Junk

1 IT Infrastructure and Management

1.1 Summary

Within a large Enterprise, there are typically distinct groups, Operations and IT, that are separately tasked to manage data center operations versus end-user computational resources including departmental or workgroup servers. In smaller companies, these services are often combined in a single functional group. This usage case focuses on the domain of IT and in some cases its broader interface to services provided by the data center such as defined earlier in the data center usage case.

In widescale use currently, Systems Management software is used to address the concerns of IT typically in the areas of: Configuration Management, Software Distribution, Asset Management, License Management, and Help Desk. In the realm of CDDLM, we will focus primarily on Configuration Management and Software Distribution, though it is understood that much of the remaining functionality will be serviced elsewhere in the Grid.

In a typical usage pattern, an end user will need to connect to some representative Grid Service. First, the user must have access to the correct software for interaction with the service. This may imply the need to deploy dependent components to the end user's workstation a priori, or in the event of use of a particular service. Also, the Grid service may be one of several configurations of services. It may be housed on a defined server or set of servers, a mainframe, or supercomputer. It may also be run on an available hive of workstations in a workgroup. In any of these scenarios, it will be important to ensure that the correct software is deployed, configured and managed through the duration of the Grid service. These instantiations must also be managed within the larger context of the Grid, with multiple jobs spanning LAN and WAN over multiple compute clusters.

The role of IT will be to ensure that end user's needs are met. They will assume the responsibility of the deployment or coordinating that deployment with the Operations staff in the data center. Even if the deployment is primarily done in the data center, IT will need to access the configuration, deployment, and event information.

1.2 Customers

The ultimate customer in this use case is the corporate knowledge worker consuming the physical service itself. The corporate IT and operations staff are also participants. And in many Enterprises, the software development team may play a role in the deployment and configuration of services for the customer.

Knowledge workers on a daily basis will interact with abstract services on the Grid in order to perform their industry specific tasks. These services may be batch oriented, or real-time submissions of commands to be processed. The user will require that a physical execution environment be created for their service, and software configured and deployed. During the execution, the end user will monitor the progress of the execution, or await its final disposition.

In the configuration of most Enterprises, these usage scenarios will mostly be within a single physical site. However, access to central or remote repositories is often needed for access to data or specialized computational resources such as mainframes or dedicated compute clusters. The number of users participating in this scenario can be quite substantial, typically several thousand users per office facility or site, and tens or hundreds of sites around the world. Additionally, there are typically one or more central data center sites where large computation facilities are housed.

1.3 Scenarios

Generalized Grid Service - A currency trader at a global investment bank has thought of a new model for evaluating intra-day pricing fluctuations. He discusses this model with the Engineering group and requests a Grid service be developed. The service is designed and its requirements are discussed with the Operations team. It is decided that the service will live in the data center and be placed in a dedicated pool of servers that have access to the historic database of global currency prices.

The trader awaits the availability of the service. The Engineering team builds a Web-based user interface for allowing the trader to create instances of the service with different configurations. He is also given screens for submitting different types of queries with different parameters and formulas. However, there is no requirement other than a general purpose web browser for his own personal machine.

End User Device Deployment - The currency trader, happy with his model's performance, decides that he needs to do some OLAP analysis of the results sets from his service runs. The Engineering team suggests that they build a Visual Basic application to front-end his Grid service. As his bank provides floating desktops, he needs to ensure that he will always have access to his service.

As part of the development of the service, he asks the Engineering team to model and configure the application as a service itself that can be deployed as needed, to whatever workstation he may be using. In this manner, when he invokes the service from the Web, his workstation will be configured with the OLAP application.

High End P2P Computation Service - The currency trader decides that his model is perfect for his extended length trades, but doesn't help him during daily trading activities. He then comes up with a modification of his model that doesn't require any historic data, but requires a lot of instantaneous compute power. He discusses this with his supervisor, who suggests using the team's extra workstations that are not in use during his trading period.

The Engineering team discusses how to modify his Grid service to be deployed on several machines at once. They propose the solution to Operations, who forwards the request to the IT staff. The IT staff and Operations team discuss how to bring the team's workstations under management of the data center's resource manager, or if it is more feasible to build a separate resource pool.

The team decides to utilize a local resource manager, able to only dedicate resources from the currency trading team's workstations. At the time the trader needs to run a simulation, the resource manager allocates available workstations, or partitions of the workstations, and invokes the deployment and activation of the service.

1.4 Involved resources

In these scenarios, the Grid system is responsible for managing a fairly large and diverse set of resources. In the data center several different hardware systems are used for storage, query,