Depending on the service offering, there are a few things that should be considered:
Each Proxy/Repo needs to have enough resources available for OS operations and VB365 server operations. Therefore we need to reserve a minimum of 15% host memory not allocated to other processes.
A simple example shows that the JetDB engine itself consumes 50% of the host’s memory. Running 250 databases on it adds another 0.1% on top per instance resulting in 25% memory usage. This leaves us with 25% of the available memory for VB365 operations and the OS.
|Default JET database instance memory consumption on Veeam Backup for Microsoft 365 repository||0.1% of host memory|
|Default JET database engine memory cache||50% of host memory|
|Recommended maximum number of JET databases per backup proxy||250|
|Max number of JET databases per backup proxy||750|
Source: VB365 BP Guide
|OS||Windows Server 2019|
|Disk 1||50 GB||OS + 300 MB for production installation + additional free space for the configuration database and logs.|
|Disk N||Backups/Cache||Size depends on the backup repository type that will be used.|
Proxy servers scale horizontally (up to 50 per VB365 Server). Each proxy adds capacity to process up to 20.000 objects or -+ 5000 users. (More details.)
Backup/Cache disk sizing:
- Using JetDB-backed repositories
- This repo type consists out of multiple folders named after the number of years of retention. For each year’s folder there is a repository file (.adb) plus a transaction and check log (.jrs and .chk). The maximum size per .adb file is 64 TB. To counter that limit an automatic rule triggers the generation of a new repo file in the same storage location when reaching 59 TB. For example, three-year retention creates three folders with one or multiple .adb files in it which each can grow up to ~60TB.
- A proper sizing must be done as disks/volumes need to be large enough to hold the entire backup data’s retention.
- Using Object storage repositories
- JetDBs are only being used as cache to hold meta data. Therefore the available space should be only around 1-2% of the source data.
- More (cache) databases fit on the same proxy and much less space is required as all backup data is stored in object storage.