Browse Prior Art Database

Calculate Optimal Placement of Cloud Groups into Expert Integrated Systems

IP.com Disclosure Number: IPCOM000236022D
Publication Date: 2014-Apr-02
Document File: 3 page(s) / 43K

Publishing Venue

The IP.com Prior Art Database

Abstract

Expert integrated systems are containers of different sizes and prices, and cloud groups, which is collection of compute resources, are items to be placed into these containers. Deciding what size or sizes are appropriate for given workloads is always a critical challenge. Using the method presented in this article, can help in decision making and create the placement plan to minimize cost or number of containers or room for growth along with other criteria.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 28% of the total text.

Page 01 of 3

Calculate Optimal Placement of Cloud Groups into Expert Integrated Systems
PureApplication* Systems are IBM** family of expert integrated systems that come packaged with high-end IBM hardware and software that provide automated cloud environments. Since these systems come in a variety of sizes, deciding what size or sizes are appropriate for given customer workloads is always a critical challenge; as underestimating or overestimating the required PureApplication System(s) could be very costly to both IBM and the customer.

Here is how this sizing is done, in simple terms:

First, customer workloads are analyzed and converted to a set of resource requirements (CPU, memory, disk, etc.).

Then appropriate VM sizes are determined based on user input. VM sizes help determine how many VMs are required to satisfy the given workload. Business requirements could very well make customers want to put their workloads in different cloud groups to guarantee physical and network separation. For example, it would make sense to keep production and dev/test workloads into separate cloud groups so they do not interfere with each other.

Cloud groups do not share compute nodes on expert integrated systems. In other words, each cloud group will have its own compute node or nodes. At this point in the sizing process, it is determined how many compute nodes are required by each given cloud group. The challenge is how to put these cloud groups into one or more expert integrated systems to meet the given criteria; such as minimizing the total cost, minimizing the number of remaining compute nodes, or minimizing the required floor space.

This article proposes a method for optimal placement of cloud groups into multiple expert integrated systems such that certain criteria are satisfied. The idea is to convert this cloud group placement problem into a packing problem and use heuristics that reduce the required computation for finding the optimal placement.

There are a number of cloud groups each consuming a certain number of compute nodes. These cloud groups are to be placed into a number of expert integrated systems. Expert integrated systems come in various sizes, with respect to how many compute nodes they contain, and have different price tags. The more compute nodes they have the more expensive they are. The question is how these cloud groups can be placed into expert integrated systems such that a given criterion is also met. This criterion can be


- Minimum number of expert integrated systems is used
- The total cost of expert integrated systems used is minimum
- Minimum number of empty compute nodes (ideally zero) is left after placing all cloud groups.

This problem can be converted into the mathematical problem below:


There are a number of items I1, I2, ..., Im (m >= 1) each with a certain size I1.size, I2.size, ..., Im.size, and container types C1, C2, ..., Cn (n >= 1) with sizes C1.size, C2.size, ..., Cn.size costing C1.price, C2.price, ..., Cn.price, re...