For many, snapshots are the go to data protection method for protecting Oracle databases. They can provide efficient, rapid protection from database corruption with high data integrity. But snapshot data protection has its limitations. First, snapshot copies can become corrupted if there is an application corruption issue. Secondly, snapshots will be lost if the storage system fails. And then of course there are retention limitations with maintaining snapshot data on primary storage as most systems only support around 250 snapshot copies. Therefore data has to be copied to another device that is independent of the primary storage device.
Consequently, many organizations have come to rely on a combination of point-in-time snapshots along with traditional backups to satisfy their various data protection requirements. The problem is that these backups often take too long to complete, can impact application performance and can be too far out of sync from the production data set to provide value in recovery. As a result, there is a gap in protection between snapshots and traditional backup that needs to be filled.
Backup Infrastructure Complexity
The traditional approach to filling this data protection gap requires a lot of extra data center infrastructure and moving parts. First, it requires a separate backup server to schedule and process backup jobs. Then, backup software agents need to be installed on the application host. Next, expensive networking equipment is required to rapidly backup application data to protection storage. This includes network or fibre channel switches and in some cases, dedicated network interface cards or host bus adapters on the application host to facilitate this data movement.
Data then has to traverse a myriad of infrastructure components before it can safely land on protection storage resources. First the backup server initiates the flow of data by talking to the backup agent on the application server. Data then moves from primary storage to the application server and then across the network to the backup server. The backup server then pushes the data over the network to the protection storage system.
But the real challenge is that this environment rarely remains static. As application data grows, this environment has to scale out to accommodate new data growth – additional backup servers, networking infrastructure, etc. And of course, all of this extra infrastructure has to be monitored, managed and tuned by skilled IT administrators. This in turn leads to more costs and complexity and ironically increases the likelihood of backup problems as there are more moving pieces that are prone to failure.
Hybrid Data Protection
What is needed is a new level of data protection intelligence that combines the best of what snapshots have to offer: frequent, near instant, non-disruptive data protection, with the cost effective retention and reliable recovery capabilities of backups. But the key is this environment needs to operate seamlessly.
In other words, busy application administrators need a way to continually protect their mission critical data without having to manage multiple data protection processes. Ideally, if they could use their existing application data protection tools to control and manage the data protection process, this would greatly simplify operational management.
For example, Oracle RMAN is a backup tool that many Oracle DBAs are very familiar with. For years, some DBAs have maintained a whole separate backup environment in addition to what their backup teams managed. As a result, there are frequently multiple silos of redundant protection storage. One way to eliminate these redundancies is to adopt protection storage systems that can be shared universally, regardless of the backup application source.
But to take it a step further, if organizations could also eliminate extraneous backup application, server and networking infrastructure by going directly from primary storage to protection storage resources, it could facilitate rapid and efficient backup and recoveries without the cost and complexity of traditional backup architecture.
Direct Backups From The Source
The concept of “server-less” backup is nothing new. Nearly twenty years ago, some backup suppliers talked about their plans to develop a way to route data directly from primary storage to backup disk or tape without going through a backup or application server. While there was strong interest from enterprise customers for this capability, the technology never made it off the whiteboard. One of the reasons for this was that backup vendors didn’t own the entire storage infrastructure stack (primary storage and protection resources in addition to backup software) and as a result couldn’t vertically integrate a server-less backup offering. Consequently, the backup data mover has typically been the application server and/or a dedicated backup host.
Backup intelligence that can initiate the transfer of critical backup data directly from the primary storage platform to a protection storage system would be ideal. This would eliminate the impact of backup processes on the application server, reduce cost and complexity by eliminating the need for excess infrastructure, (including a traditional backup application) and enable organizations to protect their critical data sets even when they are faced with exorbitant data growth.
Application Backup and Recovery
From a backup perspective, application owners would still utilize a software agent that had an API (application programming interface) into the application to be protected. But the major difference is the agent would allow application specific backup tools, like Oracle RMAN, to control the backup process from primary storage to the protection storage systems DIRECTLY. This is a critically important because RMAN maintains Oracle backup data in its native application format. This means that if a recovery had to be initiated, the data would not have to flow through a backup application server first. Instead, RMAN could initiate the recovery directly from protection storage to primary storage.
Block Level Backup Efficiency
To help reduce the volume of backup data that had to flow between the primary storage and the protection storage systems, the primary storage would ideally utilize change block tracking (CBT) technology. By analyzing data at a block level and only sending the unique changes over the wire, change block tracking (CBT) solutions can reduce backup data payloads by an order of magnitude.
This can help organizations shrink their backup windows considerably and when combined with deduplication on protection storage systems, this can also help them increase data retention on backup disk resources without having to substantially grow the backup disk footprint. Furthermore, there would be no disruption to the application as CBT processes would take place on the primary storage system.
Granular Backup and Recovery
Moreover, if the protection storage resource deduplicated and used these changed blocks to create full backups in native format, then organizations could gain several additional benefits. First, each backup would complete at the speed of an incremental backup, however, it would appear as a full backup image to the application. This would eliminate the need to concatenate multiple incrementals (sometimes referred to as “reclamation”) to obtain a “synthetic” full backup image.
Secondly, as described above, Oracle could directly access the backup image to perform a restore to production storage resources. But importantly, since there is always a full backup image available in native application format, application owners could perform either a full LUN level restore or granular, single record (table) restore. And since CBT backups on primary storage are non-disruptive to the application and can be rapidly processed, more frequent protection events can be scheduled throughout the course of the day; enabling organizations to better meet their critical application recovery point and recovery time objectives (RPO/RTO).
Protecting application data is becoming increasingly challenging due to the ongoing growth of data and the need to run critical business systems 24x7x365. Unfortunately this trend shows no signs of abating given the ease by which new business applications can be spun up on virtualized server and hybrid cloud data center infrastructure. This means that to keep up with this growth, today’s application owners and backup administrators need more intelligent, less complex and efficient ways to protect data reliably and non-disruptively.
Server-less backup offerings, like EMC’s ProtectPoint, help eliminate backup infrastructure complexity by enabling mission critical applications like Oracle to backup directly from primary storage platforms (initially EMC’s VMAX3) to Data Domain systems. Likewise, application owners can always access full backup images off Data Domain protection storage to perform rapid recoveries of Oracle LUNs to VMAX3 storage systems or perform record level recoveries of Oracle data.
These capabilities are making it possible for businesses to get the best of both world’s from their protection storage environments – non-disruptive, application consistent backup images at the speed of snapshots that also can be very efficiently and reliably stored on Data Domain protection storage systems for retention purposes. This new backup methodology reduces management complexity, helps lower infrastructure costs and enables organizations to meet their critical application backup and recovery SLAs more easily even as these environments grow.