| This Issue This article is a reprint of chapter 15, by Nelson Howell , in a new book called 'Using Microsoft Internet Information Server 4' from Que Education & Training (ISBN 0789712636) due for publication in early March 1998. The chapter covers performance tuning of the Internet Information Server version 4.0. Including: What is performance, building web sites for speed, and the performance monitor. Performance Tuning Picture the impossible. You, the Web virtuoso, have created the cr me de la cr me of Web sites. Hailed throughout the Webbed world, your site is inundated with well wishers and rubberneckers who queasily peruse your crowning accomplishment. The site is achieving a level of traffic that has never before been seen in the world. Your Web site is widely considered the defining moment in mankind's reach for understanding. The president wants to be your friend, and you're writing the pope's messages. All is well. Suddenly, your server creaks and groans, grinding to a halt with sparks flying everywhere. That one last visitor, little Timmy (with the frail leg) from Chile has put you over the edge and your server just can't handle the load. What will you do? Every Web site, large and small, faces issues of performance. Much like a Formula One racer, Web sites and servers are always in the never-ending battle for optimum speed and balance. If something in the race and environment changes, the racer's performance can suddenly decline unless the pit crew is ready to leap into action. Consider yourself the pit crew to your Web site and server. If you don't keep on top of your server's well-being, you can kiss your speedy site goodbye. Performance basics
When someone says you should tune your server's performance, what does that mean? There are fundamental concepts you need to understand before you begin. You, a server, and a monkey wrench: Tuning your Server
A healthy, stable server is a productive server. Enhancing the performance of your server will provide tremendous gains for your Web visiting users. IIS Tuning 101
A well-designed site and properly configured Web server let you squeeze that extra oomph out of your server's already prime celerity. An Introduction to High Performance Concepts Whether you want to compare your server to a speeding race car or a primed boxer, the underlying concept remains the same: complacency and poor planning can result in horrific results. It is important to constantly be aware of your server's speed and alacrity to make sure that your visitors aren't greeted by the first glacier-speed Web site, or the ever so popular "Server Busy" message. Delivering your site's contents rapidly involves tuning your server, your Web site, and your delivery. The three are intrinsically linked in what I call the Happy Web Circle, as shown in Figure 15.1.FIG. 15.1 The Happy Web Circle depicts the three most important aspects of maintaining good server performance. Click Here For Figure Let's step through the three points in the circle to define the importance of each aspect. Healthy Server, Long Life If you walk through life without any real concern for your health or well being, it is safe to assume that you aren't in top form. Servers are much the same, and rely on constant tweaking and fiddling to make them perform to the best of their processor-given abilities. Server health can be examined by three criteria: Processor Load
Everything your computer does creates a set of tasks for the processor to carry out. Each task, in turn, adds to the load that the processor must bear. A busier computer means a larger load the processor must carry. If the load is too large, the server can become bogged down in trying to keep up with the demand for tasks. A faster processor, or multiple processors, can minimize the load on a particular server. Memory Usage
As important as processor load, the amount of memory that your computer has directly impacts performance. If you think of your computer's memory like a glass, the more things you do adds more water to the glass. If you overload the glass' contents, the glass cannot hold any more and spills over. Each task your processor takes on takes more memory. The more memory your server has available, the more breathing room it has for each task. Resource Consumption
Although memory is considered a resource, it stands in a slightly different grouping. Computing resources, such as storage space and speed, as well as connection speed, have a direct bearing on your machine's performance. Slower storage devices that are filled to the brim deliver information more slowly than lean, fast devices. Fast, uncontested network connections can fire off your content rapidly to your visitors, rather than being mired up in congested network traffic. Each of these three points directly affects a server's performance. It is not good enough to tweak only one point, as the other two can easily bring your server to its knees. Tuning your server is a collective effort of concentrating on all three points to create a healthy balance. Rapid Delivery Building on your high performance server (the actual machine) is a high performance Web service. The Web service operates on top of the general server operations, delivering your Web site and its contents to the masses. Although the act of tuning up your Web server is not as involved as tuning an entire server, there are important points to consider: Limiting Bandwidth
Certain services and sites consume more of your network connection than others, typically because one is more popular than the other. Throttling the bandwidth on particular sites frees up bandwidth for other sites, preventing one site from hogging the entire network pipe. Limiting Connections
The more people you let into your house, the more crowded it gets. Rather than let your house burst at the seems, you can limit the number of people to enter at any one given time. Your Web server can act the same way, much like a doorman, only allowing so many people into your site at a time. With this, you can set your server's timeout period, which is used to determine when a connection is closed. Limiting Services
Different services use varying amounts of memory. The FTP service consumes less memory than the Web service. Carefully considering what services you need lets you eliminate the extraneous overhead on your system. If you fail to properly estimate your traffic, you could end up at either extreme of the scale. On one side, you could be overestimating and wasting precious resources that don't significantly contribute to overall performance. On the other, you could be underestimating and choking your site with too little resources allocated. Web Sites Built for Speed Improving your site's performance can begin with the server, but can ultimately fall on your site itself. Poorly planned Web sites can create the network equivalent to gridlock on an expressway. Sites that do not consider loading times, resource consumption, and the overall constraints of the medium can create a headache for both the site visitors and the host. Speaking from a delivery point of view, most poorly planned sites share at least one (if not all) of these points: No consideration for loading time
