{"id":3782,"date":"2026-10-08T10:33:49","date_gmt":"2026-10-08T10:33:49","guid":{"rendered":"https:\/\/remote-support.space\/wordpress\/?p=3782"},"modified":"2026-10-08T10:55:55","modified_gmt":"2026-10-08T10:55:55","slug":"backup-for-client-machines-selection-project","status":"publish","type":"post","link":"http:\/\/remote-support.space\/wordpress\/2026\/10\/08\/backup-for-client-machines-selection-project\/","title":{"rendered":"Backup for client machines selection project"},"content":{"rendered":"

We used duplicati because it was multiplatform and it worked up to about 100 to 200 GB max per backup. Then it started having issues. For sizes less than about 50 GB it might work ok.<\/strong><\/p>\n

We also used restic and made a backup system based on that. However it requires double the space on the client machine and is currently planned to be upgraded as a product.<\/strong><\/p>\n

Other systems are available and we shall be trying them out. If anyone is interested in participating in this project, they are welcome. Please see the project collabo<\/p>\n

For reliable multi-terabyte system backups, the open-source landscape offers several powerful options, but the “best” choice depends on your specific environment, existing skills, and whether you prefer a traditional enterprise architecture or a modern, cloud-native approach.<\/p>\n

Based on extensive use in large-scale environments, the leading open-source solutions for this task are Bareos<\/b>, Bacula<\/b>, and Restic<\/b>. Each has a proven track record, but they excel in different scenarios.<\/p>\n

\ud83c\udfc6 Top Contenders for Large-Scale Reliability<\/h3>\n

Bareos (The Enterprise-Grade Fork)<\/h4>\n

Bareos is a robust, enterprise-focused fork of Bacula, designed for growing IT environments. It is an excellent choice for organizations that need a centralized, highly scalable backup system with strong commercial backing.<\/p>\n

    \n
  • Proven Scale<\/b>: Bareos explicitly states it handles “enormous amounts of data” and offers “proven scalability into the exabyte range<\/b>“. It is used to protect a leading EU supercomputer with over 1PB of data<\/b> and backs up a network of 17,000 servers<\/b> for a French government initiative.<\/li>\n
  • Architecture<\/b>: It uses a classic client-server model with a Director, Storage Daemon, and File Daemon, providing centralized control and supporting disk, tape, and S3-compatible object storage.<\/li>\n
  • Reliability Features<\/b>: It supports parallel backup jobs, data compression, and deduplication-friendly storage formats for systems like ZFS. It is certified for SUSE Linux Enterprise Server and Red Hat Enterprise Linux.<\/li>\n<\/ul>\n

    Bacula (The Battle-Tested Veteran)<\/h4>\n

    As the original project from which Bareos was forked, Bacula (specifically the Community version) remains a powerful and highly reliable option. It is known for its robustness and is a staple in large, heterogeneous enterprise environments.<\/p>\n

      \n
    • Massive Scalability<\/b>: The architecture is designed to keep pace with enterprise data growth, with the capability to back up over 6 billion files and thousands of servers<\/b>.<\/li>\n
    • High Reliability<\/b>: Bacula is characterized by high performance and reliability, with a modern multi-threaded design that can run multiple simultaneous backups. It is common for it to run for months without problems.<\/li>\n
    • Disaster Recovery<\/b>: A key strength is that backup data is self-contained. Even if the central Director or Catalog is lost, data can be restored directly from the underlying media, which is critical for disaster recovery.<\/li>\n<\/ul>\n

      Restic (The Modern, Cloud-Native Choice)<\/h4>\n

      Restic is a different breed of backup tool. It is a single, fast, and efficient binary that is designed to back up directly to a wide variety of cloud and remote storage backends. It is an outstanding choice if your strategy is cloud-first or if you want a simple, scriptable tool for system backups.<\/p>\n

        \n
      • Proven at Extreme Scale<\/b>: Restic is the foundation for a tool called cback<\/span> at CERN<\/b>, which is used to back up about 2.5 billion files and 18PB of data<\/b>. It is also successfully used for clients with ~50 TB<\/b> of data.<\/li>\n
      • Efficiency<\/b>: It uses content-defined chunking and deduplication, which is highly efficient for large datasets and incremental changes. Its native S3 compatibility is a major advantage for cloud storage.<\/li>\n
      • Performance<\/b>: While Borg is often faster for local backups, Restic is generally the preferred choice for cloud\/remote targets (S3, Backblaze B2) due to its native integration and reduced bandwidth usage. Benchmarks show local backup speeds of 120-160 MB\/s<\/b> for Restic.<\/li>\n<\/ul>\n

        \ud83d\udd0d Detailed Comparison at a Glance<\/h3>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
        \n

        Feature<\/p>\n<\/th>\n

        		\n

        Bareos<\/p>\n<\/th>\n

        		\n

        Bacula<\/p>\n<\/th>\n

        		\n

        Restic<\/p>\n<\/th>\n<\/tr>\n<\/thead>\n

        Architecture<\/b><\/td>\n\n

        Client-Server (Director, SD, FD)<\/p>\n<\/td>\n

        		\n

        Client-Server (Director, SD, FD)<\/p>\n<\/td>\n

        		\n

        Single Binary, No Server<\/p>\n<\/td>\n<\/tr>\n

        Primary Use Case<\/b><\/td>\n\n

        Enterprise, large-scale, mixed environments<\/p>\n<\/td>\n

        		\n

        Enterprise, large-scale, traditional IT<\/p>\n<\/td>\n

        		\n

        Cloud-native, S3, modern infrastructure<\/p>\n<\/td>\n<\/tr>\n

        Scalability<\/b><\/td>\n\n

        Exabyte-scale<\/p>\n<\/td>\n

        		\n

        Billions of files, thousands of servers<\/p>\n<\/td>\n

        		\n

        18PB+ proven at CERN<\/p>\n<\/td>\n<\/tr>\n

        Storage Targets<\/b><\/td>\n\n

        Disk, Tape, S3 Object Storage<\/p>\n<\/td>\n

        		\n

        Disk, Tape, Cloud<\/p>\n<\/td>\n

        		Any S3-compatible, SFTP, REST, Local<\/b><\/td>\n<\/tr>\n
        Deduplication<\/b><\/td>\n\n

