Backup Software is Broken but Doesn't Need to be Fixed

Why backup software, as we know it, is dead and how virtualization is a game-changer.

By Ash Ashutosh, Founder and CEO, Actifio, Inc.

For a few years now, people have been saying that backup software is broken. Saying that something is broken implies that it can be fixed, but the truth is that traditional backup software is totally irrelevant. It's like saying the horse-drawn carriage is broken because it doesn't perform well on today's highways.

The reality is that today's data management requirements -- including the amount of data being protected, the need for data to be used in multiple ways, the dynamic nature of a virtualized application environments, and disk as the new backup medium (replacing tape) for data -- are all counter to the core design of backup software.

Companies need a new class of data management and storage solutions that support evolving business requirements and are explicitly designed to better manage copies of data and deliver more than just 100 percent data protection. They need protection and availability storage (PAS).

The Four "Dirty Phrases" of IT: A Brief History

To understand why backup software is not equipped to function in a modern IT architecture, it is important to understand its past. Today's backup software is rooted in the use of tape as the low-cost medium for retaining copies of data; everything about backup software is based on the behavior of tape media. Five basic functions form the core of backup software: reading data from disk; transforming disk-based block formatted data into a format designed to stream to tape; creating a catalog of the files being reformatted and writing to tape; restoring data back from tape format to disk; and managing the tape library and media.

Over the years, backup software based on these functions has left a legacy of phrases considered to be some of the most dreaded in the IT industry:

-- Agents: Deployed on every server, agents crawl file-system trees and package data to be moved to tape. Unfortunately, they interfere with the performance of business applications. Vendors have created an entirely new suite of management software intended to distribute and update agents.

-- Backup window: A period during which application performance grinds to a halt while backup agents eat away at the CPU, network, and disks to make copies of data to tape, leaving business data vulnerable to loss.

-- Restore window: The lengthy process of restoring data from tape. Backup software has to locate data on the tape, read it, and write it back to a disk connected to a server.

-- Tapes: Quickly becoming the 8-Track or VHS of the IT industry, physical tapes brought multiple challenges of handling, storing, and moving to vaults; IT hoped it would actually be able to restore the much-needed data requested by the user.

Plugging the Holes

As IT architecture evolved, those traditional backup methods quickly became outdated. As a result, several tools (such as snapshots and data duplication technology) have been employed to alleviate emerging data management growing pains.

Snapshots reduce the time it takes to capture and recover data. Snapshots are linked to production data, which means that if production data is lost, all the snapshot data is lost as well. This co-dependency on production data results in snapshots being only used as point-in-time copies of production data, for use over short periods, but cannot be used as backup. In addition, retaining snapshots for long periods becomes prohibitively expensive, as a snapshot data resides on expensive production data disks.

Data duplication technology, along with the commoditization of disks, provided a path to eliminate tapes, significantly reducing the operational burden on administrators. Backup software now reads application data blocks from disk, converts those blocks into a stream of data intended for tape, and writes to a deduplication device, which then stores these streams back as blocks. In addition to the obvious futility of these pointless data translations, the real culprit -- backup software -- remains. Although tapes are eliminated from backup in this scenario, users continue to live with agents, backup windows, and restore windows.

Flip a Coin: Heads I Win, Tales You Lose

You can't win with today's backup methods. They present false choices that require enterprises to prioritize data sets and choose different recovery point objectives (RPOs) and recovery time objectives (RTOs). Enterprises also have to decide how much data they can afford to lose and which applications will not be protected, followed by the laborious process of selecting individual point tools to meet the various RPOs/RTOs and retention time requirements. This practice results in a complex and expensive infrastructure consisting of five to 12 different point tools that create an operational nightmare. What's worse, it's nearly impossible to validate their effectiveness.

PAS: The New Path Forward

With data now moving from disk to disk, it's time to rethink and redesign backup. It is now possible to capture, store, and restore application data without having to convert it to another format. In fact, it is now possible to have applications directly access point-in-time copies of data from the backup system, just as they would from a primary storage system, eliminating the traditional "restore" operation. These design points are at the core of virtual data pipeline (VDP) technology, which is used to deliver a new class of storage: protection and availability storage (PAS).

VDP introduces virtualization into data management, bringing the benefits of reduced costs, lowered operational expenses, and radical simplicity of use. VDP is based on the premise of capturing data once and reusing it for multiple applications. It virtualizes the core primitives of data management (copy, store, move, and restore), and consists of an object file system that enables instant creation of virtual copies of point-in-time data from the collection of unique blocks of data.

It also enables a single solution to replace one or more existing tools -- including backup software, disaster recovery, business continuity, or test and development solutions – and can be used as a platform for search, compliance, and analytics tools. The resulting simplicity of operations and reduction in infrastructure dramatically drive down costs compared to legacy methods.

VDP begins with virtualizing the capture of production data, using a device-independent snapshot with change block tracking. These blocks are then stored as objects with associated application, time and service-level agreement (SLA) information in an independent, virtualized storage pool, ensuring that there is a true backup image of the production data.

Based on the SLA, VDP then sends the data to a remote site for long-term retention or disaster recovery and business continuity. When a point-in-time copy of data is needed, PAS instantly reconstitutes the data in the required application format and presents it to the application server through a Fibre Channel, iSCSI, or NAS interface.

The result is a single solution that transforms the chaos of multiple silos of infrastructure and applications traditionally deployed for data management into a single, SLA-driven solution that does far less copying, storing, and moving of data.

Organizations that suffer through their reliance on traditional backup software need to ditch their horse-drawn carriage, eliminate backup software, and embrace new solutions that support virtualized, dynamic IT environments with rapidly growing data sets. Let's stop trying to "fix" backup software. It's time to consider radically new technologies and solutions, such as PAS, that remove complexities and enable organizations to meet business requirements without crippling IT resources and budgets.

Ash Ashutosh is the founder and CEO of Actifio, Inc. (http://www.actifio.com), the Protection and Availability Storage (PAS) platform company, pioneers of industry's first storage system optimized for managing copies of production data. You can contact the author at ash.ashutosh@actifio.com.

comments powered by Disqus