Subscribe to Windows IT Pro
May 09, 2000 11:58 AM

Enterprise Backup Software

Windows IT Pro
InstantDoc ID #8725
Rating: (2)
Consider all the factors when choosing backup software

Data that you store on enterprise networks represents money, and for most organizations, irrecoverable loss of such data would be a financial catastrophe. Thus, choosing the correct backup software is like picking the right vehicle to take your company's receipts to the bank: You want the armored car, not the Yugo. Unfortunately, differences between backup software applications are not as apparent as this analogy suggests. The five enterprise-level backup applications I tested—Computer Associates' (CA's) ARCserveIT Advanced Edition 6.61, Syncsort's Backup Express 2.1, VERITAS Software's NetBackup 3.2, Legato Systems' NetWorker 5.51, and Tivoli Systems' Storage Manager 3.7.1—can deliver network data to storage media. But differences appear in the applications' performance, ease of use, scalability, manageability, and cost.

Defining the Enterprise
For software testing purposes, I defined an enterprise network as a network that has more than 300GB of data to back up within one time window. An enterprise network also has multiple network OSs (although the main backup software runs on a Windows NT 4.0 server) and at least one active mail server. Using this definition, I assembled a test network on which various servers ran NT (one of the NT servers ran Microsoft Exchange Server 5.5), Novell NetWare 5.1 and NetWare 4.11, and Sun Microsystems' Solaris 7. I generated and dispersed several diverse data sets that totaled approximately 350GB in more than 1 million files.

The Test Network
In my test network, the server that hosted the backup software was a Data General AViiON 3704 with four 550MHz Pentium III Xeon processors, 2MB of Level 2 cache, and 4GB of RAM. Disk storage consisted of thirty 18GB 10,000rpm disks in an EMC CLARiiON fibre channel disk array. I configured these disks in five RAID 5 arrays of 66.1GB each. I used one array to store performance-monitor logs and other dynamic data that was not part of the data sets I backed up. For network connectivity, the server hosted two Adaptec ANA-6944B four-port 10/100 Ethernet 32-bit NICs. I used NSI Software's Balance Suite 2.71 for Windows NT to aggregate the eight ports into one IP address for transmit-and-receive load balancing. The tape library's SCSI connectivity relied on dual Adaptec AHA-2944UW Ultra HVD 32-bit adapters.

The robotic tape library was an Advanced Digital Information Corporation (ADIC) Scalar 1000 with four DLT 7000 tape drives and 158 tape slots (for more information about the Scalar 1000, see "Scalar 1000," page 152). I divided the tape drives evenly between the two SCSI adapters in the host server. I loaded the tape library with 29 DLTs and two cleaning cartridges. Between tests, I cleaned the drives.

Service Pack 5 (SP5) ran on all NT servers in the test. The NT servers' drives used NTFS and had less than 2 percent fragmentation.

Three network clients ran on Dell Precision WorkStation 410 systems; each system had dual 550MHz processors, 128MB of RAM, and on-board Adaptec 2940 U2W SCSI host adapters. Each workstation came with a 4GB SCSI disk, and I added three Seagate Cheetah 9GB 10,000rpm hard disks to each system to increase storage capacity. I configured the first system as an NT file server and configured the Cheetahs and system disk as four logical drives. I filled the system with nearly 29GB of data. I set up the second system with NetWare 5.1, configured each physical drive as a separate NetWare volume, and filled the system with 22GB of data. I used the third system for my Exchange server. I used RAID 0 to stripe the three 9GB disks and created a logical drive to hold the 16GB private Information Store (IS) and directory database. I added a 9GB disk to the third system for the 6.5GB public IS.

To create another NT file server, I added a Hewlett-Packard (HP) NetServer E45 that had a 266MHz Pentium II processor, 128MB of RAM, and an Adaptec AHA-2910C SCSI host adapter with three HP 4.2GB 7200rpm disks. I used one disk for the OS and configured the other disks as a RAID 0 stripe set to create an 8GB logical drive. I filled the logical drive almost to capacity with random data.

