Chapter 15 - Optimizing the Network

Introduction

This chapter provides information on optimizing VINES servers in the network. It describes how to perform the following tasks:

	Configure the file system cache on the server.
	Configure communications limits on the server.
	Configure swap areas on the server.

Each of these tasks is performed at the server console from the System Maintenance menu Configure/Diagnose Server selection.

Accessing Optimization Functions

To configure the file system, swap space, communications limits, or thresholds, type 3, 4, 6, or 10 in the selection field and press ENTER. The sections that follow describe how to use each function.

Configuring VINES File System Cache

The Configure File System function enables you to change three attributes of the file system:

	Change the size of the file system cache.
	Change the block size of the cache buffers.
	Turn read-after-write verify on and off. See the Banyan Server Operations Guide for more information on the read-after-write verify feature.

To configure the file system, select the Configure File System function from the Banyan Server Configuration menu. The Configure File System menu appears, as follows.

The menu displays the file system cache limits and the total number of cache buffers that are currently in effect for your server, and whether bad block remapping is turned on or off.

The sections that follow discuss how to use the functions on the Configure File System menu.

File System Cache Limits

The file system cache limits consist of two parts, both of which you can configure:

Total Cache - The amount (in kilobytes) of server memory that is used for caching frequently accessed data. The default size is 0, which specifies autosizing.

Cache Buffer Size - The amount (in kilobytes) of server memory that makes up a cache buffer. The total file system cache is divided into cache buffers. You can set the cache buffer size at 2 KB, 4 KB (the default size), or 8 KB.

In most cases, you can use the default file system cache limits. These defaults are in effect when you bring up your server for the first time and they should be able to handle your server's processing needs.

The Configure File System function is useful when a server is becoming overloaded. Depending on how your server is performing, you may want to increase or decrease total file system cache or cache buffer size.

Use VNSM to view performance data for your server, including file system statistics. See Chapter 8 for more information.

The performance of each server is different, and is affected by many factors. Personnel who configure the total file system cache and cache buffer size should be experienced server administrators who know all the factors that affect server performance. The expertise of the administrator using the Configure File System function determines its effectiveness.

Determining New File System Cache Sizes

If you decide to change the total file system cache, first decide whether you want to use manual configuration or autosizing to determine the total cache size:

	Manual configuration lets you set the total size of the file system cache to a specific non-zero positive integer. In this case, the label <configured> appears under the Total Cache field on the Configure File System menu.
	Autosizing lets the system determine the total cache size dynamically. In this case, the label [autosized] appears under the Total Cache field on the Configure File System menu.

To enable autosizing, set the total size of the file system cache to zero (0). When the server is first installed, the cache space size is set to a default. The default value depends on the amount of memory in the server. Following installation, the server automatically adjusts the size as a percentage of total system memory in response to system usage. Table 16-1 shows some sample default cache sizes.

If you use autosizing, you do not have to determine which sizes to specify.

If, however, you decide to use specific configuration, you first should know the following facts:

	How much main memory your server has. Main memory is divided into total file system cache, the communications buffer, RAM called executable space in which services and system processes run, and space in which the kernel and drivers run. See Chapter 3 for a description of how servers manage memory.
	How much memory is available for services. Increasing or decreasing total file system cache primarily affects executable space in which processes, such as file services, run. Decreasing file system cache frees up memory for services to run in. Increasing file system cache decreases the amount of available executable space for services. Use the Siz and freemem statistics of VNSM to determine service memory requirements.
	The Siz statistic, which is accessed through the SHOW service statistics function, tells you the memory requirements of each service. The freemem statistic, which is accessed through the SHOW OS information function, tells you the amount of executable space that is currently not being used by services, but is available to them in the event that their memory demands increase. See Chapter 7 for a description of the Siz statistic. See Chapter 14 for a description of the freemem statistic.
	The current total file system cache size. You can get your server's total file system cache size through the SHOW file system statistics function of VNSM. See Chapter 8 for more information on obtaining the total file system cache size.
	The results of increasing or decreasing total file system cache. Table 15-2 summarizes the favorable and unfavorable results.

