Database Research(cid:2) Achievements and Opportunities
`Into the  st Century
`
`Avi Silberschatz(cid:2) Mike Stonebraker(cid:2) Je(cid:3) Ullman(cid:2) editors
`
`Report of an NSF Workshop on the Future of Database Systems Research(cid:2) May (cid:6)(cid:2) 
`
`
`
`Introduction
`
`In February of (cid:2) a group of database researchers met to examine the prospects for future
`database research e(cid:3)orts(cid:12) The resulting report (cid:13)Silberschatz et al(cid:12) (cid:14) (cid:15)(cid:16) was instrumental in
`bringing to the public(cid:17)s attention both the signi(cid:18)cance of prior database research and the number
`of challenging and important problems that lay ahead(cid:12) We shall not repeat here the major points
`made in that report concerning the historical development of relational database systems and
`transaction management(cid:12) Rather the reader is encouraged to consult either that report or an
`on(cid:19)line document (cid:13)Gray (cid:14) (cid:15)(cid:16)(cid:2) each of which discusses these and other historical achievements of
`database research(cid:12) In May of (cid:2) a second workshop was convened to consider anew the prospects
`for database research(cid:2) and this paper reports on our (cid:18)ndings(cid:12) The following points summarize the
`conclusions of this forum(cid:12)
`
`(cid:0) The database research community plays a foundational role in creating the technological
`infrastructure from which database advancements evolve(cid:12)
`
`(cid:0) Next(cid:19)generation database applications enabled by the explosion of digitized information over
`the last (cid:18)ve years will require the solution to signi(cid:18)cant new research problems(cid:12) These
`problems are grouped in this report into the following broad areas(cid:20) support for multimedia
`objects(cid:2) distribution of information(cid:2) new database applications(cid:2) work(cid:21)ow and transaction
`management(cid:2) and ease of database management and use(cid:12)
`
`(cid:0) A new research mandate for the database community is provided by the technology devel(cid:19)
`opments of the recent past (cid:22) the explosions in hardware capability(cid:2) hardware capacity(cid:2) and
`communication (cid:13)including the internet or (cid:23)web(cid:24) and mobile communication(cid:16)(cid:12)
`
`(cid:0) There is a heightened need for governmental and industrial support of basic database research
`in order to respond to these challenges(cid:12)
`
`In the remainder of this section(cid:2) we brie(cid:21)y introduce two major themes that weave through this
`report(cid:12) First(cid:2) the demands placed on database systems are drastically changing(cid:2) thereby requiring
`new solutions to (cid:18)t the altered landscape(cid:12) Second(cid:2) the database research community has a long and
`successful track record creating new solutions and enabling the technology transfer to put these
`ideas to practical use(cid:12) Therefore(cid:2) it is a wise investment of public research monies to ensure that
`this research community remains healthy and viable(cid:12)
`
` This workshop was supported by the National Science Foundation(cid:2) Database and Expert Systems Program(cid:2) under
`the grant IRI(cid:3)  (cid:10) All opinions(cid:2) (cid:11)ndings(cid:2) conclusions and recommendations in any material resulting from this
`workshop are those of the workshop participants(cid:2) and do not necessarily re(cid:12)ect the views of the National Science
`Foundation(cid:10)
`The attendees at the workshop were(cid:13) Phil Bernstein(cid:2) Ron Brachman(cid:2) Mike Carey(cid:2) Rick Cattell(cid:2) Hector Garcia(cid:3)
`Molina(cid:2) Laura Haas(cid:2) Dave Lomet(cid:2) Dave Maier(cid:2) Je(cid:14) Naughton(cid:2) Michael Schwartz(cid:2) Pat Selinger(cid:2) Avi Silberschatz(cid:2) Mike
`Stonebraker(cid:2) Je(cid:14) Ullman(cid:2) Patrick Valduriez(cid:2) Moshe Vardi(cid:2) Jennifer Widom(cid:2) Gio Wiederhold(cid:2) Marianne Winslett(cid:2) and
`Maria Zemankova(cid:10) Comments on this document were also contributed by Jim Gray(cid:10)
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 1 of 17
`
`