For the NetWare 4.11 server (Support Pack 8a), I used a Digital PC 3000 with a 300MHz Pentium II processor and 128MB of RAM. In this system, I installed an Adaptec AAA-131U2 PCI RAID controller and attached a Cheetah 9GB 10,000rpm hard disk. On this drive, I created a NetWare volume and loaded it with 7.2GB of data. The Solaris 7 server was a Sun Enterprise 250 server with dual 400MHz 64-bit UltraSPARC processors, 2GB of RAM, and six 9GB 10,000rpm hard disks configured as a RAID 5 volume. I loaded the Solaris 7 server with data totaling 31.5GB.

For the network, I used a 16-port autosensing full-duplex 10/100 switch with 1MB buffers to link the servers to eight aggregated ports on the Data General backup server. All servers communicated at 100Mbps full duplex. Figure 1 illustrates the test network.

Results
Price will be a big factor in your enterprise backup selection. Pricing for enterprise backup products is complex and environment-dependent. Table 1 shows vendor-supplied prices for licensing products on various system configurations. These numbers are for comparison only; contact vendors for detailed price quotes for your environment.

Each enterprise environment has unique complexities and requirements, and a product that is well suited to someone else's environment might not work as well on yours. When you evaluate a product, you need to consider scalability, support, performance, features, ease of use, and cost. Scalability means that a product can grow as your enterprise grows without burdening your IT staff with complex reconfigurations or the need to change backup products altogether. Good vendor support is usually worth much more than any cost savings you might realize at purchase. Differences in the quality of support you receive from vendors can define your downtime, which directly affects your bottom line and your peace of mind.

No product emerged as a strong favorite for enterprise environments. Taking into consideration all the test factors, including scalability and support, I give the overall edge to Backup Express for its excellent performance and midrange pricing. I place NetBackup second for its stability and scalability.

The Test
The products I tested supported a range of autoloaders and advanced features (e.g., the ability to span media and support parallel backup data streams from one client). Each product also offered Storage Area Network (SAN) support. Table 2 compares the products' important enterprise backup features.

I gave the vendors details of the test network and the test procedure and asked the vendors to recommend ways to optimize their products' performance. I tested each product's ability to simultaneously back up 349GB of data on the entire server farm. This test roughly simulated how a 9:00 a.m.-to-5:00 p.m. business would attempt to back up a large volatile data set after hours. To measure performance, I timed the backup job and recorded processor utilization. I then restored 151GB of backed-up data, timed the restore operation, and checked data integrity. Figures 2 and 3, page 138, compare the products' backup and restore performance, respectively.

ARCserveIT Advanced Edition 6.61
ARCserveIT, originally a NetWare and UNIX backup product, supports client agents for Windows 9x, Windows 3.x, NetWare, OS/2, Macintosh, and various UNIX platforms. The product has advanced autoloader support and supports several database agents. CA aims ARCserveIT Advanced Edition 6.61 at companies that use NT as a platform for backup servers. CA designed ARCserveIT to work as a component within the company's Unicenter TNG, a larger enterprise-management product, but ARCserveIT is fully functional as a standalone product. Although ARCserveIT supports features such as centralized backup-database administration, ARCserveIT is a two-tier server and client product and won't scale as readily as some of the other products I tested. ARCserveIT requires NT 3.51 or later, 32MB of RAM, and 40MB to 50MB of disk storage. The product came on a CD-ROM, and a startup guide accompanied it. The CD-ROM included more documentation in Portable Document Format (PDF) files.