It is a good idea to leave some extra executable space available for services, especially on heavily used servers. Service memory requirements are not static - they change depending on demand over time. The amount of executable space that you leave in reserve depends on the number of services on the server and the amount of demand placed on them. A range of 100 KB to 500 KB is a good average figure when considering the amount of executable space that you should leave in reserve. However, while this average figure is applicable in many instances, it may not apply to your own situation. Combine this figure with your understanding of your server's memory requirements.

For example, suppose that you want to increase the total file system cache size on a server that handles a moderate demand at the present time, and the freemem statistic tells you that you have 600 KB of executable space in reserve for services. You could increase the total file system cache size by 450 KB and leave 150 KB of executable space in reserve for services in the event that demand increases.

You should increase or decrease total file system cache only when:

	The favorable results outweigh the unfavorable results.
	The unfavorable results will not have a significant effect on server performance.

Two examples of changing the total file system cache are as follows:

	Using VNSM, you find that heavy disk usage is adversely affecting server performance. In addition, the swapping average for the server is low. Therefore, you can increase total file system cache and still have enough RAM for processes to run efficiently.
	Using VNSM, you find that the server has a high swapping average, but disk usage on your server is light. Therefore, you can decrease total file system cache without causing excessive disk usage.

After you modify total file system cache, use VNSM to detect any changes in server performance. Pay close attention to server load averages, swapping average, and disk usage statistics. These figures will help you determine whether the changes you made to the total size of file system cache improved server performance. You may have to experiment with different cache sizes until you find one that is best for your server.

Always remember that your two main goals are a cache hit ratio of 85 percent or better and a Swavg of 0.01 or less. When Swavg is non-zero, the server is paging or swapping. A Swavg of 0.01 means that paging or swapping is minimal. If you have to choose between swapping or paging and a low cache hit ratio, the low-cache hit ratio is the better alternative. If neither alternative is acceptable, it's time to add more memory.

If you find that heavy disk usage is adversely affecting server performance, you may be able to reduce disk usage without increasing total file system cache.

Frequently accessed files are usually stored in cache. If all cache buffers are in use, disk resources are required to provide buffers for these files. This results in higher disk usage. You can make more cache buffers available and improve total file system cache efficiency by reducing the cache buffer size.

Since only one file's data can occupy a cache buffer at any given time, some cache buffers are not totally used. This happens whenever the cache buffer size does not divide evenly into the size of a file. For example, if the cache buffer size is 4 KB and cache buffers are required to service a 17K file, 5 cache buffers - totaling 20 KB - are used. Four of the cache buffers are completely filled, totaling 16 KB of cache. However, only 1 KB of the fifth cache buffer is used, resulting in 3KB of unused cache that is not available to service other files.

Sometimes the file system cache can handle files more efficiently when divided into 4 KB cache buffers than when divided into 8K cache buffers. A 9K file would require two 8 KB cache buffers, but 7KB out of the total 16 KB would not be used. If the file used three 4 KB cache buffers, only 3 KB out of the total 12 KB would not be used.

Note: To determine the number of cache buffers a file requires, divide the file size (in kilobytes) by the cache buffer size (in kilobytes), and round the result upward to the next whole number.

If your server has a large number of small files (10 KB or less), setting the cache buffer size to 2 KB or 4 KB may improve cache efficiency. Remember that only one file's data can occupy a single cache buffer at a time. Keeping the cache buffer size at 2 KB or 4 KB reduces the amount of unused cache that is not available to service files.

Reducing the cache buffer size can produce unfavorable results. The server must assign cache buffers every time a file needs them, which takes up processor time. As the number of buffers increases, the amount of processor time used for file operations also increases.

For example, assume that the current cache buffer size for your server is 8 KB. For a 50 KB file, the server must assign seven 8 KB cache buffers. If you reduce the current cache buffer size to 4 KB, the server now has to assign 13 cache buffers for that same 50 KB file, which results in more processor time used.

As a general rule, reduce cache buffer size only when disk usage is high. If you notice that file operations (for example, copies) become unacceptably slow, you should reset the cache buffer size to its previous setting.

