According to the Oracle Exadata Database Machine documentation12, two recommended configuration best practices for backup and recovery on Exadata are:
A) Placing the backup network on dedicated switches installed in the top of the rack (ToR) has the benefits of isolating the backup network from other workloads and providing a greater level of control.
D) When off-site long-term backup retention is needed, use Oracle Database Backup Cloud Service as a low-cost, offsite scalable storage for a disaster recovery solution.

Exadata-specific Enterprise Manager Plug-Ins are used to monitor different components of the Exadata Database Machine.
1. ASM instances: The Automatic Storage Management (ASM) instances can be monitored directly through the use of Exadata-specific Enterprise Manager Plug-Ins. This allows you to monitor the storage space, disk usage, and performance of the ASM instances.
2. Hybrid Columnar Compression (HCC) ratios on Extended (XT) storage servers: The HCC ratios on Extended (XT) storage servers can be monitored directly through the use of Exadata-specific Enterprise Manager Plug-Ins. This allows you to monitor the compression ratios and performance of the storage servers.
3. Oracle clusterware on the database server: The Oracle clusterware on the database server can be monitored directly through the use of Exadata-specific Enterprise Manager Plug-Ins. This allows you to monitor the health and performance of the clusterware.

When applying updates on Exadata systems, database server updates are automatically rolled back to a working state on a previous release if the update fails. This ensures that the system can be restored to a working state in the event of a failed update. Additionally, storage server updates are also automatically rolled back to a previous release in the event of a failed update.
This is documented in the Oracle Exadata Database Machine X9M Implementation Essentials official text book and more information can be found on the Oracle support website [1][2].

Exadata System Software metrics and alerts work as follows:
Persistence of Thresholds: When metric thresholds are configured, they remain active and persist across server reboots. This ensures consistent monitoring behavior even after maintenance or unexpected reboots .
Default Metrics: By default, metrics are collected without predefined thresholds. Administrators must set custom thresholds to enable specific alerting based on their operational requirements.