Reference A3 Architectures
Problem: You are deploying the A3 at your site, and want to know the best design.
Solution: Choose one of these A3 templates, with escalating levels of storage isolation.
A3 Installation Target: Xen or Hyper-V
Alike Data Store: NAS via NFS
Procedure: use the A3 XVA download to deploy to your production Xen host, or the A3 VHD to deploy to your production Hyper-V host. Use a NAS and connect to it via NFS.
Discussion: Deploying your A3 on your production environment is appropriate for many small and midsize organizations. Because your NAS is physically separate, if there is a problem on your production environment, you can redeploy the A3 on a new host, and point it to your existing NAS to recover. Connecting your A3 to Windows storage is not recommended.
A3 Installation Target: Xen
Alike Data Store: NAS via NFS
Procedure: Build a dedicated standalone Xen hypervisor for your A3 VM. Use a NAS for your ADS and connect the A3 to the NAS using NFS.
Discussion: This environment will host your A3 backup server, cold standby replica VMs of critical systems, and provide an emergency environment, if needed. By deploying your A3 on Xen, you can leverage the A3's instant restore at any time. Furthermore, since this A3 is physically isolated from your production environment, it will be easier to understand, instrument, and eliminate any performance bottleneck, should one occur. This physical isolation also means if anything happens to your production environment, your A3 environment is unaffected.
A3 Installation Target: Custom Docker Environment
Alike Data Store: iSCSI or other high-performance configuration
Procedure: Build a baremetal Docker host whose only function is to run your A3 as a Docker image, then customize the A3 "YAML" for your storage needs.
Discussion: While conceptually very similar to "Isolated A3" above, this approach does not supply any emergency environment for recovery. (That environment must be provisioned separately.) Instead, it is a narrower, more focused hardware configuration with fewer moving pieces. By skipping the hypervisor layer completely, this deployment is much more streamlined and thus has fewer points of exposure to fault. With the proper hardware and storage configuration, this is a highly resilient deployment that only does one thing, and does it fast.
Which design to choose has a lot to do with your budget, expertise, and needs, not your size. Yes, larger environments benefit from increased storage isolation, but the prime reason to isolate your backup solution is not performance, but safety and reliability. A good solution puts you in a strong position to recover, even if many other things have gone wrong.
- Never use hypervisor disk for your Alike Data Store. Should your hypervisor become damaged, you will lose your backup data.
- Never use another VM as backup storage. If the VM or hypervisor below it has an issue, your backup data will be difficult or impossible to recover.
- Avoid using Windows as your backup storage backend. The A3 is a Linux appliance, and works best with other Linux-friendly storage.
- Prefer NFS over CIFS. CIFS protocol is fine for small Alike deployments. But since most NAS supports NFS, prefer it, as it scales better and is more reliable.