Increasing cache buffer size can improve the performance of the server, because a larger size reduces the number of times that server software must access the disk. For example, if a user is reading an 8 KB file from disk and the cache buffer size is 8 KB, only one disk access operation is required. Two disk access operations are required for the same 8 KB file with a cache buffer size of 4 KB.

As a general rule, increase cache buffer size only when your file operations are not performing as fast as you would like, and when disk usage is low. This means that you can increase the speed of your file operations without disk usage becoming too high.

You may have to experiment with different cache buffer sizes until you find one that is best for your server.

Using the Configure File System Function

Use this procedure to specify new file system cache limits for your server:

1. At the VINES Server Configuration menu, select Configure File System.

The Configure File System menu appears, displaying the limits currently in effect.

2. From the Configure File System menu, choose SPECIFY File System Cache Limits.

3. At the next screen, specify new limits in the following fields:

Total Cache - Specify the size (in kilobytes) of the total file system cache that you will allow on the server. Enter 0 for autosizing.

Cache Buffer Size - Specify the size (in kilobytes) of cache buffers. Valid values are 2, 4, or 8.

4. Press ENTER after specifying each limit. To keep a limit that is already in effect, press ENTER without specifying a new limit.

You return to the Configure File System menu.

5. To save your changes, select SAVE Changes and Exit or press F10. (To abort the procedure, choose ESCAPE Without Saving or press ESC.)

6. Reboot the server to put the changes into effect.

Configuring Communications Limits

Some of your server resources communicate with other nodes in the network. You can configure maximum limits for the resources the server uses to perform communications functions.

Note: Only administrators with networking experience should change the communications settings. Choosing settings requires familiarity with highly technical aspects of server communications.

Determining New Communications Limits

As a rule, do not reconfigure communication resource limits on your server. The defaults that are in effect when you bring up your server for the first time should handle most of your network communications needs.

However, increased network communications activity on your server may eventually force you to configure new communications limits. In most cases, increased activity results from adding services that produce a lot of network traffic to your server.

You can use VNSM to determine the amount and kinds of network communications activity on your server. VNSM displays all the configurable communications limits that are currently in effect.

You can configure these communications limits:

	Total size of the communications buffer.
	Maximum number of Sequenced Packet Protocol (SPP) connections that processes (for example, services) can have in use at one time.
	Maximum number of open sockets that processes can have at one time. Sockets apply to all protocol families on the server.

The memory resources that the server provides for communications functions make up the communications buffer. This buffer is the amount (in bytes) of server memory used for network communications.

Processes that communicate in the network require memory from the communications buffer. A process can be a service, a client, or any application that communicates in a network.

When two processes communicate, each process requires an open socket. Each open socket requires memory from the communications buffer. A socket acts as an interface between a transport layer protocol and a process during process-to-process communication. When a process is using a socket to communicate with another process, the socket is open; otherwise, the socket is closed.

Some processes also require an SPP connection when they communicate. The VINES SPP is a transport layer protocol that supports the transfer of packets on an SPP or virtual connection. An SPP connection acts as a data pipe between two processes in the network, such as the STDIRECT client and the STDA service.

SPP is used for communications between services and clients on DOS and OS/2 workstations only.

The following services use SPP to communicate with their respective DOS or OS/2 clients:

3270/SNA - Each active LU session requires an SPP connection to the client.

Asynchronous Terminal Emulation - Each active host session requires an SPP connection to the client.

VINES File Service - An SPP connection is established each time a user issues SETDRIVE to set a drive to a VINES file volume. The SPP connections are terminated when the user logs out.

Keep in mind that SETDRIVE commands in the user profile do not establish SPP connections for sessions with the file service. This improves login performance. For a drive that is set from the user profile, an SPP connection is established when the user performs an action on the drive, such as a read, write, CD, DIR, etc. Although a user may have many drives set from the user's profile, SPP connections for sessions with the specified file service may not be established while the user is logged in to the network.

STDA - The STDA service uses an SPP connection for a session with the STDA client.

Print - Each invocation of PCPRINT establishes an SPP connection to the specified print service.

Security - When a user logs in, the service establishes an SPP connection with the workstation to deliver profile information. The connection terminates as soon as the delivery is complete.

