Exadata Storage Architecture Overview
Exadata Storage Architecture Overview
Grid Disks in Exadata are subdivisions of Cell Disks, which in turn are created from LUNs based on physical disks. Grid Disks are the storage units that ASM Disk Groups use within the Exadata infrastructure. They provide a flexible way to allocate storage space, allowing administrators to partition Cell Disks according to storage requirements. Grid Disks facilitate efficient storage resource management and are critical for optimizing the use and performance of the Exadata architecture .
The Automatic Diagnostic Repository (ADR) plays a significant role in supporting the operational efficiency of Exadata's storage software by capturing and reporting diagnostic information. It maintains an alert log similar to an Oracle database, providing a record of notable events that aid in troubleshooting and maintenance. ADR's inclusion ensures that any operational issues in the storage software are promptly identified, which allows for quick rectification and helps maintain optimal system performance and stability .
Smart Scans are a feature of Exadata where the cellsrv (storage server process) uses the metadata provided with a data request to process the data directly within the storage tier. This processing can include filtering rows and columns, performing certain calculations, and returning only relevant data to the database layer, significantly reducing the volume of data that needs to be transmitted. Smart Scans leverage the high-performance capabilities of InfiniBand and iDB to improve query performance by offloading work from the database servers to the storage servers, which reduces CPU load and speeds up data retrieval .
In a normal Oracle database setup, disks are presented directly to ASM, whereas in Exadata, the configuration is more layered. Physical disks in Exadata are connected to RAID controllers and mapped to Logical Unit Numbers (LUNs), which represent the underlying disk to the upper software layers. Each physical disk maps one-to-one with a LUN. Cell Disks are then created from these LUNs, and Grid Disks are subdivided from Cell Disks for ASM presentation. This layered architecture in Exadata allows for greater flexibility and efficiency in storage management compared to a traditional setup .
Cellsrv is the master process in Exadata's storage server architecture responsible for executing multiple critical functions, including handling database requests for disk I/O and providing advanced SQL offloading capabilities. It manages I/O Resource Management (IORM) and facilitates Smart Scans by utilizing metadata to process data directly on the storage servers. These capabilities enhance the performance of Exadata by minimizing unnecessary data transfer and optimizing resource allocation .
ASM in Exadata serves as a key component by providing a mapping between files and objects that the database recognizes on the storage layer. Although ASM is not positioned between the storage and database directly, it provides a unified interface for managing storage and reduces complexity in file management. It handles redundancy, striping, and rebalancing of disks, thus abstracting the underlying physical storage details. This integration is essential in ensuring efficient storage management and performance optimization in Exadata environments .
The Exadata software architecture facilitates low-latency data requests by using the iDB protocol built over the InfiniBand network. This setup employs RDS (Reliable Datagram Sockets) for fast, low-latency communication that avoids traditional operating system calls. The iDB protocol integrates with the Oracle kernel through LIBCELL, a library that leverages network-based calls for direct interaction with the storage tier. This enables efficient data processing and retrieval, reducing latency compared to conventional methods where each data request requires multiple system calls and read/write operations .
The Restart Server (RS) in Exadata is responsible for ensuring the continued functioning of Exadata software and services. It does this by monitoring the heartbeat of the Management Server (MS) and cellsrv processes. If these processes fail to respond within the acceptable heartbeat interval, the RS will automatically restart them to ensure system reliability and uptime. This functionality is critical in maintaining the stability and resilience of the Exadata system, reducing downtime, and ensuring consistent performance .
The Management Server (MS) in Exadata facilitates access to cell management commands and runs background monitoring threads. It serves as the primary interface for administering, managing, and querying the status of Exadata cells via the CellCLI (Cell Command Line Interface). The MS interacts closely with the Restart Server (RS) to ensure that cellsrv and MS processes are running smoothly by monitoring their heartbeat and restarting them if necessary. This monitoring and management capability is crucial for maintaining the reliability and uptime of the Exadata environment .
The InfiniBand network in Exadata provides a low-latency, high-throughput switched fabric communications link that connects the database and storage layers. It facilitates communication by using the iDB protocol, which is implemented over InfiniBand using RDS (Reliable Datagram Sockets) for low-latency, process-to-process communication. The use of iDB with InfiniBand allows the Oracle kernel to make network-based calls to the storage tier, bypassing traditional operating system reads and writes, which enhances performance. Additionally, RDMA (remote direct memory access) can be set up over RDS in certain cases to further reduce latency in data transfer .