I launched the install wizard from the CD-ROM and chose a complete installation. ARCserveIT gave me the option of installing Microsoft SQL Server rather than ARCserveIT's proprietary database, Raima. A CA technical support engineer said the Raima database holds 16 million records (although a patch enables it to contain unlimited records), and installing a SQL database lets you expand beyond 16 million records. By default, the Raima database installed on my test system. After choosing the rest of the defaults, I rebooted the server and ran the device configuration wizard. ARCserveIT identifies SCSI tape drives at the device level and automatically sets configuration parameters. The device configuration wizard recognized the Scalar 1000 and the Scalar 1000's barcode reader. Then, because I had chosen the autoconfigure option, the wizard configured the drives and robot. ARCserveIT let me designate a clean tape slot and set the cleaning interval (clean-tape configuration wasn't this easy in the other products I tested). Finally, ARCserveIT verified the media in the tape library.

ARCserveIT includes a client push agent and uses push technology to improve data-transfer performance between server and clients. After the client receives a backup request, it pushes data across the network in large-packet bursts. The ARCserveIT host server breaks up the large packets and formats them for transfer to the designated storage device. These processes, which improve performance, occur simultaneously between the client and server.

Installing the push agent was easy and quick. In the initial ARCserveIT install wizard, I chose the option to install client agents remotely. The install wizard launched its remote setup executable file (rsetup.exe) and displayed all the Windows clients on my network. I chose the clients on which I wanted to install the agent, provided a domain account and password under which the ARCserveIT agent would run, specified the target directory to install the client files, then clicked OK. The installation process completed in a few minutes. After the initial setup completed, CA notified me that I needed to install SP1 to ARCserveIT. The update process was as simple as the installation was.

ARCserveIT's Exchange Server backup agent uses the Exchange Server's backup and restore API functions to back up an online database. The backup agent can also use Messaging API (MAPI) functions to perform online mailbox-level backup and restore. Although mailbox-level backup and restore is appealing, it doesn't provide disaster recovery or replace full Exchange database backups. The Exchange Server backup agent installed quickly and easily from the server program CD-ROM. On the server, the Exchange Server client appeared with the appropriate Exchange database options in the network client directory tree, which Screen 1 shows.

ARCserveIT's UNIX client installation was fast and uneventful. I mounted the ARCserveIT CD-ROM, then used the pkgadd command to transfer the appropriate Solaris client files to the server. I launched uagentsetup from the server's client directory and accepted the default setup options. Finally, I launched the agent with the uagent start command. On the server, the UNIX client appeared under the client tree hierarchy.

Related Content:

ARTICLE TOOLS

Comments
  • Anonymous User
    7 years ago
    Sep 08, 2005

    DISK CLEAN-UP
    Windows 98: Start, Programs, Accessories, System Tools, Disk Clean-up
    OR My Computer, Right-Click on drive letter, Properties, General, Disk Clean-up
    Windows XP: Start, All Programs, Accessories, System Tools, Disk Clean-up OR
    My Computer, Right-Click on drive letter, Properties, Tools, Disk Clean-up
    SCANDISK
    Windows 98: Start, Programs, Accessories, System Tools, ScanDisk OR My Computer, Right-Click on drive letter, Properties, Tools,
    Error-Checking Status
    Windows XP: My Computer, Right-Click on drive letter, Properties, Tools, Error-Checking
    STANDARD: Checks for file system errors
    THOROUGH: Also checks for bad sectors (damaged areas on disk)
    DEFRAGMENTING
    Windows 98: Start, Programs, Accessories, System Tools, Disk Defragmenter OR My Computer, Right-Click on drive letter, Proper
    ties, Tools, Defragmentation Status
    Windows XP: My Computer, Right-Click on drive letter, Properties, Tools, Disk Defragmenter OR My Computer, Right-Click on drive letter,
    Properties, Tools, Defragmentation

    File Recovery Software
    Recupero di Dati
    Datenrettungs Software

  • Anonymous User
    8 years ago
    Nov 22, 2004

    Forgive the long and possibly wandering comments.