        Via storage system (ZFS, etc.)<\/p>\n<\/td>\n

        		\n

        Via storage system<\/p>\n<\/td>\n

        		Built-in, repository-wide<\/b><\/td>\n<\/tr>\n
        Tape Library<\/b><\/td>\n\n

        Yes<\/p>\n<\/td>\n

        		\n

        Yes<\/p>\n<\/td>\n

        		\n

        No (not its design goal)<\/p>\n<\/td>\n<\/tr>\n

        Typical Performance<\/b><\/td>\n\n

        High (parallel jobs)<\/p>\n<\/td>\n

        		\n

        High (multi-threaded)<\/p>\n<\/td>\n

        		\n

        120-160 MB\/s local, very fast to cloud<\/p>\n<\/td>\n<\/tr>\n

        Complexity<\/b><\/td>\n\n

        High (enterprise-grade)<\/p>\n<\/td>\n

        		\n

        High (enterprise-grade)<\/p>\n<\/td>\n

        		Low to Moderate<\/b> (single binary)<\/td>\n<\/tr>\n
        Commercial Support<\/b><\/td>\n\n

        Yes (Bareos GmbH)<\/p>\n<\/td>\n

        		\n

        Yes (Bacula Systems)<\/p>\n<\/td>\n

        		\n

        Yes (Zmanda, via Zmanda Pro)<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

         <\/p>\n

        \ud83d\udca1 Other Tools to Consider (and Their Trade-offs)<\/h3>\n
          \n
        • Kopia<\/b>: A modern, fast, and reliable tool similar to Restic, with strong deduplication and encryption. It is reported to handle 98 TB of deduped data<\/b> in a single repository and is generally faster and more reliable than Duplicati for datasets over a few gigabytes. It is an excellent alternative to Restic, especially for self-hosted scenarios.<\/li>\n
        • Amanda<\/b>: A long-standing, reliable open-source option, especially for tape-based backups of networked systems. It is designed to back up multiple machines to a single large-capacity tape drive. However, it can be harder to set up and is more geared towards tape, making it less flexible for modern cloud and disk-based strategies.<\/li>\n
        • UrBackup<\/b>: While popular for its ease of use, it has significant limitations for true multi-terabyte system<\/i> backups. Image backups are limited by the VHD format to 2TB<\/b>, and users report issues restoring volumes above this size unless using VHDX and specific configurations. It is not recommended for this use case.<\/li>\n
        • Duplicati<\/b>: This tool is generally not recommended<\/b> for full system backups of multi-terabyte datasets. It is not designed for it and users report severe inefficiencies, including extremely slow restores and database operations that can run for many hours or fail on large jobs.<\/li>\n<\/ul>\n

          \ud83d\udc8e Summary: How to Choose<\/h3>\n

          Your choice depends on your infrastructure and team’s expertise:<\/p>\n

            \n
          • Choose Bareos or Bacula<\/b> if you need a traditional, centralized enterprise backup system<\/b>. They are the right choice for managing thousands of servers, supporting tape libraries, and providing a single pane of glass for your entire backup infrastructure. Bareos is the more modern, actively developed fork with strong commercial backing.<\/li>\n
          • Choose Restic (or Kopia)<\/b> if you favor a modern, cloud-first, and scriptable approach<\/b>. If your target is S3-compatible object storage or you want to avoid managing a complex server infrastructure, a single-binary tool like Restic is far simpler and has proven itself at CERN-scale.<\/li>\n<\/ul>\n

            If you can share more about your environment (e.g., number of servers, OS mix, and preferred storage target), I can offer a more specific recommendation.<\/p>\n

             <\/p>\n

            <\/div>\n

            <\/path><\/svg><\/i> \"Loading\"<\/p>\n

            <\/div>\n","protected":false},"excerpt":{"rendered":"

            We used duplicati because it was multiplatform and it worked up to about 100 to 200 GB max per backup. Then it started having issues. For sizes less than about 50 GB it might work ok. We also used restic and made a backup system based on that. However it requires double the space on […]<\/p>\n

            <\/div>\n

            <\/path><\/svg><\/i> \"Loading\"<\/p>\n

            <\/div>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_wp_convertkit_post_meta":{"form":"-1","landing_page":"0","tag":"0","restrict_content":"0"},"footnotes":""},"categories":[72],"tags":[],"class_list":["post-3782","post","type-post","status-publish","format-standard","hentry","category-backup"],"a3_pvc":{"activated":true,"total_views":8,"today_views":0},"_links":{"self":[{"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/posts\/3782","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/comments?post=3782"}],"version-history":[{"count":4,"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/posts\/3782\/revisions"}],"predecessor-version":[{"id":3789,"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/posts\/3782\/revisions\/3789"}],"wp:attachment":[{"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/media?parent=3782"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/categories?post=3782"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/remote-support.space\/wordpress\/wp-json\/wp\/v2\/tags?post=3782"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}