Raga Gopalakrishnan University of Colorado at Boulder Adam Wierman (Caltech) Amy R. Ward (USC) Sherwin Doroudi (CMU) Staffing and Routing to incentivize servers in many-server systems
strategic servers system performance Journal reviews Call centers Crowdsourcing Cloud computing Enterprise data centers … service systems
strategic servers system performance Journal reviews Call centers Crowdsourcing Cloud computing Enterprise data centers … service systems This talk: Impact of strategic servers on optimal system design Classic Queueing: Assumes fixed (arrival and) service rates, fixed control/policies. Queueing games: Strategic arrivals Service/price competition [Hassin and Haviv 2003] to incentivize strategic servers Routing Staffing
Outline The M/M/1 queue – a simple example Model for a strategic server The strategic M/M/N queue Classic policies in non-strategic setting Impact of strategic servers Asymptotically optimal policies Routing Staffing which idle server gets the next job? how many servers to hire?
M/M/1/FCFS strategic server idleness cost utility function LHS RHS
Outline The M/M/1 queue – a simple example Model for a strategic server The strategic M/M/N queue Classic policies in non-strategic setting Impact of strategic servers Asymptotically optimal policies Routing Staffing which idle server gets the next job? how many servers to hire?
M/M/N/FCFS strategic servers routing symmetric Nash equilibrium Nash equilibrium existence? performance? Blue for strategic service rates Yellow for control/policy parameters
Outline The M/M/1 queue – a simple example Model for a strategic server The strategic M/M/N queue Classic policies in non-strategic setting Impact of strategic servers Asymptotically optimal policies Routing Staffing which idle server gets the next job? how many servers to hire?
Rate-based policies Idle-time-based policies FSF SSF LISF SISF Random Common routing policies
M/M/N/FCFS routing Rate-based policies Idle-time-based policies FSF SSF LISF SISF Random When servers are not strategic…
M/M/N/FCFS routing Rate-based policies Idle-time-based policies FSF SSF LISF SISF Random When servers are not strategic… well-studied, asymptotically optimal [Lin and Kumar 1984] [de Véricourt et al. 2005] [Armony 2005]
M/M/N/FCFS routing Rate-based policies Idle-time-based policies FSF SSF LISF SISF Random When servers are not strategic… asymptotically “fair” [Atar 2008]
Rate-based policies Idle-time-based policies FSF SSF LISF SISF Random When servers are strategic… M/M/N/FCFS routing
Rate-based policies Idle-time-based policies FSF SSF LISF SISF Random When servers are strategic… Our results… M/M/N/FCFS routing
Rate-based policies FSF SSF Random & Idle-time-based policies When servers are strategic… Our results… M/M/N/FCFS routing First order condition: same unique symmetric equilibrium
Rate-based policies FSF SSF Random & Idle-time-based policies When servers are strategic… Our results… M/M/N/FCFS routing Can we do better than Random? Yes, but…
Outline The M/M/1 queue – a simple example Model for a strategic server The strategic M/M/N queue Classic policies in non-strategic setting Impact of strategic servers Asymptotically optimal policies Routing Staffing which idle server gets the next job? how many servers to hire?
M/M/N/FCFS When servers are not strategic… Random staffing asymptotically optimal [Borst et al. 2004]
M/M/N/FCFS Random staffing When servers are strategic… Problem: Explicit expression unknown!
M/M/N/FCFS Random staffing When servers are strategic…
Rate-independent Rate-dependent
Rate-independent Rate-dependent Limiting FOC: Policy:
Rate-independent Rate-dependent Limiting FOC: Policy:
Rate-independent Rate-dependent Limiting FOC: Policy:
Rate-independent Rate-dependent Limiting FOC: Policy:
Rate-independent Rate-dependent Limiting FOC: Policy:
Concluding remarks We need to rethink optimal system design when servers are strategic! Joint routing-staffing optimization? Empirical studies / Experimental evaluation? Asymmetric models / equilibria? Interaction between strategic arrivals and strategic servers? M/M/N/FCFS Random loss of efficiency? $ $$$$ $$ ? ?
Ragavendran Gopalakrishnan University of Colorado at Boulder Adam Wierman (Caltech) Amy R. Ward (USC) Sherwin Doroudi (CMU) Staffing and Routing to incentivize servers in many-server systems