Sekitei 1&2: Construction of deployment plans

Component Placement Problem (CPP) is about run-time construction and placement of component-based application on a wide-area network. Sekitei is our planner for the CPP. The algorithm gets descriptions of components and network resources and produces a deployment plan. Internally, Sekitei is an AI planner with metric resources. Sekitei supports arbitrary monotonic functions. Here are more details.

This is the first paper that introduced the problem and described the very basic algorithm.

T.Kichkaylo, A.Ivan, V.Karamcheti
Constrained Component Deployment in Wide-Area Networks Using AI Planning Techniques,
IPDPS 2003

This 5-page write-up describes some performance optimizations.

Planning with Arbitrary Monotonic Resource Functions,

This technical report summarizes the two above papers and provides evaluation.

T.Kichkaylo, A.Ivan, V.Karamcheti
Sekitei: An AI planner for Constrained Component Deployment in Wide-Area Networks,
NYU CS Technical Report TR2004-851, March 2004
TR page

Sekitei 3: Optimization of deployment plans

The next generation of the planner for the CPP introduces the notion of cost and allows for resource optimization. The extension of the model and the new algorithm are described here:

T.Kichkaylo, V.Karamcheti
Optimal Resource-Aware Deployment Planning for Component-based Distributed Applications,
HPDC 2004.

Planning for grids

I believe that application construction for snapshot applications and grids requires very similar sort of planning. Here is a 5-page write-up on very preliminary ideas on how to plan for grids.

Planning for Grid Applications with Explicit Reservations,
To appear in DC ICAPS 2004.

Application configuration in general

The following paper discusses commonalities between different instances of the application configuration problem and AI planning. In particular, it shows that the basic CPP (see Sekitei above) is PSPACE-hard.

Timeless Planning and the Component Placement Problem,
To appear in ICAPS 2004 Workshop on Planning and Scheduling for Web and Grid Services.

Information integration using Semantic Web Services

We also looked into planning for web services. Here is a piece of that work.

T.Kichkaylo, V.Karamcheti
Using Semantic Services to Answer Queries with Quantified Incompleteness,
AAAI 2004 Spring Symposium on Semantic Web Services

On-the-fly construction of user interfaces

During my internship in IBM in 2000 I was working on on-the-fly construction of user interfaces. (Yes, construction and on-the-fly are constants in my work, the rest changes over time.) The following paper describes that work.

L.Bergman, T.Kichkaylo, G.Banavar, J.Sussman
Pervasive Application Development and the WYSIWYG Pitfall,
EHCI 2001