DBPLUS PERFORMANCE MONITOR: Changes in 2024.2 - Part 3: OS STAT
19/9/2024
Tomasz Chwasewicz
In the high-tech tangle of our time, more cores usually means slower CPU clock speeds. Experienced database administrators return to the old battlefield of the battle for CPU utilization. The latest DBPLUS Performance Monitor update (version 2024.2) includes significant improvements that help in this unequal battle. In particular, thanks to the dedicated page of statistics of the operating system - OS Stat Page.
Accessible through the “Performance -> OS Stat” path or directly in DBPLUS PM OS Stat, the updated OS Stat page in DBPLUS PM offers a comprehensive overview of critical metrics that reflect host health and performance. Here's a snapshot of what to expect here.
As pointed out by the insightful article gruffdba. Ever since we jumped into the flashy world of all-flash storage about a decade ago, processor bottlenecks have been in the spotlight. Currently, they are the main factors affecting performance.
“Since the move to all flash storage solutions, about a decade ago, CPU utilization and bottlenecks have often become a major cause of performance concerns for Oracle DBA.”
This look into the past highlights the need for an OS Stat page in the latest version of our software. Now it is more important than ever before and serves as a dedicated control panel. IT solves today's problems by displaying the key statistics that are necessary to diagnose and solve these pesky performance issues. Below, a quick look at the metrics waiting behind the scenes.
This statistic represents the total number of seconds that CPUs spent executing low-priority user code. Its accumulation is spread across all processors and serves as an indicator of the time spent on less critical processes. Nice's high time may suggest that the system has adequate CPU capacity, allowing it to allocate time for lower-priority tasks without affecting the performance of more critical processes.
The LOAD statistic represents the number of processes. They are either actively running or waiting in the queue to be executed by the CPU. This metric is a direct indicator of the load on the system. An elevated LOAD value can signal a system overload where too many processes compete for CPU time, potentially leading to slower response times and increased CPU queue times.
These metrics measure the volume of paged (loaded from disk to memory) and paged (moved from memory to disk) data due to virtual memory management. High values of these statistics may indicate that the system is experiencing memory pressure, which leads to excessive swapping when the operating system tries to free up physical memory. This swap can seriously degrade performance, as disk I/O operations are much slower than memory access.
This statistic measures the amount of swap space currently available in the system. The exchange space is used as an overflow when the physical memory is fully utilized. In systems that rely on swapping during peak processing periods, monitoring this metric is critical. A low or decreasing value of this metric may precede performance problems, as it indicates that the system is running out of available memory resources.
Too many statistics? No problem. If you feel overwhelmed by the sheer number of columns, there's nothing to stress about — you can always hide the ones you don't need. Just right-click on the column header and select the “hide” option. With this feature, the dashboard is elegant and focused, presenting only the data that is most important for current analysis.
The OS Stat page includes powerful visualization tools that display these metrics in real time. These tools are designed to help administrators quickly detect spikes or unusual patterns in data, such as sudden increases in CPU load or memory swaps. By identifying these anomalies, administrators can pinpoint potential problems before they develop into serious problems, facilitating faster intervention and resolution.
This visualization capability, highlighted by features such as OSSTAT 2 - busy spike graph, ensures that DBAs can maintain optimal performance and stability on their systems, making the OS Stat page an essential resource in the toolkit of any database administrator managing modern, dynamic database environments.
The latest version introduces several new statistics from the v$osstat view, which are collected every 15 minutes and presented in both graphical and tabular form:
