Shulou(Shulou.com)06/03 Report--

Cluster Set, storage migration service, and System Insights are the three new technologies that Windows Server 2019 Lao Wang is most interested in. System Insights can help administrators realize the resource utilization forecast of Windows Server 2019 CPU/ hard disk / volume / network without the need to purchase other components, and automatically trigger remedial actions when predicting resource utilization will reach a specific state at a certain time.

The whole System Insights solution needs to be completed by Windows Server 2019+Windows Admin Center. Install System Insights function on Server, and use Windows Admin Center to analyze and predict the installed server in the web interface.

Installation prerequisites

Windows Server 2019 17692 and above

Windows Admin Center version 1804 and above

System Insights is currently only valid for Server 2019 operating systems with installed features

At present, it is only for stand-alone analysis and prediction, and the future version will achieve analysis and prediction for the cluster as a whole.

System Insights collects and stores data locally for up to a year, and if you want to retain data when upgrading the operating system to the latest version, be sure to use an in-place upgrade

System Insights can run on any Windows Server 2019 operating system, it can run on physical and virtual machines, any hypervisor, and any cloud

Start the System Insights function

Powershell enabled

Install-WindowsFeature-Name System-Insights-IncludeManagementTools-Restart

Server Manager

Installing the System Insights function also installs the powershell management module. Administrators can not only use the WAC interface to manage, but also directly use powershell to perform related management operations on Server 2019.

To install the System Insights extension for WAC, click Settings-manage extensions

Select Windows Server System Insights among the available extensions, and click install. This interface requires the WAC server to be able to connect to the network to see the list of available extensions. The extension list comes from a third party or Microsoft. The extension content includes interface display, collection rules, dashboards, etc. Currently, you can only install the extension online on the server, or offline in the future.

Click on the installation prompt to install the extension to improve the gateway mode, click to confirm automatic promotion, and begin to install the extension.

If the installation is successful, you can see that Windows Server System Insights has been installed in the installed extension

Open the WAC server management interface and click the machine name link. The Chinese version can see the system insight tool on the left toolbar, and the English version is System Insights.

Analysis and prediction of turning on or off a resource by Powershell

Enable-InsightsCapability-Name "CPU capacity forecasting"

Disable-InsightsCapability-Name "CPU capacity forecasting"

At present, the system insight tool only opens the analysis and prediction of four kinds of resources.

CPU capacity prediction

Network capacity prediction

Prediction of total storage consumption

Volume consumption prediction

Each feature analyzes daily data to predict future usage, but CPU, network, or storage usage may change frequently during the day, dynamically adjusting to the workload on the computer. Because usage is not constant throughout the day, it is important to correctly represent daily usage in a single data point, and when evaluating the filtering logic below, it is important to note that each feature attempts to inform administrators that future usage will meaningfully exceed available capacity-even if CPU utilization reaches 100% temporarily, CPU utilization may not lead to meaningful performance degradation or resource contention. For CPU and the network, it should be consistently high usage rather than an instant peak. However, average CPU and network usage throughout the day can lose important usage information, as a few hours of high CPU or network usage can have a significant impact on the performance of critical workloads. An average of up to two hours a day can avoid these extremes and still generate meaningful data for each analytical capability. "however, for volume and total storage usage, the storage usage cannot exceed the available capacity, even temporarily, so the daily maximum usage is used for these functions."

Once the system opinion has been installed, by default, each load is retrieved at 3: 00 a. M. every day. The administrator can also modify the collection plan according to the actual needs, and click Forecast Resource function name-Settings.

After modifying according to the demand, save it.

Powershell operation

Set-InsightsCapabilitySchedule-Name "CPU capacity forecasting"-Daily-DaysInterval 2-At 4:00PM

Set-InsightsCapabilitySchedule-Name "Networking capacity forecasting"-Daily-DaysOfWeek Saturday, Sunday-At 2:30AM

Set-InsightsCapabilitySchedule-Name "Total storage consumption forecasting"-Hourly-HoursInterval 2-DaysOfWeek Friday

Set-InsightsCapabilitySchedule-Name "Volume consumption forecasting"-Minute-MinutesInterval 30

In addition to the planned collection, we can also immediately trigger the data retrieval function, click on the prediction resource function name bar in the system opinion interface, and click call.

Powershell operation

Invoke-InsightsCapability-Name "CPU capacity forecasting"

There are three pieces of work to be done with a systematic view.

1. Retrieve data according to plan rules

two。 Prediction and analysis of resource load

3. Implement remedial actions based on resource tipping points

When we start to retrieve data through manual calls, or wait for the scheduled time to arrive, the prediction result will not come out immediately. It will take at least six days on the server before the system opinion can predict the data. The longer the data, the longer the length of the analysis and prediction. It is recommended that the length of the forecast should be more than three weeks.

