Multi-Bay, Workload Customizable Server
Publication Date: 2014-Sep-09
The IP.com Prior Art Database
Described is an idea to use a single, low-cost server design that addresses the differing workload requirements by implementing multiple bay designs within the server itself. This concept allows customers to configure their servers in a way that eliminates unnecessary, costly parts of server designs and allows them to achieve a much improved total cost of ownership (TCO) for their business.
Page 01 of 4
Workload Customizable Server
Workload Customizable Server
The server industry has been evolving into an application and workload dependent environment with very specific hardware needs. Recently, the server market has provided this needed hardware flexibility through the use of blade servers, but at the expense of high startup costs and poor total cost of ownership (TCO). The idea discussed here is to use a single, low-cost server design that addresses the differing workload requirements by implementing multiple bay designs within the server itself. This concept allows customers to configure their servers in a way that eliminates unnecessary, costly parts of server designs and allows them to achieve a much improved TCO for their business.
The idea of the workload customizable server is to provide a base, low-cost server (2U in height in this disclosure) that uses easily swappable bays which can be configured with numerous different options. Using industry standard buses, the customer can configure their system in any number of ways and not be restricted by what is offered by typical server architectures. For example, a high performance computing (HPC) customer can purchase this server and populate most of the bays with memory modules, resulting in a server with more memory slots that is available from other offerings. An analytics workload customer can purchase the same server but populate all the bays with storage. With this concept, customers can now tailor their server design to match their exact needs while also improving their overall cost and system flexibility.
In the recent past, IBM* has been able to design servers that were versatile and cost effective enough to be used across several different industries. Customers would be able to purchase a system tailored to their needs and not have to worry about paying too much extra for features and functions they did not require. However, with the smart server push, the server industry has been going through a major shift. There are now very distinct market segments when it comes to server workload demands. The major server workloads IBM is concerned with are internet data center (IDC), high performance computing (HPC), and Hadoop**/Netezza* Analytics.
The three major workload areas mentioned above are very different, and designing a single server to address the unique workload requirements of each are very difficult. Due to this, most server companies design multiple servers to address these needs. For example, IDC requires more processing capacity and requires much less memory and storage. Analytics on the other hand requires a server heavy in storage capacity along with the processing capacity, and HPC requires a heavy memory capacity configuration. In order to get the optimal and most competitive total cost of ownership (TCO), at least two and maybe three different servers would be required to compete in the marketplace, a very high development cost propos...