The biggest offender by far is the total lack of concern for the size of files. Large graphics, weighty animations, and bloated sound files dominate the Web in a way that seems to say overindulgence and excess are the ways to a better message. Large files have two-pronged effects that are a detriment to performance. First off, they take considerable time for the visitor to download. An enjoyable Web experience for a visitor is not waiting for an immense file to arrive. Secondly, the larger the files, the more resources on the server are consumed. Both storage and bandwidth are at risk when large files are consistently delivered to a visitor. Also keep in mind that the longer a connection is kept open for a visitor to receive a file, the more the processor load and memory usage on the server. Improper choice of medium or format
It's not likely you would use a bicycle to move a bedroom suite. Instead you would choose a more appropriate means of moving your bed to its destination. The Web is the same way. You should carefully consider what type of content you are sending to your users, and choose the most appropriate means of delivering it safely. For example, if you are delivering a lot of large files to a user, FTP may be slightly slower to connect than a Web service, but it also uses considerably less memory. Failure to consider the resource implications
When working with server side activation, such as ISAPI extensions or CGI programs, a large load can be put on the processor, as well as a significant drain on the memory pool of the server. Server-side programming is often the biggest offender for server slowdown. Careful planning and implementation of your server-side code is crucial to maintaining a healthy server. Your server programs can easily bring the server to its knees if you are not careful. When first building your site, carefully consider the consequences of your actions. You want to make sure that both your server and your site are swift to respond to the user's needs. Tuning Your Server The first stop on your hunt for better performance is the server, but where to begin? Establishing just where you need to improve your server can be a difficult task. Thankfully, Windows NT includes several tools intended for administrators to tune their servers. Without these tools the job could prove difficult, but with your small toolkit in hand you can get started. Performance Monitor: Your Window to Your Server's Health The first tool you should examine is the Windows NT Performance Monitor. Included with both Windows NT Server and NT Workstation, the Performance Monitor is a gauge to your server's overall activity and resources. The monitor, shown in Figure 15.2, is fully customizable. By adding counters, you can track almost every aspect of your server with visual graphs. Performance Monitor also includes facilities for reporting and alerting based on ongoing activity. Like most NT administration tools, the Performance Monitor also lets you connect to other machines on your network to evaluate their performance.FIG. 15.2 The Windows NT Performance Monitor is a graphical tool for measuring your server's health and ongoing activity. Click Here For Figure Because the Performance Monitor is so customizable, it can take some time to decide what criteria you want to monitor. A wealth of information, in the form of objects, is available for you to monitor. Most NT-based services add their own object to the Monitor, and each object has its own variety of counters. You can customize each counter with a unique color, scale, style, and line width for the graph. Thankfully, each counter also includes a description, letting you decide easily and quickly what to choose. Each chart that you create can be saved to disk, letting you quickly load different settings and counters for observation. The Performance Monitor is an easy tool to start using. There is no need to specifically install the Monitor, it is automatically installed with other administrative tools when you first install Windows NT. To begin, locate the Performance Monitor icon in your Start Menu. Select the icon to launch the program. When first launched, the Performance Monitor opens a blank chart, as shown in Figure 15.3. By default, the Performance Monitor is grouped with other Windows NT administrative tools in the common Administrative Tools program group. This can vary between machines, especially if you have spent any time customizing your Start Menu. FIG. 15.3
The Performance Monitor window is dominated by the chart region in the center. Click Here For Figure Like most Windows programs, a standard hierarchical menu bar is present along the top of the window. Below the menu bar is the toolbar. The toolbar contains several small icon buttons that act as shortcuts to menu bar functions. More noticeably, the large chart area is visible in the center of the window. The vertical labels are represented on the left side of the chart, with an updated display at the bottom of the chart called the value bar. The display shows the selected counter's information (including the last reading, average reading, minimum and maximum reading, and the graph time). Below the display is a legend, providing the counter color, scale, object, instance, and computer. Making a New Chart To begin using Performance Monitor, you must add counters to your new chart. After the counters have been added to your chart, their progress is updated on the display. To add a counter, follow these steps: - Click the Add Counter toolbar button. This button is represented by a large plus sign icon. You can alternatively choose Edit, Add to Chart. The Add to Chart dialog box appears, as shown in Figure 15.4.
FIG. 15.4 The Add to Chart dialog box contains a list of all objects on your server from which counters can be used. Note the Explain button on the right side to expand the dialog box. Click Here For Figure - Click the Explain button on the right side to expand the dialog box. The Counter Definition box appears at the bottom of the dialog box.
- From the Object drop-down box, select an object you want to monitor. For the sake of an example, choose the Processor object.
- The Counter list box updates with the currently selected object's counters. Select a counter that you want to monitor. In our example, select %Processor Time.
- If there is more than one instance of an object (such as more than one processor), the Instance list box displays the available instances. Select the appropriate instance you want to monitor.
- The Color drop-down box displays the color that will be used to represent this counter in the graph. Select the color you want to use for this counter.
- The Scale drop-down box defines what scale your counter will use, according to the legend. This value ranges from 0.0000001 to 100000.0. Select the scale you want to use, or select Default.
What is scale? The vertical labels on the left of the chart display is a numerical representation that reflects change in the chart. By changing the scale value, you are changing how the counter is represented on the chart. The lower the scale, the smaller the visible change on the chart. You can use scale to fine-tune your counter monitoring. - With a large number of counters to monitor, you might want to make some stand out more than others. Using the Width drop-down box, select the line width you want to use for this counter.
- Instead of line widths, you can instead use line styles to differentiate counters on the chart. Use the Style drop-down box to select a line style for this counter.
- After you have set up your counter, click the Add button to commit your changes and add the counter to the chart. If you have no more counters to add to the chart, click the Done button.
