Understanding Concurrent Disk Access Across Multiple Computers

In today’s complex computing environments, the concept of writing data to a shared disk from multiple concurrent threads is a common scenario. This article explores the implications of concurrently writing to the same disk from two different computers, providing insights into how operating systems, disk controllers, and application programming manage this task.

Concurrency and Multithreading

The statement that 'writing in two parallel threads is impossible' is a misconception. Modern systems handle multiple threads of execution quite efficiently through multithreading. In a multi-core or multi-processor environment, threads can run in parallel, performing simultaneous I/O operations such as writing to the same device. Similarly, in a single-core/single-processor environment, the operating system schedules and manages threads, ensuring that if multiple threads attempt to access the disk simultaneously, they are executed in a controlled and sequential manner.

Scenario: Concurrent Disk Writing from Two Computers

Imagine a scenario where two applications are running on two different computers, each attempting to write data to the same disk. For simplicity, let's consider these applications as free-standing embedded systems without any operating systems underneath. These applications are connected to the same disk drive, which means both computers can have read/write access to the disk. There is a communication protocol for initiating read and write operations, and this protocol is implemented both in the applications and in the firmware of the disk drive controller.

Disk Drive Communication and Command Queuing

The disk drive controller, which is responsible for managing read and write operations, typically queues commands. Commands from both computers are received and then processed serially. This means, despite the applications writing simultaneously, the disk drive can only handle one request at a time, ensuring that data is written in an orderly manner.

Handling Concurrent Commands

Commands issued from the two applications might overlap in time. Depending on the communication protocol, collisions might occur, leading to retries or sequential processing. For example, if commands from two applications arrive at the same moment, the disk controller might use a round-robin scheduling mechanism to ensure fair processing of both requests.

Application Perspective

From the application's perspective, it may seem as if writes are happening simultaneously. However, the disk drive controller processes these commands sequentially. This means that while the applications might be writing at the same time, the actual disk operations are handled in a serial fashion. This can result in either wait states or asynchronous notifications upon completion, depending on how the application is programmed.

Multi-Platter Disk Drives

It's important to note that there are multi-platter disk drives capable of simultaneous writes to different platters through independently controlled read/write heads. However, such specialized hardware is expensive and not commonly found in typical computing environments. Therefore, for the purposes of this discussion, we assume a standard disk drive without this advanced capability.

Final Considerations

When two applications running on separate computers attempt to write to the same disk, there are several factors that can affect the outcome. These include:

The communication protocol used for driving the disk operations The speed of the communication lines connecting the computers to the disk The speed of the disk controller and the rotational speed of the disk itself The size of the written data blocks

As for the applications themselves, their behavior can vary based on whether they are designed to wait for I/O completion or to receive asynchronous notifications. In the case of waiting for I/O completion, the applications will indeed wait until the operation is complete. However, if the applications are designed to handle asynchronous notifications, they will not necessarily wait and can be notified of the operation's completion in a timely manner.

Understanding these principles is crucial for developers working with distributed systems and shared resources. Proper management of concurrent disk access can significantly enhance the performance and reliability of applications running in complex environments.