Third-party - Third-party services can also use SPP connections.

Default and Maximum Communications Limits

Each communications parameter has a default and a maximum configurable limit. Table 15-3 lists these values, which apply across all VINES server models.

To determine the exact number of SPP connections and open sockets you need, you should be familiar with all communications aspects of your server, including the number of users and the types of services that use SPP to communicate with their clients. The effectiveness of the limits you configure depends largely on your expertise.

Use the following VNSM statistics to help you decide on the number of SPP connections and open sockets you need and the communications buffer requirements of your server:

	The SPP and Skt statistics, which show socket and SPP connection use on a service-by-service basis. See Chapter 7 for more information on these statistics.
	The Total Comm Buffer Size, Comm Buffer Use, Allocate Failures, Configured Sockets, Sockets in Use, and Max Open Sockets statistics. See Chapter 6 for more information on these statistics.
	The Connections configured, % Connections in use, and Max Connections (HWM) statistics. See Chapter 10 for more information on these statistics.

If your server supports VINES communications only, you must set the total communications buffer size large enough to handle the maximum concurrent SPP connections and open sockets you want to allow. Otherwise, the server will reject connection attempts.

Each SPP connection requires 80 to 100 bytes of the communications buffer. Each open socket requires 70 to 80 bytes of the communications buffer. Use the following formula to determine the total communications buffer size:

(MAX CONCURRENT SPP CONNECTIONS X 100)
+ (MAX CONCURRENT OPEN SOCKETS X 80)
=TOTAL COMMUNICATIONS BUFFER SIZE (IN BYTES)

For example, suppose that you add more services that use SPP to a server. Using VNSM, you determine that the demand for communications buffer usage, SPP connections, and open sockets has increased. You used the defaults for all three limits; now you need to increase the maximum number of SPP connections and open sockets that can be in use at one time.

You decide your server can meet the demand for communication resources if you set the maximum concurrent SPP connections to 700 and the maximum concurrent open sockets to 450.

You calculate the total communications buffer size as follows:

(700 X 100) + (450 X 80) = 106000 Bytes (104 KB)

The default of 300 KB is sufficient to handle the increase in open sockets and SPP connections.

You may not have to increase the default communication buffer size (300 KB). This size is sufficient in most situations. However, if performance problems arise, increasing the default communication buffer size by a small amount, such as 20 KB or 30 KB, is often sufficient to get your server functioning smoothly again.

Calculating the communication buffer size that is best for your server is an iterative process. If your server has 16 MB of memory and you notice that an unacceptable number of allocates failed, gradually increase the size by 20 KB increments. The number of allocate failures should not increase by more than 200 per 100000 total messages sent and received.

If your server has more than 16 MB of memory, resides on a LAN backbone in a large network or acts as a router, and you estimate that your services have more than enough memory in which to run, don't hesitate to increase the communication buffer size by a significant amount, such as 100 KB. You will still have enough memory left for services and cache space.

AppleTalk and TCP/IP share the communications buffer and sockets with VINES. If AppleTalk or TCP/IP runs on your server, closely monitor your server's communications buffer and socket use and add more buffer space and sockets if necessary.

Specifying New Communications Limits

You specify new communications limits at the Configure Communications Limits menu.

Use this procedure to configure new communications limits for your server:

1. At the VINES Server Configuration menu, select Configure Communications Limits.

The Configure Communications Limits menu is displayed, with the limits currently in effect.

2. Choose SPECIFY Communications Limits.

3. At the next screen, specify new limits in the three fields. Press ENTER after specifying each limit. To retain a limit that is in effect, press ENTER without specifying a new limit.

4. After you press ENTER at the Maximum Concurrent Open Sockets field, you are returned to the Configure Communications Limits menu.

5. To save your changes, select SAVE Changes and Exit. You are returned to the VINES Server Configuration menu.

To abort the procedure, choose ESCAPE without saving.

6. Reboot the server to put the changes into effect.

Configuring Swap Areas

This section describes the new server configuration function of adding swap area on server disks. Please read this entire section carefully before making any attempt to add swap areas.

