Disk bandwidth measures the rate of data transfer to and from physical drives (HDD/SSD) in the system. It provides separate metrics for read and write operations, reflecting the throughput achieved by the disk subsystem.
| Metric e.g. GByte/s | Description |
|---|---|
| read | Rate of data read from the disk |
| write | Rate of data written to the disk |
Disk bandwidth is an important metric for analyzing the performance of I/O-intensive workloads. High bandwidth usage indicates significant data transfer activity, which may approach the limits of the disk subsystem's capacity. Monitoring disk bandwidth can help identify potential I/O bottlenecks and optimize storage performance.
Merged requests represent the percentage of I/O requests that were combined into a single operation by the disk subsystem. This optimization reduces the number of I/O operations by combining multiple adjacent or related requests into a single request.
| Metric [%] | Description |
|---|---|
| all | Percentage of all merged I/O requests |
| read | Percentage of merged read requests |
| write | Percentage of merged write requests |
| discard | Percentage of merged discard requests (e.g., for TRIM) |
Merged requests improve disk performance by reducing overhead, particularly for workloads with sequential or contiguous data access patterns. High values indicate efficient handling of I/O requests, while low values may suggest scattered or random I/O patterns that prevent merging.
The requests metric measures the number of I/O operations issued to the disk per second. It provides a breakdown of read, write, discard, and flush operations, offering a detailed view of disk activity.
| Metric [requests/s] | Description |
|---|---|
| read | Number of read requests issued per second |
| write | Number of write requests issued per second |
| discard | Number of discard (e.g., TRIM) requests issued per second |
| flush | Number of flush requests issued per second |
The frequency of I/O requests reflects the workload's demand on the disk subsystem. Monitoring these metrics helps identify patterns in disk usage and potential performance issues, such as excessive flush or discard operations, which may impact overall I/O efficiency.
The request size metric represents the average size of I/O requests issued to the disk subsystem. It provides insights into the nature of disk access patterns for all operations.
| Metric [e.g. KByte] | Description |
|---|---|
| all | Average size of all I/O requests |
| read | Average size of read requests |
| write | Average size of write requests |
| discard | Average size of discard (e.g., TRIM) requests |
Average request size reflects the workload's access pattern. Large request sizes indicate sequential or bulk I/O operations, while smaller sizes suggest random or fragmented access. Understanding request size helps optimize storage performance and detect inefficiencies in data handling.
Disk utilization measures the percentage of time the disk is actively processing I/O requests. This metric provides an indication of how heavily the disk is being used.
| Metric [%] | Description |
|---|---|
| utilization | Percentage of time the disk is busy handling I/O operations |
A high utilization value (close to 100%) suggests that the disk is saturated, potentially leading to I/O bottlenecks and increased latency. Monitoring disk utilization is essential for identifying performance issues and optimizing I/O workloads to prevent overloading the storage subsystem.
Wait time measures the average time in milliseconds that I/O requests spend waiting to be served by the disk. This includes both queueing time and service time.
| Metric [ms] | Description |
|---|---|
| all | Average wait time for all I/O requests |
| read | Average wait time for read requests |
| write | Average wait time for write requests |
High wait times can indicate I/O bottlenecks, caused by excessive queuing or slow disk response times.