Your new counter now appears in the display area of the chart, with its information updated in real-time. The counters list at the bottom of the Performance Monitor window also lists your new counter, including the instance and scale. Use the legend's counter list to help you quickly identify the counters in your chart. When you select a counter from the counter list, the value bar below the chart changes to reflect the currently selected item. You can also double-click the selected counter to change its properties in the chart. You can freely add and remove counters to change your chart. You can also save your current chart by choosing File, Save Chart Settings. You then can choose File, Open to open your chart at another time to resume your monitoring. The display and behavior of your chart can be modified by choosing Options, Chart. Each aspect of the Performance Monitor has its own set of options, and the Chart options let you choose how the chart looks. The Chart Options dialog box, shown in Figure 15.5, provides you with a few options to alter your chart's appearance and behavior. The five check boxes alter the chart display: FIG. 15.5 The Chart Options dialog box lets you alter your chart's appearance and behavior. Click Here For Figure Legend
The legend acts as a counter list for your chart. Each counter you have added to your chart appears with its attributes in the legend. You can use the legend to select individual counters (which in turn updates the value bar) and change their properties. Value Bar
When you select a counter from the counter list, the counter's values are displayed in the value bar below the chart. If you do not use this feature, you can remove the value bar for additional screen space. If you have chosen to remove the legend from the display, the value bar is automatically removed. Vertical and Horizontal Grid
These two check boxes let you add (or remove) a horizontal or vertical set of lines to compose a grid. These grids can be used to create more distinct visual references for change on your machine's performance. Vertical Labels
The vertical labels on the left of the chart display are used to reference change in your chart. If you find that it obstructs your chart, you can remove the legend from the display. In addition to the five check boxes, there are a few other options available to you: The Gallery
This radio button set lets you change the type of display that is used for your chart. The Graph option displays a familiar graph, plotting each point of change of a counter in the chart. If you prefer, you can instead use the Histogram option to display a vertical bar for each counter. Vertical Maximum
This text box lets you define the upper limit of the vertical label. If you find that your counters rarely hit the upper level limit, lower the value to create a clearer display. Alternatively, if your counters are always peaking, extend the upper limit by entering a higher value. The Update Time
This radio button set defines how your chart is updated to reflect changes in performance. The first option, Periodic Update, lets you define an Interval (in seconds) between automatic updates to your chart. If you prefer to update your chart yourself, you can choose Manual Update instead. Using Alerts Another aspect of Performance Monitor that is incredibly useful is the Alerts facility. Let's say that you want your machine to warn you if your available memory falls dangerously low. You can create an alert inside the Performance Monitor that constantly keeps an eye on your system's condition and executes a specified program in the event that an alert is triggered. This program could be a batch file that sends an alert message to your screen or event log, or perhaps a more drastic program that shuts your machine down until you return to look at it. Alerts are incredibly simple in behavior. They simply await the trigger to be hit, and then launch the program. If no trigger is hit, the alert sits idly by and waits for it to happen. The Performance Monitor must be running on your server for the alert to take place. To switch to Alerts view, click the Alerts View button in the toolbar (the second button, which looks like a pad of paper with an exclamation mark on it), or choose View, Alert . The display changes to Alert view, which differs considerably from the Chart view. The display is dominated by the Alert Log in the center of the screen, with an Alert Interval listed above it, and the Alert Legend listed below. This display is shown in Figure 15.6. FIG. 15.6 The Alert View display is Spartan in comparison to the Chart View. The Alert Log lists every triggered alert that has occurred. Click Here For Figure Adding counters for an alert is very similar to adding counters to a chart. To add an alert counter, follow these steps: - Click the Add Counter button on the toolbar, or choose Edit, Add to Alert. This opens the Add to Alert dialog box, as shown in Figure 15.7.
FIG. 15.7 The Add to Alert dialog box differs slightly from the Add to Chart dialog box. The focus of this dialog box is to define the actions in an alert situation. Click Here For Figure - Just like the Add to Chart dialog box, you should select an object to monitor from the Object drop-down box. For example, choose Memory. If there is more than one Instance of your object, be sure to specify it.
- The Counter list box updates with the currently selected object's counters. Select the counter you want to monitor for an alert. For example, choose Available Bytes.
- From the Color drop-down box, choose the color you want to identify this counter by.
- The Alert If radio button set defines how your alert works. You can choose to trigger the alert if the counter is Over or Under the value you specify in the text box. Specify a value for the trigger and define the range as either Over or Under. For example, choose Under and enter the value of 1000 to warn you if your available bytes falls below 1000.
- The Run Program on Alert text box and radio button set define what happens when the alert is triggered. You can launch a selected program either the First Time the alert is triggered, or Every Time. In the text box, specify the program to launch (complete with the path, if it is not in the environment path). Then choose when the program will be launched. For example, in the text box enter net send administrator alert! and set it to First Time.
- After you have defined your alert, click the Add button to add your alert. When you are done, click the Done button to close the dialog box.
Your new alert appears in the Alert Legend. When an alert is triggered, it is added to the Alert Log so that you can visually track what has happened. Alerts can prove very useful for warning you of dangerous situations, such as low storage space, memory, or processing power.
Just like Chart view, Alert view provides you with options to alter Alert behavior. The Alert Options dialog box, shown in Figure 15.8, provides you with a few options. FIG. 15.8 The Alert Options dialog box lets you change how alerts behave. Click Here For Figure These options are: Switch to Alert View With this check box selected, the Performance Monitor automatically switches to the Alert view in the event that an alert is triggered. This is useful for studying the Alert Log for triggered alerts. Log Event in Application LogThe Windows NT Event Viewer (which is covered in the section"Event Viewer: Your Server's Own News Service" later in this chapter) is used to record events and alerts for your server. You can select this check box to have your alert recorded in the Application log of the Event Viewer. This creates a more permanent record of your triggered alert. This is very useful if you are away from your computer and an alert is triggered. The Network AlertThischeck box and text box let you send a network alert message to a specified user when an alert is triggered. By selecting the check box and entering a network user name, the message dialog box appears on that user's screen if he is logged in. Update Time This radio button set behaves exactly like the radio button set in the Chart Options dialog box. You can alter the frequency that the Performance Monitor updates the counters to monitor for triggers by selecting Periodic Update and specifying an Interval in seconds, or choose to Manually Update the alerts.
