Optimizing software packages for application management

Application lifecycle management (ALM) provides a wholistic approach to software products, from product requirements through operations. The advent of software product lines (SPLs) such as Microsoftpsilas Visual Studio creates challenges for ALM since product codes are combined in various ways to produce different customer releases or SKUs. This paper proposes an approach to making packaging decisions for SPLs. We begin by showing how industry best practices for application management as described in the IT Infrastructure Library (ITIL) should be extended to include packaging. Next, we develop a cost model and several algorithms that quantify the trade-offs between vendor costs and customer costs. Ideally, we want a packaging scheme that has the smallest customer cost at a given vendor cost. Unfortunately, finding this optimal trade-off by exhaustive search has computational complexity greater than O(n
²
2
ⁿ
), where n is the number of SKUs. We develop a ldquogreedyrdquo algorithm that iteratively selects SKUs to place in the same package so that the smallest packaging overheads result. This approach has complexity O(n
⁴
). Our studies of Visual Studio SKUs show that the trade-offs produced by the greedy algorithm converge rapidly to the optimal trade-offs as we increase the number of packages in a packaging scheme.