This section presents the following material:

	Overview of the swap area function
	Suggested alternatives and recommendations for its use
	Procedure for adding a swap area to a disk
	Deleting a swap area from disk

Overview of Swap Area Configuration

When the kernel pages or swaps data to disk, the kernel moves the data to the primary swap area on the server's first disk (disk1). The primary swap area is set aside on the disk solely for the kernel's paging and swapping activities (swapping is explained in the next paragraph). Nothing else may occupy that space on disk. If no paging or swapping occurs, the primary swap area remains empty, reserved for the next time it is needed.

See Chapter 3 for more information on paging and swapping.

Swapping adversely affects system performance and should be avoided. To determine how much paging and swapping is being handled by your server, look at the swapping average statistic (SWAVG) in VNSM. The ideal swapping average for a server is 0.0 (no swapping is occurring), but 0.01 is acceptable. If your swapping average over time is consistently above 0.01 on the server, there are several approaches you can use to alleviate the problem, any and all of which should eliminate the problem for most sites. These are outlined in the section, Suggestions and Recommendations, later on.

After pursuing all of those available alternatives for your server, you may find that the swapping average on the server is still unacceptably high. In that case, the primary swap area, which every server has on disk1, may not be large enough to accommodate all swapping. System performance will then degrade. Error messages may appear on the server console screen, warning that the server is about to run out of swapping space. After the server runs out of primary swap space, error messages state that processes have been canceled. You may need to add another swap area to the server's disks.

Determining the Need for Additional Swap Space

Some applications have memory requirements that place a heavy demand on the server. The application may have large storage requirements as well. For example, ORACLE®, a database application, needs a certain minimum of executable space at start-up time, but its needs grow with use. Memory that is allocated with each additional user may not be freed, reducing the amount left for other services. Eventually, a situation may develop where swapping activity increases so much that the primary swap area is overwhelmed.

To determine if you need to create more swap space, look at the number of free blocks of memory available to the primary swap area, on the new Configure Swap Space Menu. This step is explained in the Swap Area Configuration Procedure portion, later in this section. You can also determine the amount of free swap space by viewing the freeswap statistic in VNSM. See Chapter 14 for more information on the freeswap statistic.

As a conservative estimate, the number of free blocks for the primary swap area should not fall below 1000 on any server platform. If this occurs, you should then consider configuring additional swap areas on the disk. Likewise, if error messages warning that the server is about to run out of swap space appear on the server console, you should consider adding additional swap areas.

Follow the recommendations for this situation in Suggestions and Recommendations, later in this section.

Caveats and Considerations

Keep in mind the following caveats:

	The default size of the primary swap area is 32 MB for 386-based, 486-based, and Pentium-based platforms that are not manufactured by Banyan. If you upgrade a server to VINES 7.0, the default size of the primary swap area is 3.9 MB. If you fresh install VINES 7.0 on your existing server, you have a choice of making the swap are 32 MB or leaving it at 3.9 MB.
	The maximum number of swap areas allowed on any server is 16. You may add up to 15 swap areas, in addition to the primary swap area.
	On any disk but disk1, you may create up to eight swap areas. On disk1 you may add up to seven swap areas (though Banyan does not recommend it), because the primary swap area resides there.
	Once a new swap area has been configured, it is disk space that cannot be used for any purpose other than paging or swapping. Each swap area added means less disk space for other uses.
	Once a swap area has been created, it cannot be made larger or smaller. You can only configure additional swap areas after the first one has been added. Consider what size to make each additional area before you actually configure it. The system will reject any configured swap area under 8 blocks of 512 bytes each (4 KB), and any area over 16,384 blocks of 512 bytes each (8 MB). (512 bytes = 1/2 KB)
	If the disk cannot accommodate the size of swap area you have chosen, even though the size is within the 8 through 16,384 block range, the area will be trimmed to a size the disk can handle. Each swap area created is an area of contiguous blocks of space found on the disk. The Configure Swap Space Menu will remind you of the size of each block and the upper size limit for the total additional space.