Log Files When you want to store the activity of your server and look back on it later, you can choose to use the Performance Monitor's Log view. To switch to Log view, click the View Output Log file status toolbar button (the third icon, represented by a cylinder), or choose View, Log. The display changes to the Log view, as shown in Figure 15.9. This display is dominated by the legend list box at the bottom of the display, with the log file information appearing above it. FIG. 15.9 Outputting your performance monitoring to a log lets you study the results later. Click Here For Figure Instead of logging individual counters, Performance Monitor logs entire object activity. This lets you use the log data at a later time to create charts of individual counters. To create a log file, you must define the objects to be monitored: - Click the Add Counter icon, or choose Edit, Add to Log. The Add to Log dialog box appears, as shown in Figure 15.10.
FIG. 15.10 The Add to Log dialog box is very simplistic, letting you add complete objects to be logged. Click Here For Figure - From the Objects list box, select the object you want to have logged. Remember that each object's counter is included in the logging.
- Click the Add button to add the object. If you are done adding objects, click the Done button to close the dialog box.
The objects you have added appear in the display's legend. Before logging begins, you must set the Log Options. To do so, open the Log Options dialog box (as shown in Figure 15.11 by choosing Options, Log. FIG. 15.11 The Log Options dialog box is where you set the log file's name and location, as well as the frequency of the updates. Click Here For Figure The Log Options dialog box is a specialized file selector, letting you choose the location and name of the log file. To begin your log file, follow these steps: - Using the Explorer interface within the dialog box, locate the directory where you want the log file to be stored, or you can enter the path with the file name.
- Enter the log file's File name, ending with the .log suffix.
- Set the Update Time using the radio button group at the bottom of the dialog box. For automatic updates, choose Periodic Update and enter an Interval (in seconds) in the text box. For manual updates, choose Manual Update.
- To begin logging immediately, click the Start Log button. If you do not want to begin logging yet, click the Save button. (You need to return to this dialog box to start the logging later.)
When the logging begins, the upper display changes to reflect the file name and location, the status (either opened or closed), the interval (in seconds), and the size of the log file. When you are done logging, you must return to the Log Options dialog box and select Stop Logging to close the log. After you have a log file saved to disk, you can use that log within the Performance Monitor whenever you want. To load your log data into the Performance Monitor, follow these simple steps: - Choose Options, Data From. This opens the Data From dialog box, as shown in Figure 15.12.
FIG. 15.12 The Data From dialog box lets you specify what data the Performance Monitor will use. Click Here For Figure - Select Log File and enter the log file path in the text box. Alternatively, use the Browse button beside the text box to select the log file using the Explorer interface.
- Click OK to close the dialog box and load the data.
After you load the data, all information for your charts use the logged information. You must add your counters as before, but the logged information is used rather than the current activity. Note that certain options (such as intervals) might not be available in all dialog boxes. To return to using the current activity in Performance Monitor, follow these steps: - Choose Options, Data From. This opens the Data From dialog box.
- Select Current Activity.
- Click OK to close the dialog box.
Performance Monitor Reports Reports reflect current conditions on your server. Instead of using a chart to display the information, reports use numerical representation. To switch to Reports view, click the View Report Data button on the toolbar (the fourth icon, represented by a pad of paper with lines on it), or choose View, Report. This changes the display to Report view, as shown in Figure 15.13. FIG. 15.13 Report view can be used to get an overall status of your server without using charts. Click Here For Figure Reports display information in columns for each individual instance of an object. You can create a report on all the counters for a given object and then watch them change under various conditions. To add counters to your report, follow these steps: - Click the Add Counter button, or choose Edit, Add to Report. This opens the Add to Report dialog box, as shown in Figure 15.14.
FIG. 15.14 The Add to Report dialog box is in the same vein as the Add Chart dialog box. Click Here For Figure - From the Object drop-down box, select the object you want to monitor. For example, choose Physical Disk. If there is more than one instance of an object (such as more than one physical disk drive), specify the Instance to monitor.
- The Counter list box updates to represent the counters for the current object. Select the counter to add to your report. For example, choose %Disk Time.
- When you are finished selecting your counter, click the Add button. If you are finished adding counters, click the Done button to close the dialog box.