    Syncsort was our 'one and only' software for backing up our systems, since about 1999.
    The product worked as advertised, and allowed us to backup our Unix and Windows clients from one system.
    Then, about 1-2 years ago, the problems started.
    We received an upgrade, and gladly installed it, only to find out that during the upgrade our 'catalog' was destroyed. Apparantly the process consumed all disk space (Needed to have 110% of existing catalog size free to install, and we did not, but no warning was given), and failed to convert the catalog correctly. We also didn't find this out, until after the next catalog save completed, which erased the last catalog save. Although we should have had more than 1 tape for the 'catalog', we assumed that the catalog was also being backed up with the normal saves. We were informed from tech support, that the 'data' directory is automatically exlcuded from the normal system backups, even if you explicitly select it.
    So, we were forced to 'start over', with no way to reload the catalogs from the old tapes, as most other systems are able to do.
    Recently having upgraded to AIX 5.3, we decided to upgrade to BackEx latest 2.2 version, with the web interface.
    The upgrade went smooth with no signs of problems.
    We succesfully restored a few files, and were happy, until we attempted to do a backup.
    The system reported hardware problems on the 'SCSI' interface.
    Syncsort support blamed the hardware. We disagreed, since all was working until that upgrade, but gladly contacted IBM and exabyte.
    IBM replaced the SCSI cards, and the cables. Exabyte tested the drives, and found no problems. Syncsort still insisted there was a hardware problem.
    We purchased another SCSI card, and installed the library onto that controller, but BackEx still reported hardware problems, on every SCSI interface that had a tape drive attached (3 scsi cards now). It was apparant to us, that the problem was related to syncsort, and possibly it running on AIX 5.3
    IBM's tools reported no errors, and was able to backup to every tape drive in the system
    An engineer from Syncsort attempted a few fixes, they even sent a tech on site, who simply reviwed the cabling and installation...
    Well, after 30+ days, we still do not have a backup, and have decided to switch to another vendor, and funny enough, there are no SCSI bus errors, and we're using the same hardware / drivers as was used with syncsort's backup express.


    Goodbye Syncsort.

  • Todd Edwards
    10 years ago
    Oct 30, 2002

    Our company is in the process of reviewing enterprise backup software. I read Tom Iwanski's "Enterprise Backup Software" (June 2000, InstantDoc ID 8725) and wondered whether the author has done any recent reviews.

  • Tom Iwanski
    10 years ago
    Oct 30, 2002

    For a more recent look at backup products, see Ed Roth's "Enterprise Backup Solutions" (October 2001, InstantDoc ID 22239). I suggest that you thoroughly test any products in your own environment, if possible. Some vendors provide a timed trial version of their software that's viable for comparative testing. Defining what features are most important for your environment and which product addresses them the best is crucial. For example, performance might be critical for your organization because you have a narrow backup window. Ease of use is also important, especially if you'll have moderately skilled technicians performing backups and restores. The list is unique to each environment.

  • Brandon McFeron
    10 years ago
    Mar 18, 2002

    I read this article, and being a TSM administrator, I was suprised to see Netbackup held in such high regard. After reading the article I came across the following post on www.adsm.org (a TSM support website) and thought it to be VERY Pertinent to the information presented here:

    -----Original Message-----
    From: Seay, Paul [mailto:seay_pd@NAPTHEON.COM]
    Sent: Monday, January 21, 2002 12:28 AM
    To:
    Subject: Re: VM TSM migration options: Veritas vs Netbackup