Suggestions and Recommendations

There are two sets of recommendations included in this section:

	Suggested alternatives to try before you consider configuring additional swap areas.
	Recommendations once you have tried the alternatives and have decided that adding the swap areas is the best alternative.

Solutions to Try Before Configuring More Swap Areas

If you have set up optimal use of server memory and disks, you may be able to avoid having to add additional swap space. Before you attempt to add a swap area, try any or all of the following solutions:

Reduce the total file cache size. Reducing the file cache size increases the amount of RAM available as executable space for the kernel, drivers, applications, and so on. See the section "Configuring VINES File System Cache" earlier for more information.

Increase the amount of server memory. Consider increasing RAM to the maximum supported by your server platform.

Stop unused or lightly used services.

Reduce the number of users of a heavily used service. Ask users to use an alternate service, if available, on another server.

Move heavily-used services off of disk1. Move them to disks that are accessed less. Try to balance the work load in your server, so that no one disk (especially disk1) has more to handle than other disks. This should improve performance, but may not affect the amount of swapping.

Reduce the number of services on the server. Move services to other servers that are less heavily used. Make sure that the other servers have sufficient memory capacity to take on the additional work load.

Shut down and restart server software if the swapping average (SWAVG in VNSM) remains above 0.01 for a long period of time.

Move applications whose memory requirements are too high for the server. You may be using a third-party application that is not well-suited to the type of server on which it is installed. Move the application to a server that is better matched to its needs.

Configuring Additional Swap Areas

If the only alternative remaining for you is to configure additional swap areas on the server disk, please consider the following recommendations carefully before making an attempt:

	It is recommended that you do not place the additional swap areas on disk1.
	Use the SHOW disk usage statistics function of VNSM or the OPERATE command to view the disks on the server. Place any additional swap areas on your least-used server disks. If all of your disks appear to be heavily used, consider adding another disk before you configure an additional swap area.
	If the primary swap area is heavily used, consider adding more than one swap area to the chosen disk or disks. This will provide load-balancing of the swapping activity on the server.

For example, if you create only one additional swap area on disk2, the swapping load will be balanced between the primary swap space and the new one. Each will be tapped 50 percent of the time by swap requests. If you add two swap areas to disk2, disk2 will be tapped almost 67 percent of the time, while disk1 will be tapped approximately 33 percent of the time. The server continues to cycle through the swap areas that have space available, beginning with the primary swap area, tapping each in turn, until no swap space is left for paging or swapping.

	The free blocks available to each swap area, including the primary swap area, are shown on the new Configure Swap Space menu, accessed from the server console. The free blocks number decreases in an amount proportional to how often that swap area is being tapped. For example, disk2 may have 8000 blocks total, but only 7990 blocks are free; the rest are currently in use.
	Before adding any swap areas, consider your server's future needs and try to anticipate how they will affect the server's performance. Consider:

- How do you plan to use the disks?

- What additional services or applications will need to reside on the server?

- How large are its file volumes, and by how much will their data storage grow?

- Will the Network Mail service handle an increasing number of requests?

- Will you be moving the large application, the one that is pushing up the swapping average, to another server in six months?

Procedure for Adding a Swap Area to a Disk

Before you configure an additional swap area on the server disks, send a message to users notifying them that the server is about to shut down services. Then, proceed as follows:

1. At the console, choose Shut Down Server Software from the Operator Menu.

2. Select System Maintenance from the Operator Menu.

3. At the System Maintenance menu, choose Configure/Diagnose Server.

4. At the VINES Server Configuration menu, select the Configure Swap Space option.

5. At the Configure Swap Space menu, check the number of free blocks available to the primary swap area. If the number is below 1000 blocks, proceed to the next step.

If the number is above 1000 blocks, you do not need to configure additional swap space. In this case, press ESC to return to the VINES Server Configuration menu. Make sure that you restart services.

6. Choose CREATE swap area from the Configure Swap Space menu. The Create Swap Area screen appears.

7. In the Disk name field, enter diskn, where n is the number of a disk in your server. Do not put a blank space between "disk" and the number.

For example, to specify disk3 for the new swap area, enter disk3 at the prompt.

