Name and Address Resolution for R/3 Servers 

Overview

The R/3 System uses IP names and IP addresses to identify R/3 servers and external communication partners. Therefore, the correct configuration for resolving IP names and IP addresses is extremely important for the R/3 System to function properly.

Read and take these rules into consideration before installing an R/3 System.

All the operating system platforms supported by SAP offer various mechanisms for name and address resolution (for example, the Hosts file and DNS). These are explained briefly in the second part with respect to the R/3 System.

As a prerequisite, you must have basic knowledge of the TCP/IP protocol. (See also: Basic Principles of TCP/IP Configuration for R/3 Servers.)

 

Terminology

Translating an IP name into an IP address or the reverse, is called resolution. This can occur in two directions (see graphic). The most common case involves resolving an IP name into an IP address. For the R/3 System, the opposite resolution is also very important (this is called reverse lookup).

 

 

Requirements for the R/3 System

Due to the client/server architecture of the R/3 System, there are specific requirements for R/3 servers and R/3 clients when resolving IP names and IP addresses.

R/3 Servers

The sections Host Names for R/3 Servers and Configuring R/3 Servers with Multiple Network Cards list the rules for assigning host names, IP names, and IP addresses for R/3 servers. Follow these rules to ensure smooth operation of the R/3 System.

You must meet the following requirements for your R/3 System to function properly.

The following requirements also apply for communication functions:

In addition to the name resolutions necessary for the system to function, other reverse lookups occur that are not functionally necessary. This type of name resolution occurs, for example, so that the actual names are displayed in monitors and in the system log, instead of the numerical IP addresses.

In these cases, the resolution does not have to return a valid result for the R/3 System to function properly. However, if the configuration for resolving names is incorrect (in the operating system or in the name server), there may be long wait times due to timeouts, which can have a highly negative affect on R/3 System performance. Therefore, ensure that the configuration is correct when using name resolution with DNS (Domain Name System).

Buffering Names and Addresses in the R/3 Server

The SAP Network Interface (NI) is a part of the R/3 System, and stores the result of the name resolution and reverse lookups in a special buffer. This means that the operating system only has to perform the name resolution once. Successful resolutions remain in the buffer until the R/3 System is shut down, or until an overflow, or until the buffer is manually deleted (Transaction SM51). Manually deleting the buffer does not include the message server. For this reason, restart the R/3 System to make the changes of IP names or IP addresses effective in the system.

If a resolution fails, this is noted for a duration of 10 minutes (as of R/3 Release 4.0). This is because the name server may not be available. If another attempt is made to resolve the name after this period of time, the R/3 System repeats the operating system call.

The advantage of buffering in the R/3 System is that it increases the performance during name resolution and it only very rarely reaches a critical situation.

R/3 Frontends (SAPgui)

To be able to use all of the R/3 functions, the frontend computer must be able to resolve the names of the message and application servers into IP addresses.

During load balancing, the SAPgui receives a list of IP addresses from the message server. Resolution does not occur.

If the frontend computer cannot resolve the names of the R/3 application servers, you can also start the SAPgui by using numerical IP addresses. However, the functions may become limited, for example, when you use certain desktop components.

External RFC and CPI-C Communication Components

An RFC or CPI-C client establishes a connection to the R/3 System. To do this, it requires the IP address of an application server. You can enter this address directly to establish the connection. In this case, no resolution occurs. You can also use load balancing with logon groups for RFC. Here, the RFC client connects to the message server and receives a list of IP addresses from the server. No resolution occurs here.

Server programs are usually started by an SAP gateway, or by a SAPgui. Each R/3 instance has a gateway. You can also operate an SAP gateway without an R/3 instance (standalone). After the server program starts, it must establish a connection with the gateway within a specific period of time. You can start the program in different ways:

Start using a remote shell: On the command line, the server program receives the IP name of the gateway to which it should connect. This name must be able to be resolved.

Start using the SAPgui (only RFC): The command line communicates the information with which the SAPgui was started to the RFC program. It also communicates a SAProuter string is used. This means that the resolution is identical to the resolution with the SAPgui.