    There has been a lot of recent discussion on the list about the subject area
    of Veritas on Intel versus TSM. The comments here are for everyone, not the
    author of the question, nothing is personally meant by any comments here.
    Bottomline, NetBackup doesn't scale at all. We are ripping it out of the
    Windows environment right now. We worked for 18 months with Veritas
    Engineering to try to fix the product. They simply gave up. The word
    "compete" should not even be put in the same sentence when speaking of
    Netbackup vs TSM on Intel. If you have less than 20 clients to backup and
    none with over 50GB of data, Netbackup will be OK. That is if you never
    need to create duplicate offsite copies or need a deleted file policy.
    Duplication on Windows is an impossibility in the Netbackup world unless you
    buy 4 times the hardware you have in comparison to TSM and a 24 to 72 hour
    window to create those duplicates. Deleted file policy, Netbackup, asks
    what is a policy? No such animal, so you get stuck when you do not catch
    that a file has been deleted before your tapes expire in Netbackup. The key
    word here is tape expiration, not backup object expiration. NetBackup has
    no such thing as storage management. I refer to it as NetBackup,
    GrossNoRestore. In other words, NetBackup backs up some of your stuff, but
    you will never be able to restore it all.

    Yeah, UNIX is next. After the debacles of implementing 3.4 of Netbackup,
    Veritas really dug the grave deep. Oh, I forgot to mention that our Windows
    Netbackup 3.4 migration lead to a down (backups lost) situation for weeks
    and we ended up figuring out what the problems were.

    Because we lost half the performance from 3.2 to 3.4 on Windows, we were
    faced with needing to change. More or better hardware would not fix the
    problem, hell, we are using ESS disk and Magstar FC tape with high-end
    servers. Before the migration we were getting 4.5MB/sec and up to 10 in
    certain situations. Veritas could not figure out how we were getting these
    levels of performance. They could not reproduce them with our own server
    and identical hardware in their labs. Simply, Netbackup cannot scale in the
    Windows environment.

    I consider myself an expert on Netbackup and a knowledgeable person on TSM.
    I believed the Netbackup hype, thought the product was the best because it
    had the features that I thought were needed. When actually, implementing
    you find out the features differences with TSM are gimmicks to get you to
    buy and really never scale making them unusable. These gimmicks cause you
    to overlook the real issue of being able to restore your business, which
    implies having control and the ability to direct what is backed up.
    Netbackup's GUI is impressive, it is the registry hackers dream. Wait till
    all the timeout crap hits the fan and you start tweaking registry entries,
    creating undocumented touch files and finding out there is poor to
    non-existent Windows support at Veritas for Netbackup when you have a
    critical problems. When you are paying 23% maintenance from a large account
    you would think that having half a dozen critical down situation open calls
    would get someone from Development engaged to work with your account. We
    finally surmised these people did not exist anymore.

    Yes, TSM has its quirks and customers have lost data over the years, but
    probably mostly of their own doing and not really learning the TSM product.
    After 911, everyone should be taking backup and recovery at a different
    seriousness. If not, you are in the wrong business. That means if you are
    not an expert in the backup product you are using and doing regular disaster
    recovery tests, then shame on you, get to be an expert. If you are not
    capable, choose a vendor that has support, Tivoli is one of them. The shame
    if it is we automatically set the support expectation bar 2 notches higher
    when it is an IBM company, but we will pay more to a fly-by-night
    organization and make excuses for them when they do not answer the phone.

    This all said make your NetBackup/TSM decision on facts, not likes or
    dislikes. Your business depends on you getting this right and ultimately
    your job and reputation.

    Consider one final note. Your understanding of TSM is an irreplaceable
    asset. You could spend 50K training people alone on Netbackup and still not
    be able to support the product. The cost of TSM is much less than Netbackup
    in the long term. See if you can work a deal with IBM to convert your
    drives to FC or SCSI for a nominal fee. The "staying with TSM" carrot may
    be all that is needed to push the button hard enough to get someone's
    attention. ESCON is relatively slow compared to SCSI and fibre channel.
    FICON is a different story. If you have MVS, that is ultimately the
    cheapest answer to your problem.

You must log on before posting a comment.

Are you a new visitor? Register Here

advertisement

advertisement

Windows is a trademark of the Microsoft group of companies. Windows IT Pro is used by Penton Media Inc. under license from owner.