How to choose your virtual machine backup scheme

Data Center resource management is complex. Many techniques are required to ensure that virtual machines are always available, ensuring rapid disaster recovery, and reliable recovery capabilities.

We talked about some new technologies in the field of virtualization, and some areas still lack effective solutions. In the first part of this article, we will describe the process of virtual machine backup. In the second part, we further discuss the fault recovery and cluster in virtual environment.

Backup

There is not much difference between a virtual data center backup and a traditional backup. Install the backup agent on each of the child operating systems, and then replicate the virtual machine's files, partitions, or entire virtual disks in other locations.

This approach works, but there is a big downside to applying this method in a virtual environment because each virtual machine uses the same I/O channel of the host operating system. Therefore, if multiple virtual machines start backup at the same time, you will inevitably encounter I/O bottlenecks.

To prevent such congestion, administrators should carefully plan backups so that there is a time lag between each virtual machine backup to prevent the child operating system from overlapping during intensive operations.

Unfortunately, this method is not extensible. In other words, when there are many virtual machines, there will inevitably be backup overlap. Because depending on the application requirements, if each virtual disk data reaches 20GB, then each backup may take several hours, so it is inevitable that the backup time overlap.

Sub-operating system backup the administrator has a few things to do when it comes to recovery. First rebuild an empty virtual machine, and then start the virtual machine from the bare Metal recovery CD.

Risky approach.

In addition, an alternative scenario is to do a child operating system backup at the host layer.

Since a virtual machine is a single, stand-alone file stored in the filesystem of the host operating system, just like a spreadsheet or picture file, many virtualization novices may think that backup is a very simple thing. However, it is not the case that backups are much more difficult than they might think.

First, the virtual machine is considered open file, locked by a process or application (think about Microsoft Outlook.) PST Mail archive file). These files can only be accessed in a special way-that is, freezing their state mirrors (which we usually call snapshots), and then performing a backup.

Backup software can perform backup tasks only if it knows how to handle these open files, although sometimes the host operating system assists in backup. For example, Windows Server 2003 has a feature called VSS (Volume Shadow Copy service) that can perform snapshots with a third-party solution.

Even knowing how to handle these open files, we still have to face another challenge when performing online backups: Virtual machines are not just open files, but a complete operating system that accesses the entire set of virtual hardware.

Every time a snapshot is made, everything stops, including virtual memory and interrupt programs. This is called a power outage in the virtual realm, which can damage the filesystem structure of the child machine.

A handful of vendors support this approach, even if a powerful operating system does not cause data corruption during power outages. Vizioncore is a product that supports this method and is very popular with Esxranger. Esxranger can perform virtual machine online backup in VMware ESX Server, and provides a considerable amount of automation.

Massimiliano Daneri has released the famous VMBK script, although they do not support this method, but have bravely tried this method, can also perform basic online backup for VMware ESX Server virtual machine.

Microsoft will start with the well-known service Pack 1 and provide this support for its virtual Server 2005. However, you will not be allowed to use standard Microsoft Backup for backup.

The most common way to backup

Typically accepted, and the only virtual machine method that is truly recognized by a virtual trader is to suspend or shut down a running virtual machine, and then perform backup and recovery or restart the virtual machine. Unfortunately, this process conflicts with highly available services, allowing administrators to leverage traditional agent-based backup methods to back up mission-critical virtual machines.

When the operating system is more adaptive to virtualization, the online backup problem will eventually be resolved. However, it is noteworthy that this second method also brings some pressure on the host I/O channel.

To solve this problem thoroughly, we must change the backup point from the host to the storage facility. Operating a virtual machine file in a storage facility does not directly affect the virtualization platform.

VMware was the first to use this solution, but now its product VMware Consolidated Backup (VCB) has a number of noteworthy limitations: Available only to ESX server; only as a Third-party backup solution agent (which allows the user to have to make a different Product configuration and installation of different scripts); and it cannot perform the recovery process.

At the storage layer, there is also a different method of backup: Leveraging Storage Area Network (SAN) management software and LUN cloning. Typically, this method provides insufficient granularity (granularity) because the storage facility does not recognize the LUN format and therefore cannot provide a single virtual machine backup.

The LUN format recognition depends on the storage management software we purchase and what file system is supported. It may identify NTFS-formatted LUNs and allow us to back up the Windows virtual machines for VMware server. However, it may not support VMFS format and we cannot back up VMware ESX Server virtual machines.

If the LUN format is not recognized, or if we do not have any good storage management solutions, we will only be able to clone the entire LUN. There are multiple virtual machines in the LUN, and even if only one of the virtual machines needs to be restored, we can only restore all the VMS at the same time.