`

` (cid:2) The Changing World of Database Management
`
`At its core(cid:2) a database system is a computerized record(cid:19)keeping system(cid:25) it stores and provides access
`to information(cid:12) Reduced to its basic components(cid:2) a database system consists of data(cid:2) hardware(cid:2)
`and software(cid:12) Although these simple components remain fundamental to database systems(cid:2) their
`scope(cid:2) magnitude(cid:2) and complexity have expanded mightily(cid:12)
`The last (cid:18)ve years have witnessed an unparalleled movement toward data of increasing complex(cid:19)
`ity(cid:12) The simple business(cid:19)data(cid:19)processing information expressed in numbers and character strings(cid:2)
`while still important(cid:2) has been joined by large numbers of multimedia (cid:23)documents(cid:2)(cid:24) images(cid:2) time(cid:19)
`series(cid:2) procedural or (cid:23)active(cid:24) data(cid:2) and myriad other complex data forms(cid:12) Representative examples
`of expanded data sources are found in Sections (cid:12)(cid:2) (cid:12)(cid:2) (cid:12)(cid:2) (cid:2) (cid:12) (cid:2) and (cid:12)(cid:12)
`Additionally(cid:2) low(cid:19)cost(cid:2) high(cid:19)speed hardware components such as multiprocessors based on fast
`and inexpensive microprocessors(cid:2) have become widely available(cid:12) (cid:23)O(cid:3)(cid:19)the(cid:19)shelf(cid:24) or commodity
`disks and memories show increased capacity and reduced cost each year(cid:12) The impact of hardware
`advances is reviewed in Section (cid:12) (cid:12)
`Lastly(cid:2) a new breed of sophisticated DBMSs has begun to appear to manage the demands of
`this new collection of data and to fully exploit the processing advantages of new hardware(cid:12) These
`systems are explored in Section (cid:12)
`Not only is the computing infrastructure changing(cid:2) but also the user community is undergo(cid:19)
`ing a similar revolution(cid:12) Nearly every human enterprise now includes computerized information
`processing as an integral component of its operation(cid:12) From the free(cid:19)ranging connectivity of the
`World(cid:19)Wide(cid:19)Web (cid:13)Section (cid:12)(cid:16) to the truly astronomical proportions of the Earth Observing Sys(cid:19)
`tem (cid:13)Section (cid:12) (cid:16)(cid:2) the world is on(cid:19)line and exchanging information(cid:12) User expansion is illustrated
`in Sections (cid:2) (cid:12)(cid:2) and (cid:12) (cid:12)
`
` (cid:2) The Case for DBMS Research
`
`A primary goal of this report is to provide a sound and reasoned argument that database systems
`are pivotal to current and developing information(cid:19)management needs and that(cid:2) consequently(cid:2) a
`commitment to the (cid:18)nancial support of database research is a worthwhile investment(cid:12) A second(cid:2)
`major goal of this report is to build on the report in substantiating the payo(cid:3) from funding
`basic database research(cid:12) The history of database systems research clearly illustrates the connection
`between basic research(cid:2) commercial success(cid:2) and job creation(cid:12) We observe a consistent pattern of
`theory building and derivation of working principles(cid:2) leading to experimental studies and prototype
`implementations(cid:2) which evolve(cid:2) in turn(cid:2) into commercial products(cid:12)
`We begin(cid:2) in Section (cid:2) with examination of several areas where research(cid:2) often dating back a
`decade or more(cid:2) began to in(cid:21)uence products in the (cid:18)rst half of the s(cid:12) This section demonstrates
`the continued payo(cid:3) to research funding in this area(cid:12) Then(cid:2) Section continues with a sampling
`of the new database applications that will drive future research(cid:12) We continue in Section  with
`the context in which this required research must take place(cid:12) Lastly(cid:2) Section  lays out the research
`agenda which the workshop participants felt strongly was required(cid:12) Section  closes this report
`with some (cid:18)nal thoughts(cid:12)
`
` Recent Research Achievements
`
`A number of new developments in the database marketplace took place since our report(cid:12) As
`was the case prior to (cid:2) the great majority of the companies active in these new markets are
`US(cid:19)owned companies(cid:2) and the products these companies provide have their origins in research and
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 2 of 17
`
`