The forecasting function requires the use of forecasting models to predict future usage, and for each forecast, the model will be trained locally based on your machine data. This model is designed to help detect long-term trends and retrain each Windows Server instance to adapt to the specific behavior and nuances used by each computer.

Determining the type of model to use requires testing many models with a dataset of tens of thousands of machines. After analyzing and adjusting these models, Microsoft decided to use the autoregressive prediction model because it produces highly accurate and intuitive predictions and does not require much time for training. However, the model requires three weeks of training data, so each ability uses a basic linear trend until three weeks of data are available.

Through observation, Lao Wang believes that this prediction and analysis function can not be completed only locally, because the installation process does not generate a database, and no analysis tools are installed, and it is still a 40-megabyte installation package, so Lao Wang speculates that if you want to achieve such a prediction and analysis effect, you still need to connect the WAC server to the network, and you need to rely on the Azure machine learning function to complete the training presentation of the prediction model.

When the server is placed for a period of time and can be provided to the prediction model to complete the analysis, click on the name of a resource function, and you can see the analysis data provided by the system. On the left side of the forecast view is the data used every day by the resource. The gray area on the right side of the graph is the system insight function. According to the input date of the resource analysis data, you can predict the use of the resource for a period of time in the future. Administrators can refer to the forecast results for capacity planning

The system opinion supports obtaining the history of data collection directly through Powershell and exporting it to JSON format. WAC also automatically displays the JSON file through the extension program.

# Specify the History parameter to see the last 30 prediction results.

Get-InsightsCapabilityResult-Name "CPU capacity forecasting"-History

# Use the Output field to locate and then show the results of "CPU capacity forecasting."

# Specify the encoding as UTF8, so that Get-Content correctly parses non-English characters.

$Output = Get-Content (Get-InsightsCapabilityResult-Name "CPU capacity forecasting") .Output-Encoding UTF8 | ConvertFrom-Json

$Output.ForecastingResults

OK, how is the data collected? what does Predictive Analysis mean? I know, the last remedy, an ambitious function, so, when will it be remedied?

The prediction analysis function periodically shows us the forecast status of the current resource under the forecast view, which can be divided into the following five categories

Normal: it is not expected to exceed the available capacity.

Warning: the forecast exceeds the available capacity for the next 30 days.

Serious: the forecast exceeds the available capacity for the next 7 days.

Error: this feature encountered an unexpected error.

None: there is not enough data to predict. This may be due to a lack of data or no recent reported data.

Through WAC, we can set what to do when the prediction status of a resource reaches normal, warning, serious, error, or none. Currently, only Powershell script remediation is supported. Click on the prediction resource function name-Settings-Actions, you can enter a script for each state, and automatically execute when the prediction resource reaches this state, or you can plan to set different levels of remediation scripts for different states. For example, it is predicted that disk capacity may be insufficient, disk cleanup is performed at warning level, and storage migration is performed at severity level.

Powershell direct management

$Cred = Get-Credential

Set-InsightsCapabilityAction-Name "CPU capacity forecasting"-Type Warning-Action "C:\ Users\ Public\ WarningScript.ps1"-ActionCredential $Cred

Set-InsightsCapabilityAction-Name "CPU capacity forecasting"-Type Critical-Action "C:\ Users\ Public\ CriticalScript.ps1"-ActionCredential $Cred

Remove-InsightsCapabilityAction-Name "CPU capacity forecasting"-Type Warning

Get-InsightsCapability | Get-InsightsCapabilityAction

Remediation script referenc

In this article, Lao Wang uses Server version 2019 17733 WAC version 1804. If you need resources, you can ask Lao Wang for resources. You can see that the newly introduced system insight function in Microsoft 2019 still has certain applicable scenarios. It turns out that this kind of predictive analysis needs to be integrated with SCOM+SCVMM or SCOM+Veeam to see the effect of predictive analysis. Now there is no need to buy System Center products, only the license of Windows Server 2019 can be completed. Help IT infrastructure management to a higher level, automatically provide hints for IT through intelligent analysis, and automatically perform part of the operation and maintenance, which is also the development trend of IT operation and maintenance in the future. I'm glad to see Microsoft launch this function.

For the system opinion of Lao Wang's suggestion, there are still too few types that can be analyzed and predicted. If you can analyze the load of the front end of the website, or the status of a service, and automatically remedy it, the effect will be better. Currently, remedies are still limited. It would be better to support more remedies, such as interfacing with SCO,SMA. When the prediction reaches a certain status, a runbook is automatically triggered to execute. At present, Lao Wang sees that prediction analysis can only be effective on stand-alone servers. It would be better if it can be designed to analyze the entire cluster or a resource pool in the future. At present, it is mainly to predict the utilization rate of analysis resources, and it would be better if we can cooperate with BPA to make suggestions on the configuration of the server.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.