Whistler Offers New Storage Features

Backup andrecovery is an essential function for mission critical systems, so it is ironicthat backup and recovery software is difficult to certify for Microsoft’splatform for mission critical systems, Windows 2000 Datacenter Server.Microsoft intends to address this issue in its next release of the Windowsoperating system, Whistler.

Whistlerwill have native functionality that enables administrators to take “snapshots,”a copy of system settings and information at a single moment in time, then sendthe snapshots to third-party back up and recovery software for storage on tapeor other backup media.

Backup andrecovery software are kernel-touching applications: They communicate directlywith the operating system’s kernel. Although this is necessary for informationon the computer to be stored to tape, kernel touching applications can alsoincrease the chances of an operating system going haywire. Because Microsoftwants Datacenter Server to compare with proprietary Unix and mainframe systemsin terms of reliability and availability, it made it difficult forkernel-touching applications to receive certification. Redmond had found thatkernel problems were a major factor contributing to system failures. While thedecision helps ensure the uptime of mission-critical systems, it does little toaddress scenarios when the system may fail.

Backup andrecovery software from vendors such as Veritas, Legato, and Computer Associatesoften rely on agents, small low-level programs that gather information about asystem. Agents sit locally on each server and send information to the backupand recovery console, which then writes the information to tape. In addition tostorage agents, anti-virus software and device drivers are also kernel-touchingsoftware.

Inlieu of kernel-level agents,Whistler will use a function for taking system snapshots. Presumably, Microsoftwill ensure that the snapshot functionality will not conflict with the rest ofthe operating system, guaranteeing both a stable and recoverable operatingsystem.

SteveWiden, research director for storage software at IDC, is reluctant to indictbackup and recovery software as a cause of system crashes. “Backup and recoveryis a pretty stable product set,” he says. One scenario he can see where backupand recovery software would cause a system to fail is where a tape is filledand the software is not sophisticated enough to move to another tape. Widensays that most well-regarded backup vendors do not have that problem.

Microsoftsays the snapshot functionality will not replace third-party software. ISVswill be relied on to provide a “writer,” which will sit on each server and sendthe snapshot to the backup and recovery console. Microsoft will provide thesnapshot API to each of the backup and recovery vendors.

“[Microsoft]wants to be able to embed more and more functionality in the operating system,”Widen says. Widen compares the addition of snapshot functionality to theearlier addition of protocol stacks in Windows. At one time, Windows PCs neededthird-party protocol stacks for networking. Microsoft eventually added them toWindows, effectively killing off competition. “It also eliminates the number ofcompanies that may potentially compete,” Widen says.

SteveDuplessie, senior analyst at the Enterprise Storage Group, disagrees. Hebelieves it’s a good idea for Microsoft to provide native snapshot ability.“Backup is a universal problem, and NT is no exception,” he says. Vendors oftenface challenges in developing agents for backup and recovery, and Duplessiesays integrating snapshot into the operating system will aid these vendors.“They would much rather have Microsoft pave the way for them,” he explains.

Snapshotsprovide a point in time view of a machine, to simplify and accelerate backupand recovery. The snapshot routine briefly freezes the I/O of a machine,flushes the memory, then “takes a picture” of the system’s status at thatmoment. A third-party writer can then begin to back up the system based on thesnapshot. Snapshots allows administrators to backup machines without takingthem offline and can improve the reliability of backups.

Becausemany mission-critical files such as databases and Exchange stores are inconstant flux, software needs to isolate a moment to use for backup. Ifsoftware simply copied a file from one volume to tape, it would change from thestart of the backup to the end of the backup, rendering the backup unreadable.The snapshot allows the software to isolate one moment to backup.

In thepast, administrators took systems offline to prevent the problem. But with therise of the Internet and 24x7 uptimes, it is impossible for administrators totake a system offline for backup. Momentarily freezing the system, taking asnapshot, and performing the backup from a snapshot allows systems to stayonline continuously.

In additionto the snapshot API, Whistler will contain functionality to lubricate recovery.Microsoft will introduce a feature known as Automated Systems Recovery (ASR),for saving and restoring applications. If an administrator, for example, usesASR on Exchange, settings and rules are restored with the application.

ASR alsoenables plug-and-play functionality for failed systems. In the case of a harddisk failure, administrators who used ASR can replace the hard drive and makean exact duplicate of the previous hard drive, once the disk has been replaced.

Index:
Microsoft Corp., Redmond, Wash., www.microsoft.com
IDC, Framingham, Mass., www.idc.com
Enterprise Storage Group, Milford, Mass., www.enterprisestoragegroup.com