`

`prototypes funded primarily by US government agencies(cid:12) In this section we brie(cid:21)y outline some of
`the key new developments(cid:12)
`
`(cid:2) Object(cid:4)Oriented and Object(cid:4)Relational Database Systems
`
`In there existed a few research prototypes of object(cid:19)oriented database systems (cid:13)OODBs(cid:16)(cid:12)
`There was considerable debate regarding the nature of such systems and the relationship between
`OODBs and relational systems(cid:12) Today there are a variety of commercial OODBs(cid:12) In aggregate(cid:2) it
`is a (cid:28)M(cid:29)year market growing at about (cid:30) per year(cid:12)
`Likewise(cid:2) in there were a few research prototypes that combined the good features of
`relational DBMS (cid:13)SQL access to simple data types(cid:16) with the good features of OODBs (cid:13)model(cid:19)
`ing of complex data(cid:16) to create new breeds of (cid:23)object(cid:19)relational(cid:24) database systems (cid:13)ORDBs(cid:16) and
`(cid:23)deductive(cid:19)object(cid:19)oriented(cid:24) database (cid:13)DOOD(cid:16) systems(cid:12) Today(cid:2) technology transfer of these ideas
`and prototypes has created new markets(cid:12) Moreover(cid:2) the traditional relational vendors are beginning
`to move their products in this direction(cid:12)
`
`(cid:2) Support for New Data Types
`
`It was recognized in the report that new applications may require data that is not necessarily
`composed of records of simple numbers and character strings(cid:12) Since then(cid:2) research on a number of
`other forms of information has found its way into products(cid:12)
`
`(cid:0) A decade or more of work on storage and retrieval of spatial data has now appeared in
`commercial GIS (cid:13)geographic information systems(cid:16)(cid:12)
`
`(cid:0) Long(cid:19)term exploration for appropriate models of temporal data has now resulted in a number
`of di(cid:3)erent proposals for extending query languages to better support temporal data(cid:12) Among
`these(cid:2) the one which currently has the widest support is the TSQL proposal(cid:2) which extends
`SQL(cid:19) (cid:12)
`
`(cid:2) Transaction Processing
`
`A core responsibility of a database management system is to support the coordination of many
`simultaneous users of shared information(cid:12) The clear de(cid:18)nition of transaction management require(cid:19)
`ments was one of the key contributions of the DBMS research community during the (cid:17)s and
`early (cid:17)s(cid:2) as noted in the report(cid:12) However(cid:2) traditional transaction management is not
`always the appropriate requirement for today(cid:17)s distributed information systems(cid:12) The study of
`policies and algorithms for supporting alternatives to atomic transactions dates back to the s(cid:12)
`Today(cid:2) we see that this long research chain is beginning to bear fruit(cid:2) such as(cid:20)
`
`(cid:0) Some commercial products support replicated data(cid:2) allowing consistent views of the same
`information to exist at various nodes of a network(cid:12)
`
`(cid:0) Some object(cid:19)oriented systems now support (cid:23)long transactions(cid:2)(cid:24) where data is (cid:23)checked out(cid:24)
`and not restored for hours or days(cid:12)
`
`(cid:0) Some commercial database systems also support (cid:23)versions and con(cid:18)gurations(cid:2)(cid:24) the ability
`to construct a history of related objects (cid:13)(cid:23)versions(cid:24) of that object(cid:16) and to combine spe(cid:19)
`ci(cid:18)c versions of di(cid:3)erent objects into (cid:23)con(cid:18)gurations(cid:12)(cid:24) These capabilities(cid:2) as well as (cid:23)long
`transactions(cid:2)(cid:24) are important support for design activities such as software engineering(cid:12)
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 3 of 17
`
`

