Object-oriented framework for generic adaptive control

Abstract

A system and method are described for constructing and implementing generic software agents for automated tuning of computer systems and applications. The framework defines the modules and interfaces to allow agents to be created in a modular fashion. The specifics of the target system are captured by adaptors that provide a uniform interface to the target system. Data in the agent is managed by a metric manager, and controller modules implement the desired control algorithms. The modular structure and common interfaces allow for the construction of generic agents that are applicable to a wide variety of target systems, and can use a wide variety of control algorithms.

1 . A computer-based tuning system for automatically tuning one or more target systems, comprising: 
 a metric manager for managing at least one set of metrics corresponding to the one or more target systems;    one or more controllers for implementing one or more control strategies for adaptively tuning performance characteristics of the one or more target systems based upon the at least one set of metrics, wherein the one or more control strategies are independent of a particular architecture of any of the one or more target systems, and    one or more adaptors providing an abstract interface to the one or more target systems with respect to the one or more control strategies, wherein at least one of the one or more adaptors is specific to a corresponding one of the one or more target systems,    wherein the one or more controllers invoke the one or more adaptors (i) to obtain metric values from the one or more target systems to compute tuning control values, and (ii) to tune the one or more target systems using the computed tuning control values.    
   
   
       2 . The tuning system of  claim 1 , wherein each of the one or more control strategies corresponds to a separate computer program.  
   
   
       3 . The tuning system of  claim 1 , wherein said one or more controllers comprise a master controller for resolving conflicts between the one or more control strategies.  
   
   
       4 . The tuning system of  claim 1 , wherein the set of metrics comprise read-only metrics and read/write metrics.  
   
   
       5 . The tuning system of  claim 4 , wherein said one or more adaptors directly obtain latest values of the read-only metrics from the one or more target systems.  
   
   
       6 . The tuning system of  claim 4 , wherein said one or more adaptors set values corresponding to the read/write metrics on the one or more target systems.  
   
   
       7 . The tuning system of  claim 1 , wherein said metric manager has a capability of adding, deleting, and listing the at least one set of metrics.  
   
   
       8 . The tuning system of  claim 1 , wherein the tuning system has a capability of invoking other tuning systems to form a hierarchical tuning system with respect to the one or more target systems.  
   
   
       9 . The tuning system of  claim 8 , wherein the tuning system and the other tuning systems operate cooperatively to implement the one or more control strategies.  
   
   
       10 . The tuning system of  claim 1 , further comprising an administrator application programming interface (API) for specifying service-level requirements of the one or more target systems.  
   
   
       11 . The tuning system of  claim 1 , further comprising an administrator application programming interface (API) for monitoring an operation of the tuning system.  
   
   
       12 . The tuning system of  claim 1 , wherein said metric manager is capable of receiving an input specifying at least a subset of metrics to be stored from among the at least one set of metrics.  
   
   
       13 . The tuning system of  claim 1 , wherein more than one of the one or more controllers is employed in a given application of the tuning system to a given one of the one or more target systems.  
   
   
       14 . The tuning system of  claim 1 , wherein at least some of the one or more controllers are modular and have a capability of being deleted from the tuning system, modified, or replaced.  
   
   
       15 . The tuning system of  claim 1 , wherein at least some of the one or more adaptors are modular and have a capability of being deleted from the tuning system, modified, or replaced.  
   
   
       16 . The tuning system of  claim 1 , further comprising a shared facility for logging metric changes.  
   
   
       17 . A computer-implemented method for automatically tuning one or more target systems, comprising the steps of: 
 managing at least one set of metrics corresponding to the one or more target systems;    providing one or more controllers for implementing one or more control strategies for adaptively tuning performance characteristics of the one or more target systems based upon the at least one set of metrics, wherein the one or more control strategies are independent of a particular architecture of any of the one or more target systems;    providing one or more adaptors providing an abstract interface to the one or more target systems with respect to the one or more control strategies, wherein at least one of the one or more adaptors is specific to a corresponding one of the one or more target systems; and    automatically invoking the one or more adaptors (i) to obtain metric values from the one or more target systems to compute tuning control values, and (ii) to tune the one or more target systems using the computed tuning control values.    
   
   
       18 . A computer-based tuning system for automatically tuning one or more target systems, comprising: 
 a metric manager for managing at least one set of metrics corresponding to the one or more target systems;    one or more controllers for implementing one or more control strategies for adaptively tuning performance characteristics of the one or more target systems based upon the at least one set of metrics, wherein the one or more control strategies are independent of a particular architecture of any of the-one or more target systems; and    one or more adaptors for abstractly interfacing with the one or more target systems with respect to the one or more control strategies, wherein at least one of the one or more adaptors is specific to a corresponding one of the one or more target systems;    wherein the one or more control strategies invoke the one or more adaptors to adaptively tune the performance characteristics of the one or more target systems to set control parameters for adjusting resource allocations of the one or more target systems.