Four Cost-Effective Mirroring Solutions Provide NetWare Redundancy

Global Data Security USA LANshadow 4.6 LANshadow 4.6 is an excellent low-end mirroring package. If you have a small, well-defined environment, can afford the mirror latency of its round-robin scheduling approach and have the patience to go through the details of configuration, this product will work well for you. However, for environments of any real size with real-time needs, this product is going to fall short. LANshadow is relatively easy to install via a Microsoft Windows administrative application that is reasonably intuitive, though somewhat awkward. We found that the product's simple design let us slip the package into an existing environment with no changes or additional software for the protected servers. Thi s is a plus for server administrators who are leery of installing additional NLMs (NetWare Loadable Modules) into their environment. All of Global Data Security's NLMs reside on the mirror server. Unfortunately, the simplicity of LANshadow is its greatest weakness. During installation, the interface for selecting directories of the protected servers to be mirrored does not let you make multiple selections and it doesn't create destination paths on the mirror server. There is a facility for grouping directories into families, but the selection problems don't make this facility much better. What Global Data Security's package is trying to accomplish is relatively straightforward. Unfortunately, the tools provided lack the sophistication that would make this product as easy to use as Global Data Security intends. Issues with the interface aside, our primary disappointment with the functionality of the product was the polling mechanism used for updating the data store on the mirror server. In this system, t he mirror server polls all of the protected servers' volumes sequentially. This method can create sizable delays depending on the size of the environment. You can sidestep this process by changing the scheduling for the mirror of particular protected servers, but the scan of the protected directories is still round-robin. Scott S. Campbell is a network systems analyst at Syracuse University. He can be reached at sscampbe@syr.edu.
	Breaking Out Two Basic Mirroring Product Types We found that the products tested can be categorized according to two primary architectural or design issues: the method used for updating the mirror (polling or real-time system calls), and whether the mirror server automatically fails over and stands in for a protected server, or simply provides an alternate location for accessing critical data. Interestingly, with the exception of LANtegrity (which is a hybrid), the four products we tested fit neatly into this matrix (see "Updating Versus Failover" below). The products that don't support automatic failover and stand-in, like NSI Software's Double-Take for NetWare and Global Data Security's LANshadow for NetWare, require that administrators establish their own failover procedures to accommodate users during a protected server failure. These can take the form of routines that redirect users' network environments (so the data stores on the mirror server are referenced instead of the failed server), or that actually down the mirror server and reboot it under the failed server's name. Unfortunately, the former solution would require that you determine the state of your data servers every time a client logs in, and the latter solution won't automatically take into account applications unless the administrator establishes failover routines for those functions as well. On the other hand, the products that do support automatic failover and stand-in, Vinca's StandbyServer for NetWare and Network Integrity's LANtegrity for NetWare, emulate all of the functionality of the failed machine. This is also true in the case of StandbyServer, which even reboots under the failed server's version of NetWare. The administrative load then is significantly less than the approach of manual failover and stand-in. For automatic failover, it's a hands-off situation, but the administrator has less control over the process of failover and stand-in. The concern driving the polling versus system call debate is primarily an issue of mirror latency. Systems that update the mirror via NetWare system calls (Double-Take, LANtegrity and StandbyServer) will be more in sync than one that polls in a round-robin schedule (LANshadow and LANtegrity), especially as the systems involved become busy. Even when the set of directories or files being mirrored is small, if there is a s teady stream of modifications, then the product that must scan the file store to notice changes won't be anywhere near as fast as one that listens to the actual file system for changes when and where they occur. Interestingly, LANtegrity takes a hybrid approach, offering both round-robin polling as well as real-time mirroring via system calls. With round-robin polling there is cache on the mirror server's disk for files that have been recently accessed, and all other files are packed off to the tape system, which makes scaling the system easy by adding tapes. However, the files that are maintained by real-time mirroring must have 1:1 disk space available on the mirror server, just as is the case in the other solutions. With this flexibility, we could use round-robin mirroring for files that are not very volatile nor time critical and use real-time mirroring for volatile or time critical files.

Other Reviews Web-Based Teamware Lets Users Work in Concert By Nancy Cox