Your new counters are added to the Report view, listed according to object and instance. To modify the update interval for your report, open the Report Options dialog box by choosing Options, Report. The Report Options dialog box, as shown in Figure 15.15, provides you with the ability to change the Update Time. You can have the report updated automatically by choosing Periodic Update, and specifying an Interval (in seconds) in the text box. If you would rather update the report manually, choose Manual Update. FIG. 15.15 The Report Options dialog box is very simple, only letting you change the frequency of the report updates. Click Here For Figure There is much more to Performance Monitor than meets the eye. You should explore Performance Monitor further by experimenting with the tool and referring to the Microsoft documentation. One of the key aspects of Performance Monitor that was not covered here is its capability to monitor multiple computers over a network. If you are monitoring more than one server over a network, you are bound to find these features useful. Event Viewer: Your Server's Own News Service If the Performance Monitor is your window inside your server, the Event Viewer is the narrator. The Windows NT Event Viewer, shown in Figure O.16, is automatically installed when you first set up Windows NT Server and Workstation. It is used to record events and audit situations on your machine that might require your attention. The Event Viewer should be monitored closely, as it often provides the first clues to possible problems. FIG. 15.16 The Event Viewer is an invaluable tool for administrators. You can customize what events will be recorded. Click Here For Figure As defined by the Microsoft Event Viewer online help, an event is "any significant occurrence in the system or in an application that requires users to be notified." Events that do not require immediate attention by the user are recorded to the Event Viewer, whereas critical events are usually recorded to the Event Viewer and brought to the attention of the user with an on-screen message. The Event Viewer splits events into three separate logs: System
System events directly apply to the underlying system functions of Windows NT. This includes device drivers and system components. Any problems of this nature are recorded to the system log. Security
Security events are situations that might indicate possible breaches to the security of your system. You can use the security log to record both successful and unsuccessful security events. Application
For programs that encounter errors, typically while running invisible to the user, the application log is used. Services report status and error messages to the application log for your reference. Within each log, events are displayed by a single line. Each line contains the following: Date and Time
The actual date and time that the event occurred is recorded in the Event Viewer log. Source
The source of the event is the software that reported the condition. This can be either an application or a component of the system, such as a device driver. Category
Events are classified by the source for easier sorting. Sample categories are logons and logoffs for security. Event
To assist in diagnosing events, particular events have event numbers to identify situations. User
If a user account was involved, such as a service running under a specific user, the user's name is provided. Computer
The computer that the event occurred on is listed. Type
Beside each event entry is a colored icon representing a classification for the event. Sample classifications include Error, Warning, Information, Success Audit, and Failure Audit. To move between logs, you can choose the log of your choice from the Log menu. When you choose a different log, the display changes to represent the current log's contents. Examining an Event To look more closely at an event, select it from the log and double-click it. This opens the Event Detail dialog box, as shown in Figure 15.17. FIG. 15.17 The Event Detail dialog box provides you with more specific information on the event, including a data breakdown. Click Here For Figure The Event Detail dialog box provides you with more comprehensive information on the event: The upper third of the dialog box displays the event time and date, user, source, type, category, and computer. This is a mirror of the information displayed in the log. The Description box, in the center of the dialog box, displays a text report of the event. This message varies according to the event, but usually provides specific information on what happened to the source involved. Below the description box is the Data box. This lists a data breakdown of the event for debugging. You can choose to see this in the form of Bytes or Words. Along the bottom of the dialog box is a row of buttons, including a Close button to dismiss the dialog box, a Previous button to move to the preceding event in the log, and a Next button to move to the following event in the log. Moving to the next and previous events depends on how your log is being sorted. From the View menu, you can choose to have Newest events listed first or Oldest events first. Filtering Your Log In addition to choosing what order your events will appear, you can choose to list All events or Filter the events to view only specific messages. To filter your log view, follow these steps: - From the View menu, choose Filter Events. This opens the Filter dialog box, as shown in Figure 15.18.
FIG. 15.18 The Filter dialog box lets you choose what events will appear in your log view. Click Here For Figure - The top two grouping boxes let you specify where your view will begin and end. You can choose to view from the First event to the Last event, or specify particular starting and ending dates and times (Events On for starting, Events On for ending). Specify the starting and ending criteria for your view.
- The center box provides you with five check boxes that represent event types. Select the check boxes for the types of events you want to include in your view.
- The Source drop-down box lets you specify the event source that the view will contain. This lengthy list includes every source on your server. Select the source that you want to view.
- The Category drop-down box lets you select the event category (if applicable) to view. Select the category you want to view.
- If you want to view all events belonging to a particular user, enter the user name into the User text box.
- Because you can monitor events from more than one machine, you might want to specify a particular computer to monitor. If you want to do so, enter the computer name in the Computer text box.
- The Event ID text box lets you specify a particular event number to view. If you want to filter based on an event ID, enter the number into this box.
- If you are satisfied with your filter criteria, click the OK button. Otherwise, you can click the Clear button to reset the criteria or the Cancel button to abort the filter.
Log Settings Logs can become cumbersome and large to sort through. Thankfully, the Event Viewer provides you with a few options for making things easier. The Log Settings dialog box, as shown in Figure 15.19, is opened by choosing Log, Log Settings. FIG. 15.19 The Log Settings dialog box lets you limit the size of a log file and control how events are overwritten. Click Here For Figure - The dialog box lets you control the overall size of each log, as well as how the log manages overwriting old events. To modify a log's settings, follow these steps: 1. From the top drop-down box, Change Settings For, choose the log that you want to modify. You can choose the System, Security, or Application log.
- The Maximum log size text box lets you enter a maximum size for the selected log. The log size must be defined in 64-kilobyte increments. You can optionally use the flywheel arrows on the side of the text box to control the size.
- The Event Log Wrapping radio group controls how old events are overwritten in the log. If you want the Event Viewer to overwrite events as it needs, choose Overwrite Events as Needed. If you prefer to have the events overwritten after so many days, choose Overwrite Events Older than and specify the number of days in the text box. You can optionally use the flywheel arrows on the side of the text box to set the number of days. Finally, if you prefer to manually clear the events, choose Do Not Overwrite Events.
- Repeat steps 1 through 3 for each log. When you are satisfied with your changes, click OK to accept them. If you prefer to return to the defaults, click the Default button.
If you have reduced the maximum size of the logs, your change will not take effect until the log is first cleared. Clearing Your Log When a log is filled with outdated information, or if you have changed the maximum size on the log, you might want to clear the log contents and start fresh. Clearing your log is a simple process. Follow these steps: - From the Log menu, choose the log that you want to clear.
- Return to the Log menu and choose Clear All Events. This opens a confirmation dialog box.
- You are asked if you want to save the contents of this log before clearing its contents. Click Yes to save the log using a standard file selector. To clear the log without saving, click No.
- A final warning and confirmation dialog box appears, warning you that the clearing of a log is irreversible. If you are still sure that you want to clear this log, click Yes; otherwise click No.
