 Improving Data Server Utilization and Management through Virtualization
This document describes the best practices for deploying the IBM DB2 Version 9 product with the IBM System p™ virtualization technology. When you run the DB2 product on the System p platform, selecting the right blend of virtualization features and their configurations to achieve desired business goals, while improving the utilization of IT resources, is a challenge. Achievable business goals are reducing administration, power, cooling, or floor space costs by consolidating data servers. Examples of ways to improve resource utilization are optimizing the performance of the DB2 product, improving processor utilization, sharing system resources, using dynamic resource allocation without rebooting, and using workload management.
This paper describes the primary System p virtualization technologies that concern the selection of the types of logical partition, disk I/O, and network interface, as well as workload management considerations. The following brief summaries describe the major considerations discussed in this paper and how they can benefit your business.
Logical partition type
Increasingly, with most hardware systems being heavily under-utilized due to sizings based on forecasted peak server activity, businesses today continually face the challenge of driving the average level of system processor utilization even higher in order to maximize their return on investment (ROI). By using shared processor partitions, businesses can efficiently consolidate multiple databases that are housed on different physical servers or dedicated partitions onto shared processor partitions on a single physical server. This sharing of processor resources, while balancing the processor requirements for both peak and average operations, reduces the total cost of ownership (TCO). One can then assign a quality of service for each of the shared processor partitions to ensure more important workloads will always get the processor resources they need while lower priority workloads will get resources on a best-effort basis. Hosting test and production applications together, on different shared processor partitions, can also help improve the quality of the test results as the test environment faithfully mimics the production environment.
Disk I/O type
With the capability to create multiple shared partitions, each with a fractional entitlement capacity, it is possible to exhaust all the physical I/O slots of the machine if a dedicated I/O slot was assigned to each logical partition. Also, in many production environments with multiple databases consolidated into multiple logical partitions, the I/O performance requirements vary quite significantly between many applications. In these cases, the virtual I/O server (VIOS) enables sharing of the disk adapter and I/O resources across multiple applications to better utilize the overall storage infrastructure and meet varied performance needs while maximizing the ROI. The VIOS feature also provides additional value-added capabilities, such as Live Partition Mobility - a feature on the POWER6™ processor family - that allows for the movement of a running partition from one POWER6 server to another with no application downtime, resulting in better system utilization, improved application availability, and energy savings.
Network type
Similar to the reasons already mentioned for the sharing of storage resources, the VIOS also handles the sharing of network adapters and, therefore, the sharing of network bandwidth across various partitions on a system. This maximizes both system resource utilization and ROI.
Workload management considerations
The types of workload management capabilities available with System p virtualization technologies are most important to businesses running customer relationship management (CRM) or transactional workloads with CPU intensive batch jobs during off-peak hours and less intensive CPU activity on the transactional system during peak business hours. These capabilities also have applicability in industries similar to retail in which typically the demand on the data server system is much higher on particular days of the year, such as Black Friday after Thanksgiving or Boxing Day after Christmas, than on other days. Efficient workload management maximizes both system resource utilization and ROI, while reducing TCO.
Test these best-practice guidelines in your test environment before implementing them in your production environment.
|