Naming Resolution Mechanisms

There are four common mechanisms for assigning names with addresses:

The operating system determines which mechanisms are available, which are active by default, and how they are configured. The administrator can decide which of the available mechanisms are used. In most operating systems, you can configure several mechanisms with different priorities. These are used one after the other, if the first mechanism does not give you the result you want.

Choosing a Naming Resolution Mechanism

The configuration of the naming resolution mechanism on R/3 servers must meet the following requirements:

Hosts File

The Hosts file contains a list of IP addresses and the accompanying IP names. The resolution occurs by searching through the list. Under UNIX, the file is called /etc/hosts , under Windows NT it is called \winnt\system32\drivers\etc\hosts .

The Hosts file is a robust and simple solution, since all of the information for naming resolution is available on the computer. No network access to other computers is required to resolve names. The advantage is very high performance. However, maintaining the file requires some effort, especially if there are many hosts, or the hosts are divided among many separate networks.

When you maintain the Hosts file, you must enter the IP name that corresponds to the host name as the first name directly after the IP address, and not as an alias (second name).

NIS

The Network Information Service (NIS) is a distributed database system that replaces configuration files, which usually have multiple copies, with a central administration. Files such as /etc/hosts , /etc/passwd , and so on do not have to be administrated in each host. Instead, the information is retrieved as needed from the database.

If you use NIS to administrate the Hosts file, only the settings in NIS are valid. The local Hosts file is no longer used. This behavior depends, however, on the operating system.

NIS is a centrally administrated system with a master server. To improve availability and performance, you can set up additional slave servers that answer the client queries. Since access to an NIS server is required for each resolution, you must have high availability, short wait times, and you must carefully plan the NIS system for your system to work smoothly.

WINS

WINS (Windows Name Service) is suited for use in a pure Microsoft Windows environment and cannot be directly compared with the other mechanisms described here for naming resolution. One main difference is that no IP names, but NetBIOS names, are mapped to IP addresses. The advantages of WINS are its simplicity, robustness, and flexibility. The disadvantage is that WINS is only suited for smaller, flatly structured and homogeneous Microsoft networks.

DNS

The Domain Name System (DNS) can map complex, hierarchical structures with a large number of hosts. It lets you use efficient centralized and decentralized administration. For example, the entire Internet uses a single DNS infrastructure for naming resolution. DNS is the most robust of all methods of naming resolution, and is the most well suited for the future, for example, for potential integration with directory services. DNS is available on all server and frontend platforms supported by SAP.

With regards to availability and wait times, the same applies here as it does for NIS. With DNS, planning is particularly important, since configuring and maintaining DNS is complex and, therefore, prone to errors. The most dangerous errors are not those that lead to a total failure, rather those that lead to performance problems, because they remain undiscovered.

Tips for DNS Configuration

The effects of an incorrect DNS configuration are explained in the following example:

If a reverse lookup in the DNS is not maintained, and the name server in the corporate internal network is configured incorrectly so that it accesses the central name server in the Internet for unknown domains (this is the default setting for certain name servers), the name server waits for several minutes for a response from the inaccessible Internet name server. During this time, the R/3 work process from the operating system is stopped and the user cannot continue working.

Wait times can also occur due to connection problems between the application server and the name server, or between two name servers.

If you want to use DNS (you need some basic knowledge of DNS):

Conclusion

Deciding on which mechanisms you should use for naming resolution depends on your priorities. When making your decision, remember that the name resolution is a critical function in the R/3 System that influences the system operation.

WINS is only used in pure Windows environments. In the future, it will not be supported in certain circumstances, and we therefore do not recommend it.

DNS is the most flexible and powerful solution, almost a standard. However, its use requires a high level of knowledge about implementation and maintenance. Buffering in the R/3 System means that wait times are not critical, within certain limits. However, you must have high availability for the DNS server and the related network connections. If you already have a functioning and well-maintained DNS infrastructure, we recommend using DNS on the R/3 servers and frontends.

Simplicity and robustness are the advantages of the Hosts file. It is therefore highly recommended for the R/3 servers. Generally, the R/3 server does not have to use the same mechanism as in the rest of your corporate network. However, maintaining and distributing the Hosts file may require a great deal of effort to administrate.

On UNIX servers, you can replace the Hosts file by NIS. Here, like DNS, you must pay attention to proper implementation and high availability of the NIS server to ensure that the R/3 System operates properly.