InformationWeek
Defining The Business Value Of Technology
March 6, 2000
 Printer ready |
NetWare Cluster Services Expand On Reliability
Novell's clustering solution allows high-reliability failover service among much larger collections of servers than competing alternatives. Sharing a failed server's load over more nodes makes clustering even more affordable.
| Related links: |
|
|
| And from our sister publications: |
|
|
| TechEncyclopedia |
 Send Us Your Feedback |
s more companies rely on the Web for business transactions of all sorts, from E-mail to Web servers selling products, it becomes crucial to provide service 24 hours a day, seven days a week, with no interruptions or downtime. The old notion of being able to take systems down for maintenance at night fails completely in the Internet world--if you have a Web server, there will always be customers wanting to log on from somewhere where it's not the dead of night.Clustering is one way to increase availability. It allows a set of servers to be connected, sharing the load among the clustered systems if one server goes down. In the mainframe world, clusters also allow load balancing, so an application can be automatically moved from a heavily loaded server to one that's not as heavily loaded. That feature hasn't been realized in the network operating-system market yet. Rather, most load-balancing needs must be addressed with other technologies: load-balancing routers or features such as the IP load-balancing support in Windows 2000. In some cases, such as relatively static Web sites, these load-balancing solutions provide many of the same benefits as a full-blown cluster solution. You'll have to manage mirroring updates to the Web site's content across all the load-balanced servers, but that's relatively trivial. For other applications, such as transactional databases or groupware, maintaining two or more copies of the data in perfect synchronization is a major obstacle--hence the need for a real clustering solution.
Until now, clustering in that marketplace really meant only two-node failover, where no more than two nodes could be clustered, with one taking over applications from the second in case of a failure. The problem with two-node clustering is that it severely limits the use of the servers. If the servers normally run near 50% capacity, diverting the combined load to a single server would overload the surviving platform.
Novell is the first major network operating-system player to market with a multinode clustering product. NetWare Cluster Services for NetWare 5 allows up to eight nodes in a cluster out of the box and will support up to 32 nodes with special support from Novell. This allows servers to be used more efficiently--individual servers can carry heavier loads, since each application can be distributed to a different server, rather than having only one other server to which to failover.
NetWare Cluster Services for NetWare 5 also supports failover clustering of both applications and network services, including file and print services as well as services such as Dynamic Host Configuration Protocol (DHCP) and domain name system. This lets administrators cluster critical network services as well as applications such as Web servers or database servers.
NetWare is installed on two or more servers in what is to be the cluster. The servers share a storage device, either a Fibre Channel storage system shared through a daisy-chain or Fibre Channel hub or switch, or a shared SCSI device. The shared volume allows files from one server to be accessed by the server that's taking over its duties. The servers communicate synchronization information over a standard network interface--a separate network hooking the servers together isn't required, although it can be set up that way.
Shared storage isn't strictly required either--it's possible to replicate data between the servers using a product such as Novell Replication Services to replicate data to all servers and do without shared storage, although this would require that each server have more storage than a shared storage solution would use. Also, if applications were all using static data, such as Web servers, it would be possible to have all the data reside on all servers--meaning that replication would not be necessary.
NetWare Storage Services and TCP/IP are required. NetWare Storage Services ensures that volumes can be dismounted from the inactive server and remounted by the next server in a timely manner. NSS volumes mount much more quickly than standard NetWare volumes, regardless of the size of the volume or the amount of data it contains.
TCP/IP lets clients reconnect to the new server without having to reboot the client. IPX/SPX isn't supported in the cluster environment, since clients would generally have to reboot their workstations if a server in the cluster failed.
continued...page 2, 3
Back to Labs
Send Us Your Feedback
Top of the Page
