Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Son of SAN - Storage Networking Technologies -- Ignore unavailable to you. Want to Upgrade?

To: Wei Yu who wrote (1580)	11/8/1999 4:08:00 PM
From: Douglas Nordgren	Respond to of 4808

Wei, here is a link for ordering the entire proceedings for the FCTC ($295) - fibreconference.com Thanks for the CPQ clustering info. I believe NT clusters have now gone up to 8 (4 pairs of 2-node failover). However, I do not feel that fail-overs are true "clustering" a la Unix clustering. They just add a fault-tolerant element to NT to barely meet High Availability requirements. Will rebooting one server reboot the other fail-over server? (that is meant to be a joke, sorry.) Douglas

To: Wei Yu who wrote (1580)	11/9/1999 1:36:00 AM
From: Douglas Nordgren	Read Replies (1) \| Respond to of 4808

Now for true Clustering. Need some Load Balancing and Application Partitioning? It doesn't come with Win 2000, yet, despite the misnomer of "Microsoft Cluster Services." Blame the HAL. This story appeared on Network World Fusion at nwfusion.com. Why clustering beats SMP for Web sites By LAWRENCE GENOVESI Network World, 11/08/99 Organizations with large Web sites face dynamic demands that require servers to be highly scalable, manageable and reliable. IT departments traditionally have turned to symmetric multiprocessing (SMP) systems to meet these needs. But in today's Web environments, SMP systems lack the scalability to keep pace with growing user traffic, and they are complex and expensive to implement. Clustered Web servers, on the other hand, easily scale with demand, and are straightforward to implement and manage. Traditional SMP systems can be powerful, but they require an enormous amount of technological expertise to install, maintain and upgrade. A further limitation is that SMP systems share every resource, meaning that the bus, network, memory and I/O systems are shared by the CPUs. The system uses a single copy of the chosen operating system and database. Multiple CPUs share the processing work, and contend for resources such as disk drives and network cards. Diagram of how it works:http://www2.nwfusion.com:8001/cgi-bin/print.cgi?article=http://www.nwfusion.com/news/tech/1108tech.html#diagram SMP-based systems communicate using a common shared memory pool via interconnect buses. However, as the number of processors in each system increases, the amount of traffic on these buses does as well. The increased traffic becomes a bottleneck that limits system throughput and, therefore, overall system performance. In contrast, clustered systems connect system resources and CPUs in a manner known as loose coupling. Clustered servers have load-balancing software that distributes work to the server within the cluster that has the most available resources. Each server has its own operating system and associated resources from which to draw, minimizing share-based performance degradation. Workloads can even be distributed among multiple clusters, thus increasing overall system throughput and performance. Including intelligent load balancing as an integral part of a Web server configuration is a cost-effective way to achieve a high degree of scalability, maximum server performance and consistent service availability. Servers clustered with load-balancing functionality allow network administrators to monitor workload more effectively and distribute it to optimize use of available capacity. Justas important, administrators can increase capacity by adding Web servers knowing that the workload will automatically be balanced and optimized.With built-in load-balancing software, network administrators have the flexibility to achieve redundancy and support growth. First, they can achieve greater gateway redundancy in larger site configurations by simply installing the gateway service on multiple Web servers. In addition, through the use of agents, load balancing can be extended even further by balancing the workload not just across servers but also within each server. This is achieved with a technique known as logical partitioning, combined with intelligent agents that reside in each server. Through logical partitioning, a certain percentage of each Web server is assigned to handle individual applications. The agent will monitor the usage of each application and Web site as it compares to the percentage specified for that Web server and balance the load as needed across other servers within the application partitions assigned to them. The network administrator can dynamically adjust the percentage an application occupies on a given Web server at any time. If an application xperiences increased usage, its percentage can be increased while reducing others. This technique provides a powerful means to manage capacity in an efficient and cost-effective manner. Internet growth and the demand for access to content make it difficult to provide and manage large Web sites using the complex tools of the traditional client/ server model architecture. Network administrators no longer have the luxury of bringing systems down to upgrade them - there are no off-hours on the Web. Maintaining a flexible clustered architecture allows companies to respond quickly to fully meet user demands