Flylib.com

Books Software

 
 
 

Tivoli Storage Manager

[Previous] [Next]

Tivoli Storage Manager

Tivoli Storage Manager is the new name for Tivoli Adstar Distributed Storage Manager (ADSM). Version 3.1 is an enterprise backup/restore software solution originally from IBM. Tivoli Storage Manager is usually configured to perform one initial full backup and incremental backups thereafter, although you can run full backups any time.

Tivoli Storage Manager can run multiple backup processes concurrently; while individual performance might be slow, overall performance is usually acceptable. The slow individual performance is a consequence of Tivoli Storage Manager's unique "one full, incremental forever" architecture. As Tivoli Storage Manager attempts to back up each file, it must first verify that the file meets the policy that the administrator has specified. Having to check each file before backup degrades the performance somewhat. Tivoli Storage Manager masks this by performing only incremental backups (by default) that require the movement of much less data across the network, and that, theoretically, result in a shorter backup time. This characteristic is evident when performing backups of larger files, as Tivoli Storage Manager can move large files at reasonable speeds.

Another performance indicator is how well the software restores large amounts of data. Because of the architecture, when recovering multiple files, Tivoli Storage Manager restore times are substandard as compared to other products. Before you implement a Tivoli Storage Manager environment, you should investigate other features that are relevant to its performance, such as colocation and reclamation.

Server Recommendations

An important factor when configuring a Tivoli Storage Manager backup/restore server is the inclusion of sufficient disk space for the Tivoli Storage Manager database and recovery logs. With more versions of a file to be kept and longer retention times, Tivoli Service Manager requires a significant amount of overall disk space. The minimum server specifications should consist of 256 MB of RAM, 200 MHz processor(s), and sufficient disk space for the database and recovery log files.

TIP
For improved reliability, allocate database and recovery log volumes in an NTFS file system, not a FAT file system. By using NTFS, you can take advantage of the Windows 2000 capability to recover from problems that can occur during I/O to a disk.

[Previous] [Next]

Configuring a Backup Server

When configuring a Windows 2000 backup/restore server, be sure to configure physical hardware according to the software and the demands that are to be placed on the server (that is, the number of tape drives ). Wherever you can, configure as much physical memory as possible, which allows products such as ARCServe, BackupExec, NetBackup, NetWorker, and REELbackup to use and further adjust memory as required.

Also keep in mind that performance can vary according to the type of data that is being written or read from tape. For example, a file that is 100 MB can, in most cases, be written to tape faster than 100 files of 1 MB each.

The Windows 2000 default configuration is also limited to a maximum block size of 64 K when writing data to tape. However, this can be increased to 128 K blocks using the variable block size technique. Higher block sizes typically aren't allowed because of host bus adapter (HBA) restrictions.

[Previous] [Next]

The Price of Backup and Restore Solutions

Dedicated telecommunication lines—one for voice, one for video, and one for data—are no longer required. Telecommunications transport methodologies and packetized multiplexing methods of data transfer support a variety of signals on the same line. Recent technological advances in data concentrators and multiplexers allow users with different types of data distributed throughout their enterprises , created at different locations, to maximize their investment by sharing common telecommunications lines for various functions, sources, and types of data.

Research indicates that on average, costs for T1, T3, and Synchronous Optical Network (SONET) lines are coming down at a rate of 30 percent annually in the U.S. In 1992, line charges were about $80,000 per month for T3 services installed. Today, that equivalent line charge is about $1,000 a month. Table E-1 compares the transfer times for 1 gigabyte of data across different channels.

Use of private networks, dark fiber, and SONET are enabling increased communications bandwidth. All these advances are driving increased competition and lower-cost transmission and communication lines.

Table E-1. Transfer times by topology

Topology Time to Transfer 1 G (in hours) Time to Transfer 1 G (in minutes)
56 K Link * 41.7711 2506.27
T1 Link * 1.5189 91.14
T3 Link * 0.0520 3.12
10BaseT 0.3086 18.52
100BaseT 0.0309 1.85
1000BaseFX 0.0033 0.20
FDDI 0.0234 1.40
Token Ring 4mb 0.5848 35.09
Token Ring 16mb 0.1462 8.77
ATM OC-3 0.0221 1.32
ATM OC-12 0.0055 0.33

Hardware compression has reached the high end of computing. T3 data compression, impossible until recently, is now enabling twice the effective data rate from 44 Mb to over 88 Mb across a single dedicated line.

Internet- and intranet-based remote data storage and access open a new frontier for business continuity. Secure, offsite, and high-availability data solutions are technologically ripe for the picking.