8. In the Number of 512-byte blocks field, enter the number of blocks, in multiples of 8, that will comprise the new swap area. The acceptable range is 8 blocks through 16,384 blocks. (Each block is 512 bytes, or 1/2 KB, in size.)

You should enter the number of blocks in multiples of 8 because the kernel swaps data to and from disk in 4 KB pages. For example, 8 512-byte blocks equals 4 KB (one 4 KB page). 16 512-byte blocks equals 8 KB (two 4 KB pages), and so on.

If you enter a number that is not a multiple of 8, the system rounds the number down to a multiple of 8. For example, if you enter 8001 blocks, the system rounds the number to 8000.

9. The Configure Swap Space menu displays the newly configured areas below the primary swap entry. In this example, the system administrator added swap area to disk2, and it has 7800 blocks of space:

If you want to configure another swap area, select CREATE swap area from the Configure Swap Space menu and repeat steps 7 and 8. Otherwise, select SAVE changes and exit or press F10 and return to the VINES Server Configuration menu. You do not have to reboot the server in order for these changes to take effect.

10. Return to the Operator Menu and restart services.

11. Return to the Configure Swap Space menu to look at the newly created swap area, and to observe how the server distributes data to the different swap areas.

Deleting a Swap Area from Disk

When you delete a swap area from the server disks, consider the following rules:

	You can delete swap areas that you have added. You cannot delete the primary swap area.
	You cannot delete a specific swap area from a disk. You can only delete all the swap areas on a disk.

For example, suppose that you have added two swap areas to a disk and you want to delete one of them. When you select one of the additional swap areas to be deleted, the system deletes both of them. You must then add the swap area that you want to retain.

Before you delete a swap area, have a complete system backup available. Send a message to users notifying them that the server is about to shut down services.

To delete a swap area, proceed as follows:

1. At the console, choose Shut Down Server Software from the Operator Menu.

2. Select System Maintenance from the Operator Menu.

3. At the System Maintenance menu, select Configure/Diagnose Server.

4. At the VINES Server Configuration menu, select the Configure Swap Space option.

5. Choose DELETE swap area from the Configure Swap Space menu.

6. At the DELETE swap area screen, enter the name of the disk on which the swap area resides. The Configure Swap Space menu appears. An asterisk appears next to each swap area on the disk that you specified. Keep in mind that you are deleting all the swap areas on the specified disk.

7. To delete the swap areas, select SAVE changes and exit or press F10.

- To cancel the deletion without exiting from the Configure Swap Space menu, select UNDO selections. The asterisks next to the selected swap areas disappear. You can then choose another swap area for deletion.

- To cancel the deletion and exit from the Configure Swap Space menu, press ESC.

8. Reboot the server. Changes will not take effect unless the server is rebooted.

Configuring Thresholds

From the Banyan Server Configuration menu, you can configure the Z-drive threshold and the router threshold for purposes of load balancing in the network.

Configuring the Z-drive Threshold

You can set a threshold on the number of clients for which the server provides a Z-drive. When the threshold is reached, the server delays before responding to additional Z-drive requests. This delay allows slower servers to respond and accept the client requests for a Z drive, and ensures that the requests are answered if no other servers are available.

Valid threshold values are zero (0) to 32766. Zero means that the server will not respond to client Z-drive requests even after a delay period. The default threshold is 32766.

When you change the threshold, recycle (stop and start) VINES Files or reboot the server for the new threshold to take effect.

To configure the Z-drive threshold, see the Banyan Server Operations Guide.

Configuring the Router Threshold

You can set a threshold on the number of clients for which the server acts as the routing server. When the threshold is reached, the server delays before responding to additional routing requests. You can also configure the delay period.This delay allows slower servers to act as routing servers after the faster servers reach their thresholds. It also ensures that clients get a routing server even if all servers are at their thresholds.

Valid routing threshold values are zero (0) to unlimited (32766). Zero means that the server will not respond to client routing requests even after a delay period. The default threshold is 32766, or unlimited.

The delay period is in units of 200 millisecond ticks. Valid delay values are from 1 (200 ms) to 50 (10 seconds). The default delay period is 1.

To configure the router threshold, see the Banyan Server Operations Guide.