Coming from the FOSS area, there are few backup solutions which are fit for use in larger environments.
There's an abundance of existing commercial backup software and I'm interested how they measure up when used in the field. Most offerings look quite nice on paper, but they tend to show their ugly side as soon as they're rolled out and used for the first time. Common sources of trouble are important things you forgot in your product evaluation, features which were presented by the marketing people more vividly than they really are or bugs/problems which compromise functionality, data integrity or the product as a whole.
I'd like to know how the backup software you know & use fares in the following categories:
- What platforms are supported on server & client side?
- Is communication between client, server and storage encrypted?
- Does the architecture allow for scalability? Can you easily grow to 100, 500 and 1000 clients? (Split out "controller" and "storage" daemons, no areas with known scalability issues)
- What storage devices are supported?
- Is the configuration revision-safe? (Can you notice accidental/malicious configuration changes easily)
- Does the software have interfaces for monitoring? (e.g. failed backup jobs)
- How about performance data collection? (e.g. used volumes, average job size, days till backup space will be full, etc.)
- Does the product have some sort of data deduplication? (Don't store identical files twice)
- Do incremental backup jobs handle deleted files correctly? (E.g. don't restore files which were deleted between backup runs when restoring from an incremental backup)
- Are there any tools/features which aid in Bare Metal Recovery? (Think Live-CD, automatic setup of block devices & filesystems, restoring of the bootloader, etc.)
- How does job scheduling work? Are there any features which help in balancing backup jobs over a given timeframe?
- Are there plugins/helpers for databases and other daemons which need to be put in a consistent state first? Is there a proper Pre/Post-Backup script execution framework?
- Is job migration supported? (E.g. Move a Backup Job from Disk to Tape or from Tape to an external/offsite archive)
Please mention any kinks, issues, problems, etc. which are affecting you in daily use as well as features which are unique/set it apart from the rest.
I know this is a lengthy question, but I think it'll help the whole community to flesh this out as thorough as possible so that an informed decision can be made without much trial & error ;). Feel free to comment on things I forgot in my list.
-
This is by no means a complete answer as I simply don't have a complete answer, but I can comment on two products...
- Legato
- Symantec NetBackup - (AKA NotBackup, AKA NetStuffup)
Firstly, NetBackup would not be such a disaster if the team responsible for the servers was also the team in charge of the backups, but in the case of an enterprise-sized company, this is often not the case. That is, there is a separate backup team.
The problem here is that NetBackup has assumed this for you - so suddently, the backup team ask you for all sorts of access (root access to be specific).
Another flaw of NetBackup is that - say you want to revive a server from scratch, from backup - you actually can't. You need to install the Operating System, install the netbackup client, and then restore the files.
Legato does not have any of these problems.
Oh - the reports that you get from NetBackup are also inferior to that which you get from Legato. NetBackup reports (which I've seen to date as managed by our backup team) is very useless - to the point that we have asked them to tell us if it went ok or not, and we depend on them (i.e. we don't monitor the backup email ourselves anymore).
This is again due to the fact that they can't generate good reports because we don't give them root access to the severs.
From Xerxes -
After 10+ years and thousands of dollars, I have given up on both tape and commercial solutions. We now backup to disk, and I look for a backup solution to be:
Free/open archive formats (e.g., tar, zip, etc.)
Scriptable -- can involve native commands or utilities, but these must be callable from a common backup script (e.g., exmerge, osql, mysqldump, gnupg, etc.)
Platform independent where possible
Generates log files or status that can be emailed/directed as part of the overall job
While these fail many of the "enterprise" criteria specified above, in my experience the keep-it-simple, standardized approach is more likely to result in your data being both recoverable and secure. Keep in mind there is no one ideal solution for backing up all systems, and developing effective backup strategies for each one involves more than just buying an agent for a commercial program.
From nedm -
Bacula is an Open Source Backup solution. It's architecture is designed after that of major commercial solutions.
Platforms
Bacula supports Windows as well as any POSIX compliant Unix system.
Encrypted communication
Authentication between daemons is always secure, to encrypt the communication between the daemons, a full rollout of PKI (client certificates) must be done.
Architecture & Scalability
Bacula consists of file daemons (clients), one or more directors (Job Schedulers) and one or more storage daemons. Each director can use multiple catalogs (SQL Databases) to store information about jobs.
See the Bacula documentation for a more in depth look at Bacula architecture.
Storage devices
Bacula stores all file contents in volumes which can be stored on local filesystem, on tape drives or even on DVD media. Support for tape autochangers exists.
Configuration safety
All configuration exists in the form of plain text files, which can be easily versioned by a SCM. Additional state of the software is entirely in a SQL database, which can be audited by appropriate tools on the database side.
Statistics Interfaces
No native interfaces for monitoring or performance data exist, although this information can be extracted from the SQL databases without too much troubles. There've been some improvements in this area in Bacula 3.0, which I haven't explored in detail yet.
Data deduplication
Bacula doesn't support any form of data deduplication.
Deleted file handling in incremental backup jobs
As of Bacula 3.0, there's a mode to handle deleted/renamed files correctly.
Disaster recovery
Bacula is working on a Live system which aids in bare metal recovery. See the release notes for further information.
Job scheduling
Job scheduling is very rigid in Bacula. You can define intervals and start times for every backup job, but there's no automatic tuning/distribution of the jobs based on (changing) amounts of data which need to be backed up.
Plugins & Helpers
Bacula 3.0 comes with a new plugin system and native support for Exchange 2003 & 2007. Additionally there's an extensive framework to run custom scripts on client side before/after backup jobs.
Job Migration
Jobs can be easily migrated between different pools.
lukecyca : +1 for Bacula. We use it to back up data from two sites with 2 tape drives, an autochanger, and 2 media types. It has a learning curve, but a flexible and solid architecture which is especially well-suited for integrating multiple locations, media types, and devices. I haven't upgraded to Bacula 3.0 yet, but it promises some pretty slick features such as synthetic full backups, more flexible job scheduling, and better ACL support.From Michael Renner -
I use rdiff-backup. It's basically like rsync, but you can easily access old versions of files/directories and manage how many old versions are kept. It's only useful when backing up to disk, but disks are cheaper than tape these days anyway.
You need to have it installed on both the server and the client, and it runs on OS X, Linux, and Windows. I haven't used it on Windows, but according to the project's mailing list it runs pretty much the same there.
From Brad Beyenhof -
Sometimes we use HP DataProtector. But I prefer to use tar. HP DP also has a nice command line but tar is much more flexible.
From disserman -
I think all of the major commercial backup software has its quirks. I'm actually starting to like HP's dataprotector. I've used legato and netback in the past and hadn't even heard of data protector until my current gig. After using it for a year I'm fairly impressed.
From Jim B
0 comments:
Post a Comment