requires storage space, so storage monitoring and capacity planning can be very useful when using snapshots. In addition, you may only store a limited number of snapshots, so Be sure to configure the snapshot tool to delete the old snapshot and free up the storage space for the new snapshot. After performing the full VM snapshot, clear the data buffer in time. Otherwise, the VM status may be inconsistent or the VM file may be Damage, further making recovery more difficult.

A similar technique to snapshots is Continuous Data Protection (CDP), which tracks and records any changes to the data set, such as the state of the virtual machine. Although you can perform snapshots frequently, CDP is more suitable for changing. Virtual machine, CDP continuity reduces the need to empty the buffer, but still needs to manage storage consumption.

VM live migration between physical servers is clearly not a backup solution, but it can affect the data protection behavior, administrators must take it into account, although VMs can be migrated as needed, they must be able to To access the data, the application also needs to know where the virtual machine has migrated.

Many times, in order to accommodate virtual machine migration, data files are also moved, and administrators must consider migration and interaction with snapshots, replication, backups, and other data transfer tools.

There is another trouble to consider. The VM usually resides on the virtual platform - Hypervisor. This means that the underlying system platform also needs to be backed up, usually including an operating system and hypervisor, such as VMwareESX, CitrixXen. Or Microsoft Hyper-V.

Greg Schulz, founder and senior analyst at TheServerandStorageIO Group in Minnesota, said: "Virtual server backups need to consider restoring clients and the entire system."

Any backup strategy or policy change Both should include backup testing. In a virtual environment, testing disk-based backups is usually easier than testing traditional tape backups. Snapshots and CDP files can be quickly restored to idle servers without any impact on the production environment, which makes training IT People and presentations have become easier.

More Effective Backup Strategies

Today's data protection technology is more flexible than traditional tape systems. As you saw earlier, snapshots of each virtual machine can be captured in minutes. If the incremental acquisition may take only a few seconds, it will not affect the production environment.

Deduplication can be considered for storage systems. VM files contain a lot of vacant space and redundant data, for example, if you run 50 virtual machines on 10 servers, host servers and clients. The operating system running on the side may be the same, eliminating duplicate data and greatly reducing the storage space required for snapshots. Smaller data sets mean faster backups to other media or replication to remote sites.

In any case, data protection always requires some storage space and network bandwidth, so you should try to stagger the VM snapshot scheduling time, for example, if a physical server hosts 10 virtual machines and 10 virtual It may not be a good idea to perform a snapshot, and it may be reasonable to perform a snapshot of one or two virtual machines so that the high peaks of storage and network I/O can be effectively controlled.

Finally, it is common for virtualization users to adopt multiple data protection schemes. Different approaches are adopted for each VM's importance to the organization. For example, CDP can be used to protect mission-critical virtual machines. A virtual machine only needs to perform a snapshot every hour, and a non-critical virtual machine can only take a snapshot a few times a day. All data at a specific point in time can be periodically replicated to the disaster recovery site, or other backup media such as VTL or tape. Administrators should develop a reasonable backup strategy based on the specific circumstances of each virtual machine.