Windows NT Server Consolidation
Solaris PC NetLink has always had the ability to consolidate several Windows NT 4.0 servers within the same NT domain into one Solaris PC NetLink server. The following example illustrates this kind of consolidation:
In FIGURE 1, all three Windows NT servers names: futomaki, ebi, and unagi are consolidated into one Solaris PC NetLink server. Only one edit to the /etc/opt/lanman/lanman.ini file will accomplish the task. Add the line aliases=futomaki,ebi,unagi to the [lmxserver] section of the file. Once this is done, the Solaris PC NetLink server will respond to all three NetBIOS names. The PC clients that were using the Windows NT servers need not be changed in any way. However, a reboot at transition time may be needed to access the new server.
FIGURE 1 /etc/opt/lanman/lanman.ini Server Aliases
While this is an excellent way to consolidate servers that reside in one Windows NT domain, it does not allow consolidation of servers that are in multiple Windows NT domains. Prior to Solaris PC NetLink 2.0, the only way to accomplish multi-domain consolidations in one server was by use of Sun EnterpriseTM hardware domains, as illustrated in FIGURE 2:
FIGURE 2 Sun Enterprise 10000 Hardware Domain Used to Consolidate NT Servers
In FIGURE 2, each Sun Enterprise 10000 domain (up to 16 per server) can run a copy of Solaris PC NetLink. Each copy of Solaris PC NetLink, in turn, supports a separate Windows NT domain.
Multiple Instance Support
Solaris PC NetLink 2.0 can support multiple instances of Solaris PC NetLink running simultaneously on one server. This new functionality allows Solaris PC NetLink to support multidomain Windows NT consolidations even on the smallest Sun Enterprise server.
FIGURE 3 Multiple Windows NT Domains Supported on One Sun Fire V480 Server
In FIGURE 3, all three Windows NT domains have been consolidated onto one Sun Fire V480 server. The Sun Fire V480 server supports up to four processors and runs one instance of the Solaris operating environment. Sizing, performance, and availability options should be considered fully before proceeding into a full-scale consolidation of multiple domains.
It may be difficult to determine exactly how heavily Windows NT servers are being used, especially during peak periods. With this new multi-instance functionality of Solaris PC NetLink 2.0, a customer might decide to move Windows NT domains to the new server one by one. After each transition, performance and resource consumption should be monitored during peak use periods to determine if further consolidation is possible on the new server. Solaris servers that have under-utilized resources would continue to be candidates for additional Windows NT server consolidation.
Support for multiple instances of Solaris PC NetLink not only increases the number of ways used to consolidate Windows NT servers, but also plays an extremely important role in increasing the efficiency of servers used in HA clustered solutions.
High Availability and Clustering
Support for Sun Cluster 2.2 was introduced with Solaris PC NetLink 1.2. This support allows the creation of a virtual server. This virtual server would be supported by a cluster of physical servers or nodes. The node that manages the resources in the cluster is called the default master node. The other nodes in the cluster are called backup nodes. When Solaris PC NetLink is interfaced with cluster data services it can become part of an highly available application that can move from node to node.
If a copy of Solaris PC NetLink running on one node in the cluster experiences a problem, Sun Cluster detects the problem and it relocates the application to a backup node. This process of relocating the application service from a failing node to a surviving node is called failover. The Sun Cluster must encapsulate all the information associated with the data service into a logical host. All the client accounts, IP addresses, data management structures, and processes associated with the data service need to move across nodes without modification for the failover to work.
Unlike Solaris PC NetLink 2.0, Solaris PC NetLink 1.2 only allows one instance of the product to run on any one server.A logical host is defined an IP address (or a hostname) associated with an application service that can be hosted by an active cluster node. The file structures, exported by Solaris PC Netlink, are supported by disk storage mirroring which is a basic Sun Cluster requirement to support availability
In cluster environments of two to three nodes, only one logical host can be created, which can support only one Windows NT domain. FIGURE 4 shows a typical cluster with Solaris PC Netlink 1.2.
FIGURE 4 Sun Cluster 2.2 Support With Solaris PC Netlink 1.2 Software
In this configuration, both the master and backup nodes of the cluster have direct access to Storage A and Storage B. Both Storage A and Storage B are mirrored volumes to offer High Availability. Under normal operation, the master node is the active side of this two-node cluster and supports the logical host PCNL1. If a problem occurs on the master node, the failover occurs, and the backup node takes over the role of supporting the local host PCNL1. FIGURE 5 illustrates the result of this failover.
