How does Exadata integrate with Oracle ASM at hardware level?

How does Exadata integrate with Oracle ASM at hardware level?

While Oracle Automatic Storage Management (ASM) can run on almost any hardware, its integration with Exadata is unique. In a standard environment, ASM is just a volume manager; in Exadata, ASM and the storage hardware are "interlock-aware."

They don't just coexist—they communicate via a specialized protocol called iDB (Intelligent Database Protocol) to bypass traditional storage limitations. Here is how that integration works at the hardware and architectural level.


1. From Physical LUNs to Grid Disks

In a typical SAN, ASM sees a LUN and treats it as a "black box." In Exadata, the hardware is carved into a specific hierarchy before ASM even touches it:

  • Physical Disk/Flash: The raw NVMe or SAS hardware.

  • Cell Disk: A software layer over the physical disk that handles low-level formatting and error tracking.

  • Grid Disk: These are the actual "slices" of the Cell Disk presented to ASM.

Because ASM sees Grid Disks, Exadata can present the "fast" outer tracks of a spinning disk as a high-performance Diskgroup and the "slow" inner tracks as an archival Diskgroup.


2. Partner-Aware Placement (The Anti-Affinity Rule)

ASM is responsible for mirroring data (Normal or High Redundancy). On standard hardware, ASM might accidentally put a primary copy and its mirror on the same physical storage tray.

On Exadata, ASM is "Exadata Aware." It automatically understands the physical boundaries of the storage cells. It ensures that a Primary Allocation Unit (AU) and its Mirror AU are never placed on the same storage cell. This hardware-level awareness ensures that an entire storage server can catch fire, and the ASM Diskgroup will remain online.


3. The RoCE/InfiniBand Fabric Integration

The hardware interconnect (RoCE or InfiniBand) allows ASM to use Remote Direct Memory Access (RDMA).

When ASM needs to rebalance data (move an AU from Cell A to Cell B), it doesn't use the Database Server’s CPU as a middleman. The storage cells use Direct-to-Wire communication to move ASM chunks between themselves. This hardware-offloaded rebalancing means that adding a new storage cell doesn't slow down your SQL queries.


4. Failing Disk "Fencing"

One of the most powerful hardware-level integrations is how ASM handles a "sick" disk.

  • Traditional: A disk might become slow, causing ASM to hang while waiting for a timeout.

  • Exadata: The Storage Server software monitors hardware latency at the controller level. If a disk is performing outside of its hardware spec, the storage cell tells ASM to "Drop Force" the disk.

ASM immediately stops trying to read from that hardware and switches to the mirrored copy on a different cell. This is called Predictive Failure Analysis, and it prevents a single hardware component from dragging down the entire cluster's performance.


5. Intelligent Data Placement (IDP)

On High Capacity (HC) cells, ASM works with the hardware to perform Intelligent Data Placement.

  • The most frequently accessed ASM metadata is placed on the fastest part of the physical disks.

  • The Exadata Smart Flash Cache acts as a hardware accelerator for the ASM Diskgroups, automatically caching the most active AUs.


Summary of Integration

FeatureStandard ASMExadata + ASM
VisibilitySees LUNs as "Black Boxes."Sees physical cell boundaries and disk tracks.
CommunicationStandard SCSI/Fibre Channel.iDB Protocol over RDMA (Kernel Bypass).
MirroringLogic-based.Hardware/Cell-aware (Partner Awareness).
Disk FailureReacts after the timeout.Proactive "Fencing" based on hardware latency.

The Result

This integration is why Exadata can offer "True Scale-Out." Because ASM and the hardware are tightly coupled, you can add or remove storage cells on the fly, and the system will automatically rebalance itself at the hardware level with zero downtime

Looking for servers Rental ?

Call Our Expert :


  • (call for rental enquiries)

Email us :