Your log file is cleared and will record any new events. You might need to refresh your log to see any new events. You can do so by choosing View, Refresh. Task Manager: Real Time Server Monitoring One other useful tool in your war for server performance is the NT 4 Task Manager. The Task Manager displays all of the ongoing tasks and threads on your machine, as well as their resource usage. You can use the Task Manager to switch tasks, launch new tasks, or end processes. The Task Manager is pictured in Figure 15.20. FIG. 15.20 The Task Manager can be your front line tool for checking up on server performance. Click Here For Figure The Task Manager window is broken into a few divisible areas. Along the top is the obligatory menu bar. Below that is a row of three tabs, which let you switch views (more on that in a second). Below the tabs is the current view, which displays the constantly updated information appropriate to the selected view. Finally, a status bar lies at the bottom of the window. This status bar displays the current number of running processes, percentage of processor usage (CPU), and memory usage (in the form of physical memory followed by kernel memory). The Task Manager has three views: Applications, Processes, and Performance. Applications View If you think of applications as programs running in the foreground of your computer (that is, you can interact with it), the Applications view lets you control these programs. As shown in Figure 15.21, the Applications view lists all of the currently running application tasks and their status. You can select a currently running program and use the End Task button to close the program, or the Switch To button to move focus to that program. Otherwise, you can use the New Task button to launch a new program. FIG. 15.21 Applications view displays all of the currently running "foreground" tasks. Click Here For Figure Processes View When you consider that all activities on your computer are broken up into processes, your machine is filled with a lot of them. Each program has one or more processes running, and most of them are invisible to you. Processes view, shown in Figure 15.22, lets you see what processes are currently running on your machine and what resources they are using. You can select a process from the list and click the End Process button to close the task (if possible). FIG. 15.22 Processes view lists every process currently running on your machine. Click Here For Figure You can also control the process priority. Certain processes on your server might be of greater importance than others; thus, they might be running with more resources allocated to them. This is called priority. You can control the priority of a process by right-clicking it in the list and choosing Set Priority. From there, you can choose to have the process operate in one of four different priorities: Realtime
If a program relies on instantaneous response and no delay, select Realtime. This allocates the highest priority to the process. High
If you want a process to run at better-than-normal performance, choose the High priority to allocate more resources to the task. Normal
If you want a process to run without any resource boosts or special priority, select Normal. As the computer slows down or speeds up, so will the process. Low
When you have a task that is not all that important, such as an idle program in the background, you might want to move it to Low priority. The changes you make in the process's priority is only effective while the process is running. If you end the process and restart it, or restart your machine, you will need to adjust the priority as needed. Performance View Performance view, as shown in Figure 15.23, does not involve any interaction. It displays the status of your machine, including the percentage of CPU usage, memory usage, as well as total processes and threads. In addition to this, a visual graph depicts both CPU usage history and memory usage history. This view is very useful for keeping an eye on how your computer is behaving. FIG. 15.23 Performance view acts like a dashboard, giving you a constant read out as to what is going on under the hood. Click Here For Figure Task Manager provides minimal customizability, mostly in the form of update intervals. This tool is most useful for taking a quick look at how your machine is doing. For comprehensive studying of your machine, Performance Monitor is more ideally suited. Of course, Task Manager is the only tool that provides you with the ability to instantly end processes and adjust their priority. Resource Management The tools mentioned in the preceding are used for monitoring your server. The actual task of keeping your server afloat relies on the management of your server's resources. Using the information given to you through the Performance Monitor, Event Viewer, and Task Manager, you can make educated decisions on where you need to adjust your server's resources. The essential areas you can look into are: Storage
Everyone knows that you can never have too much disk space. This holds especially true when you are delivering content onto the Internet or an intranet. As you need to grow your storage, you can begin by adding hard disk space to your server. This process and the limitations depend on your hardware platform. Beyond simply adding hard disk space, or removing files, you can look to larger storage solutions such as RAID arrays. Memory
The only commodity more valuable than disk space is RAM. As your server's load increases, expanding your system's memory will add relief in many ways. By increasing RAM, your server will be able to not only deliver content to more people, but also will be able to deliver more faster. Processing power
If you've looked to more memory and that simply hasn't helped the growing pains, you might want to look to a faster processor, or even multi-processor options. NT's support for multiple processors and symmetric multi-processing makes it a great platform for delivering content in high volume sites. Distributing the load
Growing your one server might be a losing battle if you are better off moving some services to separate servers. Windows NT's network operating system (NOS) heritage enables you to distribute your services to other NT Servers running within the same NT Domain (no relation to Internet domains). By selecting one server in your pool to become the Primary Domain Controller (PDC) and each subsequent server a Backup Domain Controller (BDC), your servers can take advantage of shared user accounts and permissions. If you are running several systems with different user bases, you might want to look at setting up multiple PDCs. Clustering
So, you've tried more storage, more memory, more processors, and more servers. It still didn't do the trick? Maybe it's time to look at clustering. Clustering is the powerful capability to group similar or even dissimilar servers to share the load. Clustering allows an entirely new level of fault-tolerance for servers, as well as performance increases. Clustering is available through third-party products, such as Digital Clustering from Digital Equipment, and from the Microsoft WolfPack Clustering API. Improving IIS Performance Beyond improving your server's importance, you can look to improving the performance of your Web server. Internet Information Server 4.0 has made considerable strides forward in handling server performance, thanks in part to the move to the Microsoft Management Console. As you discovered in other chapters in this book, the MMC is your centralized tool for controlling your Web server and the Web sites it hosts. In addition to global modifications, each individual Web site has the capability to be fine-tuned for particular performance ranges. To begin tuning your IIS services, open the Microsoft Management Console. Global Performance Changes Internet Information Server lets you define performance aspects for your entire server. These parameters in turn have an effect on each Web site and FTP site you host on that server. When you are in the Management Console, each of your IIS capable servers are listed in the Console tree, as shown in Figure 15.24. FIG. 15.24 The Microsoft Management Console lists each IIS 4+ server in the console tree. Click Here For Figure To begin modifying the global parameters for your IIS server, follow these steps: - Right-click your server in the tree and select Properties. The Computer Properties dialog box for your server appears, as shown in Figure 15.25.