FIGURE 5 Standard Failover With Solaris PC NetLink 1.2 Software
This configuration has a high level of availability, but has several limitations. Remember, Solaris PC NetLink 1.2 supports only one instance of the software running on any node. Likewise, the cluster can only support one of Windows NT domain . A more significant limitation is that only one of the nodes actively supports the Solaris PC NetLink data server for the cluster during normal or failover operation. This means that other nodes in the cluster are completely inactive with respect to Solaris PC NetLink operation. If other services, such as email or a web application, are required in this HA environment, the other node could be actively supporting these services, more fully utilizing the available resources.
Solaris PC NetLink 2.0 allows multiple instances of the software to run simultaneously on the same node of a cluster. This allows many Windows NT domains to be supported on a cluster simultaneously. In addition, all the various nodes of the cluster can actively support different domains. This allows all resources to be utilized under normal Solaris PC NetLink operation. FIGURE 6 shows a simple example of such a configuration.
FIGURE 6 Solaris PC NetLink 2.0 Multiple Instance Support and Clustering
With the configuration shown in FIGURE 6, Windows NT servers in both Windows NT domains CORP_DOM and ENG_DOM can be consolidated into a highly available clustered environment. If a failure occurs in this cluster, either node of the configuration is capable of supporting all the servers involved in the consolidation for both Windows NT domains If the instance of Solaris PC NetLink software supporting the logical host PCNL1 and the Windows NT domain CORP_DOM failed for a software or hardware reason, the result after the failover is shown in FIGURE 7.
FIGURE 7 Multiple Instance Support After a Node Failure
FIGURE 7 shows the right node supporting both the CORP_DOM and the ENG_DOM domains through the logical hosts PCNL1 and PCNL2 simultaneously.
This example is very simple. A consolidation of a real Windows NT environment could involve many Windows NT servers, from several Windows NT domains all going to multiple instances of Solaris PC NetLink, to several Sun servers all involved in a large cluster configuration.
One goal of any Windows NT server consolidation is to minimize or eliminate any system administration of the multitude of Windows clients that use the servers. As support for each Windows NT server is brought into the cluster of Solaris PC NetLink logical hosts, the PC clients that routinely use the server would not even be aware that the change has taken place.
Performing a large Windows NT server consolidation is best done in phases, as each Windows NT server is replaced by clustered Solaris PC NetLink configuration.
Again Solaris PC NetLink 2.0 allows a more efficient utilization of resources in the cluster to be utilized under normal operation. Careful thought is required to determine the acceptable failover performance requirements during peak use periods. All failover states should be analyzed to determine acceptable failover performance. While Solaris PC NetLink will continue to operate properly even when the server is fully saturated, latency times will continue to grow as the load is increased.
For most installations, nodes expected to act as backup nodes for clusters should have excess resource capacity in the form of additional CPUs and memory to support a failover as required. Except on very busy peak periods, most users of the servers will not notice the difference between normal and failover performance. Typical productivity applications may take a little longer to read and file to the server.
Other Situations When a New Instance of Solaris PC NetLink Is Needed
A new instance of Solaris PC NetLink, running on the same server, is needed whenever you want to support a new Windows NT domain. Other situations where you will want to create a new instance of Solaris PC NetLink are as follows:
When Windows NT exported file systems (shares) from different servers have the same path nameDuring a Windows NT server consolidation, it is likely that the shares from different Windows NT servers will have the same path name(For example, \\server1\cdrive and \\server2\cdrive). Unless you are willing to juggle directories within the shares to avoid collisions, it may be best to create a separate instance of Solaris PC NetLink to support some servers separately.
When there are security issuesIf, during a consolidation, it is determined that the shares of one server are protected differently than those from others, then a new instance of Solaris of PC NetLink may be warranted. For example, users of one server may not want other users seeing their shares.
When the concurrent user limit is reachedDifferent versions of Solaris PC NetLink have different limits for the maximum number of concurrent connections they will support. When this limit is reached, the PC connections can be distributed across the multiple instances of the software.
When quality of service is an issueMultiple instance support can help manage the resources consumed by a group of users. Solaris Resource Manager (SRM), or some other resource management technique, can ensure that VIP users or mission critical applications have a higher priority than other activity on the server.
Improvements in Performance Elsewhere
Both Sun Enterprise servers and the Solaris operating environment have improved the ability of Solaris PC NetLink to perform better and scale further. Benchmarks done with Solaris PC NetLink 1.1 running on the older Solaris 2.5 operating environment, compared with those done on the current Solaris 8 and 9 operating environments, show that even this older versions of Solaris PC NetLink can scale further with the new operating environments. This improved performance is mostly due to performance improvements in the file system and the memory manager as well as the drivers for supporting the network interfaces.