Autonomic ComputingMajor vendors are applying decision-support techniques to service-centric computingby Seth Grimes Continued from Page 1 In the case of SQL Server 2000, the wizard works with sets of materialized views — relational database views join and restrict data from multiple underlying tables, and materialized views turn these into actual, rather than just virtual, tables — and indexes. The wizard simulates the effect of creating additional materialized views and indexes, invoking the query optimizer, which figures out the best way to handle a query, to help in the evaluation. IBM, Oracle, and other vendors have added similar self-tuning and self-management features to their competing products. But I'll note that five out of five feature articles in the Quarter 3, 2002 issue of DB2 Magazine are on tuning and design for performance. I wonder when any of these projects will really deliver? Dealing With ComplexityYet Chaudhuri and Weikum stated in the cited paper that "the key for simplification lies in substantially limiting [SQL's] functionality and expressiveness," which obviously hasn't occurred. Instead, recent SQL versions have added roll-up (data summarization) and other analytic functions to the mix. Chaudhuri and Weikum also stated, "there should be no dynamic resource sharing among components," yet Microsoft's Internet Explorer browser and Internet Information Server are notorious for their security holes, many introduced by Microsoft's component strategy. And of course the Windows registry is a prime example of problem-prone, dynamic resource sharing among components. The road ahead may have been mapped, but it's full of potholes. DiversityMicrosoft's focus appears to be primarily on application software and less on the operating system, hardware, and network, unlike such efforts as IBM's eLiza, Hewlett-Packard's Utility Data Center, and Sun's nascent N1, which aim to manage increasingly complex, networked operating environments. I'll cite just a few other examples, first Rendezvous, technology from Apple for networking computers and devices that take over once physically connected, "automatically broadcasting and discovering what services each is offering for the use of others." Chaudhuri pointed me to one work he particularly likes, the Berkeley-Stanford Recovery-Oriented Computing (ROC) project, which views hardware, software, and human failure as unavoidable and unpredictable and seeks to create novel approaches to system recovery. Key ROC principles include:
The diversity of these projects underscores that we work in a heterogeneous computing world. In this larger world where the true Grand Challenge resides, the networking and distributed computing and industrial competencies of companies such as Sun Microsystems and IBM come into play. Although Microsoft's Chaudhuri, in an email to me, welcomed IBM's autonomic-computing conversion — IBM Research director Paul Horn first articulated his company's new vision only in March 2001 — I see IBM's ability to execute far ahead of any other vendor's. Regardless, the larger industry-academic-user community can all contribute because so much of the new architecture is based on open standards and published interfaces. Notably, Web services provide the universal glue that binds service-centric computing. I had been puzzling for a couple of years whether Web services would ever do more than simply replace proprietary protocols in select cases, for instance, by more easily providing access to servers for a range of heterogeneous clients from PDAs to large servers. In researching a first look at XML for Analysis last year, it seems that extended access was the sole ambition of analytic-tool vendors. There are variations on the theme: Sun Microsystems' Jini network technology offers a compelling alternative for creating "network-centric services — whether implemented in hardware or software — that are highly adaptive to change,... scalable, evolvable, and flexible as typically required in dynamic computing environments." Jini provides Java code mobility, that is, Java-object deployment to networked machines, and enables self-configuring and self-healing and componentized service networks, a key autonomic-computing goal. Given the common goal, I'm confident that the diverse approaches will evolve into compatibility. Any approach that doesn't will fall by the wayside. Where's BI?You haven't read the terms OLAP, or business intelligence (BI), or analytic applications before now in this column. Query and reporting tools, "online" slice-and-dice analysis, key performance indicators, and dashboard displays all can and will benefit by the access and integration possibilities provided by Web service interfaces, but I've seen little evidence that vendors are implementing self-tuning and self-management features in what are, for the most part, highly focused packages. And in researching an earlier column on grid computing, I couldn't find a single mainstream analytics vendor that was working to enable products for parallelized, distributed processing. This point helps clarify the difference between BI and decision support. The latter encompasses techniques that may be applied in a spectrum of applications, in BI and also in embedded form in the guts of database and other servers, in operating systems and storage systems, and in computing grids and machines that regulate distributed, service-centric architectures. The techniques draw from control theory, pattern recognition, predictive analysis, simulation, and optimization, techniques that are, for the most part, only now entering the BI milieu. The KeyAutonomic computing will call on the range of analytic techniques to realize the ambitious goal of creating dynamic, service-centric networks of self-managing computing resources. And at every level, from software subsystem to operating platform to Web-service network, a decision-support framework emphasizing measurability, analysis, and action will provide the key. Seth Grimes [grimes@altaplana.com] is a principal of Alta Plana Corp., a Washington, D.C.-based consultancy specializing in analytic computing systems and demographic and economic statistics. RESOURCES"AutoAdmin: Self-Tuning and Self-Administering Databases:" research.microsoft.com/dmx/autoadmin/default.asp IBM Autonomic Computing: researchweb.watson.ibm.com/autonomic Jini Network Technology: wwws.sun.com/software/jini "Recovery-Oriented Computing:" roc.cs.berkeley.edu/roc_overview.html Rendezvous: www.apple.com/macosx/jaguar/rendezvous.html "Rethinking Database System Architecture: Towards a Self-Tuning RISC-Style Database System:" ftp.research.microsoft.com/users/AutoAdmin/RISC-vldb00.pdf "Self-Tuning Systems Software:" www.barrera.org/selftune/selftune.htm
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| |||||||||||||||||||||||||||||||