FIG. 15.25 The Computer Properties dialog box defines the global parameters for your IIS 4+ server. Click Here For Figure - To limit the overall bandwidth consumption by all IIS services, select the Limit Network Use check box. With this check box selected, you can enter a bandwidth value in the Maximum Network Use text box, in the form of kb/s.
- To control the master properties for the WWW or FTP services, select the appropriate service from the Master Properties drop-down box and click the Edit button. Master properties for these services are covered after this list.
- When you are satisfied with your changes, click Apply and then OK to close the dialog box. If you are not content with your changes, click the Cancel button to abort and close the dialog box.
WWW Service Master Properties Using master properties, you can define the standard settings for all of the Web sites hosted on your server. These settings directly impact every Web site you host, and in turn can be altered for each site. Master properties are ideal for establishing common parameters between Web sites, such as connection limits and bandwidth consumption. Keep in mind that you are altering the overall settings for your Web service. Certain Web site-specific properties might not be available. Four main global properties that you can control directly apply to server performance. Each of these properties can also be modified for individual Web sites. These four properties are: Connection limitations
By setting a "house limit" for your Web site, you are less likely to have your server performance fall to a crawl while trying to service a large number of visitors. In addition to limiting the number of simultaneous connections, you can also define the Web service time-out. This time-out is used to determine when a visitor's session is closed if a network error occurs, rather than leave the session open indefinitely. Performance considerations
IIS4 provides you with the ability to tune your Web site for a particular size of audience. You can increase the priority to a particular Web site by allocating more resources to it. Less popular sites can be lowered in priority. Bandwidth throttling
If you want to limit the overall network bandwidth usage by the Web service, you can define an upper limit. This is convenient if you want to ensure that your entire network pipe is not used up by your Web server. Connection Configuration
HTTP Keep-Alives are used to hold Web server connections open for users. Because a connection is being held open, it can drastically affect overall performance by forcing more connections to be used. Limiting Connections When you open the WWW Service Master Properties dialog box, you are greeted with a tabbed series of pages. Each of these pages provides you with global properties that can be modified for your server. The first tab, Web Site, houses the Connections option group. This page is shown in Figure 15.26. FIG. 15.26 The Web Site tab lets you limit the total number of connections to your Web server. Click Here For Figure The Connections option group is located in the center of the page. It contains two radio buttons that control the total number of connections and a text box for defining the time-out value. To control the number of connections each Web site globally allows, follow these steps: - On the Web Site tab, if you want to limit the number of connections, select the Limited To radio button (deselecting the Unlimited button). The text box to the right of the radio button appears.
- Enter the total number of connections you want to allow to each Web site in the connections text box.
- If you want to modify the time-out value for all Web sites, enter a new value (in seconds) into the Connection Timeout text box.
- When you are satisfied with your changes, click the Apply button to commit your changes. If you are finished with this dialog box, click the OK button to close it.
Controlling Performance The third tab, Performance, contains the remaining global properties that affect performance. This tab, shown in Figure 15.27, contains controls for performance tuning, bandwidth throttling, and connection configuration. Controlling performance for your server is a simple matter of estimating the total number of visitors you are expecting for your server. With your estimate in mind, you manipulate a slider into a range fitting the total number of expected visitors. This alters the resources allocated to the Web service, in anticipation of your visitors. FIG. 15.27 The Performance tab contains three options for controlling performance. Click Here For Figure To tune your Web service's performance, follow these steps: - On the Performance tab, use your estimated total number of visitors to gauge where you should move the Tune slider. Moving it to the left allocates few resources for less traffic, moving it to the right allocates more resources expecting more traffic.
- Click the Apply button to commit your change. If you are finished with this dialog box, click the OK button to close it.
Limiting Bandwidth Bandwidth throttling sets a global limit on how much network bandwidth the Web service can use. All values for bandwidth throttling are based on kilobyte per second (KB/S). To limit the overall bandwidth for your Web service, follow these steps: - On the Performance tab, select the Enable Bandwidth Throttling check box. The text box below the check box appears.
- In the Maximum Network Use text box, enter the total number of kilobytes per second that you want to limit this service to. Note that you can override the master bandwidth setting for each individual Web site.
- If you are satisfied with your change, click the Apply button. If you are finished with this dialog box, click the OK button to close it.
Configuring Keep-Alives Finally, you can control the presence of HTTP Keep-Alives on your server. Some HTTP requests might contain HTTP Keep-Alives, which ask the site to keep the connection open during processing. Although HTTP Keep-Alives are enabled by default, you can choose to turn them off to minimize resources. To do so, follow these steps: - On the Performance tab, select the HTTP Keep-Alives Enabled check box. To have Keep-Alives turned off, make sure the box is not checked. If you want Keep-Alives enabled, make sure the box is checked.
- If you are satisfied with your change, click the Apply button. If you are finished with this dialog box, click the OK button to close it.
For users who are visiting your site with a slower connection, Keep-Alives prevent the connection to the server from being prematurely closed. If Keep-Alives are not used, connections are closed indiscriminately by the server. Keep-Alives involve both the client and the server, and can benefit you in medium load situations. FTP Service Master Properties Global properties for FTP sites are much simpler in nature. You are limited to controlling the total number of connections that your FTP service will allow. When you limit the total number of connections by your FTP service, you are setting an upper limit on how many individual FTP sessions can be opened on your server. More than one session can belong to the same user. If a user tries to connect and there are no more sessions available, he cannot enter until a session closes. To limit the total number of connections for your FTP service, follow these steps: - After you have opened the FTP Service Master Properties dialog box, switch to the first tab, FTP Site. This page is displayed in Figure 15.28.
FIG. 15.28 The FTP Site tab lets you control the total number of FTP connections allowed. Click Here For Figure - The Connection group acts identically to that of the WWW Service. To limit the total number of users, select the Limited to radio button (deselecting the Unlimited radio button). The text box to the right of the radio button appears.
- Enter the total number of connections allowed into the connections text box.
- If you want to modify the FTP service time-out value, enter a new value (in seconds) into the Connection Timeout text box.
- When you are satisfied with your changes, click the Apply button. If you are finished with this dialog box, click the OK button to close it.
Modifying Individual Sites When you alter the master properties for a service, all individual sites under that service are affected. If you prefer to alter each individual site separately, you can do so in the Management Console. Functionally identical to modifying the global properties, the properties for individual sites are exclusive. In addition to fine-tuning particular sites, you can carry out more extensive modifications on specific sites. When modifying global properties, not all options are available to you. That changes when you are dealing with each site separately. To modify individual site performances, right-click the site name in the Management Console tree. This opens the Properties dialog box for the selected site. Rather than repeat the instructions, use the references for modifying global properties when editing your individual sites. The principals remain the same, except that you are dealing only with one particular site at a time. You can use values that override the global properties, where needed. Internet Information Server 4 also provides you with the ability to tune your server-side programming. For comprehensive information on tuning your server for programs, see Chapter 16, "Understanding and Planning for Server-Side Loading." Web Site Considerations As mentioned earlier, the contents of your site can drastically affect your server and site performance. With each new standard and frill available to Web designers, more content is being forced on the visitor without any real regard to its effect. If you take into consideration the fact that most people who are visiting your site are likely connecting with a 28.8Kbps modem (or slower), it isn't difficult to see how more doesn't necessarily mean better. Several studies have indicated that the average user will wait no longer than 30 seconds for a page's contents to appear. If you overload your pages with sizable graphics or multimedia files, you'll easily push the load time beyond that. Considering the Implications of Files Everything you send to a user from your site is a file. It doesn't matter whether it is a graphic, a Web page, or an audio file. Everything comes down to a file, a logical grouping of information. Each different type of content is represented by different file types. Images can appear in several different file types (or formats), such as JPEG, GIF, or PNG. Audio files also have a variety of file types, ranging from WAV and AU files to RealAudio RAM files. When you are planning your site, it is important to consider what file type will have the most profound effects on your site's performance, for both speed and size. File size and delivery speeds are closely linked, but not necessarily mutually inclusive. Although a large file creates a slower delivery time, it is not absolute. A small file can also have a slower delivery (or display) speed if it requires additional software or processing to display. For example, if you choose to deliver streaming video to your visitors, you are already choosing to send a great deal of information downstream. To extend this further, you can choose to deliver the video in a format that requires considerably more overhead on the user's machine than simply downloading the sound file and playing it. In addition to causing more overhead for the user, your selection also causes more overhead for the server. First off, large files cause more strain on server resources. Connections are open longer, more memory and bandwidth is used, and of course storage space is consumed. In addition to this, if you choose a format that requires special server-side handling (such as streaming video rather than downloadable video files), the overhead of the streaming services can have considerable impact on your server. Here are a few points to consider when deciding on file types for your site contents: Calculate the download time
If your average user is connecting at 28.8Kbps, or lower, how long will it take him to view your page using different formats? If you are exceeding 30 seconds, you should carefully look at the necessity of such a large load time. Consider the additional overhead involved
Content of any kind uses resources, but some content uses more than others. If you are looking at additional services on your server, is it worth it? Do you need to deliver such large files? Is there a simpler way of doing things? Can you keep the pace?
If you are looking at delivering large content now, what about in the future? You can't always count on things getting smaller. More often than not, sites are getting larger and the content files are growing to match. Will your server and network connection be able to sustain your growth from where you start? Match your content types to your growth rate. Don't overload yourself too soon. Planning for Quicker Service Through careful planning, you can create sites that minimize delays and increase performance. Both the real-time server performance and the perceived performance on the user's end contribute to your appeal. Creating a Web site that is quick and fluid to use will no doubt add to your appeal. When planning your site, try to minimize extraneous content. Duplication or unneeded content is just more to send down the wire. Carefully look at what has to be sent down the line, and what complements it. Here are a few quick tips you can use to create quicker sites: Use the same graphics
Where possible, reuse your image files. Most browsers cache the graphics they receive, allowing for quicker loading on the user's end. This perceived timesaving reduces load on your server and improves response time for the visitor. Plan around slow connections
If possible, offer the ability to choose a high bandwidth/low bandwidth site. If you can provide a usable site to those who are connecting at low speed, you'll keep your visitors happy. And for those visitors who do have the speed to take advantage of, load them up with a richer experience. Use the most efficient means of delivery
Simply put, choose the most appropriate way to deliver your content. If you can deliver some content quicker through other means, do so. But always keep in mind simplicity for the visitor. Prune out the old and excess
As your site grows and progresses, don't forget to update your older pages. If a newer way of delivering your content comes around, move your content over to it. Don't force users to visit old pages that might have incredibly large and outdated file types. Recognize the effects of add-ons
Whether it's Java, ActiveX, or plug-ins, you might be adding more overhead to your page. Considerable system resources are used with many add-ons, and can create a slower (and more frustrating) experience for your visitors. Keep it simple
Don't overcomplicate. Only send what's needed. Don't waste the visitor's time with something that isn't needed. From Here With your Web server ready to take on the world, there are other important aspects to consider. The following chapters also cover performance tuning issues, and should be closely studied: Chapter 16, "Understanding and Planning for Server-Side Loading," specifically covers all of the elements involving server-side functionality. If your site uses these technologies largely (and why not?), flip to this chapter. Chapter 18, "Troubleshooting Techniques for Your Server," can be considered your emergency reference. If your server isn't performing like it should be, or you've stumbled on a huge problem, earmark this chapter. |