`

` New Database Applications
`
`To understand our proposed agenda for database research(cid:2) we illustrate the next(cid:19)generation appli(cid:19)
`cations that drive requirements for new capabilities(cid:12) In this section(cid:2) we brie(cid:21)y discuss (cid:18)ve such
`applications(cid:12)
`
` (cid:2) EOSDIS
`
`The Earth Observing System (cid:13)EOS(cid:16) is a collection of satellites to be launched by NASA starting
`in (cid:25) their purpose is to gather information that will support earth scientists concerned with
`long(cid:19)term trends regarding our atmosphere(cid:2) oceans(cid:2) and land(cid:12) These satellites are to return to
`earth about (cid:29) of a petabyte (cid:13)  bytes(cid:16) of information per year(cid:12)
`This data is intended to be integrated with existing data and information from other sources
`such as foreign satellites or nonsatellite observations(cid:2) and will be stored in the EOSDIS (cid:13)EOS Data
`and Information System(cid:16)(cid:2) a database on a scale not heretofore seen(cid:12)
`EOSDIS is intended to supply the information needs of both scientists and nonscientists(cid:12) For
`example(cid:2) it is imagined that school children will be able to access EOSDIS information to see
`simulations of world weather patterns(cid:2) the e(cid:3)ect of vulcanism(cid:2) and so on(cid:12) Among the many
`challenges presented by the EOSDIS project are(cid:20)
`
`(cid:0) Providing on(cid:19)line access to petabyte(cid:19)sized databases and managing tertiary storage e(cid:3)ectively(cid:12)
`
`(cid:0) Supporting thousands of information consumers with very heavy volume of information re(cid:19)
`quests(cid:2) including ad(cid:19)hoc requests and standing orders for daily updates(cid:12)
`
`(cid:0) Providing e(cid:3)ective mechanisms for browsing and searching for the desired data(cid:2)
`
` (cid:2) Electronic Commerce
`
`There are a number of active projects attempting to make information available to support on(cid:19)
`line browsing of catalogs and subsequent electronic purchasing of goods(cid:12) The general goal is to
`allow companies to supply information about their products to on(cid:19)line customers (cid:13)perhaps using
`an electronic broker as an intermediary(cid:16)(cid:12) Brokers can aggregate data from several sources(cid:2) for
`example by collecting all electronic catalogs on wearing apparel(cid:12) In turn(cid:2) a broker can o(cid:3)er (cid:23)one
`stop shopping(cid:24) to the ultimate customer(cid:12)
`Like EOSDIS(cid:2) electronic commerce involves a very large number of participants interacting
`over a network(cid:12) While EOSDIS involves one principal supplier and many consumers(cid:2) procurement
`involves many suppliers and many consumers(cid:12) Further(cid:2) the participants are mutually suspicious
`and often have already installed proprietary information systems(cid:12) Among the challenges of this
`environment are(cid:20)
`
`(cid:0) Heterogeneous information sources must be integrated(cid:12) For example(cid:2) something called a
`(cid:23)connector(cid:24) in one catalog may not be a (cid:23)connector(cid:24) in a di(cid:3)erent catalog(cid:12) Such (cid:23)schema
`integration(cid:24) is a well(cid:19)known and extremely di cult problem(cid:12)
`
`(cid:0) Electronic commerce needs reliable(cid:2) distributed authentication and funds transfer(cid:12)
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 4 of 17
`
`

`

` (cid:2) Health(cid:4)Care Information Systems
`
`Physicians need to draw upon many di(cid:3)erent kinds of information in the course of their work(cid:12) For
`example(cid:2) medical records about a single patient may exist at many di(cid:3)erent hospitals(cid:2) medical
`o ces(cid:2) and insurance o ces(cid:12) A history of the patient must be gathered from these(cid:12) Information
`about procedures(cid:2) drugs(cid:2) diagnostic tools(cid:2) and other aids to treating patients are in principle
`available through multiple systems and databases(cid:12)
`In addition(cid:2) the notes of physicians on a patient(cid:2) records of diagnostic tests(cid:2) and insurance and
`billing information(cid:2) can be captured electronically and made available for later use(cid:12) Transforming
`the health(cid:19)care industry to take advantage of what is now possible will have a major impact on
`costs(cid:2) and possibly on quality and ubiquity of care as well(cid:12) But there are problems to be solved
`before we can reach that point(cid:2) such as(cid:20)
`
`(cid:0) Integration of heterogeneous forms of legacy information(cid:12)
`
`(cid:0) Access control to preserve the con(cid:18)dentiality of medical records(cid:12)
`
`(cid:0) Interfaces to information that are appropriate for use by all health(cid:19)care professionals(cid:12)
`
` (cid:2) Digital Publishing
`
`The publishing industry(cid:2) like health(cid:19)care(cid:2) will undergo profound changes in business practices over
`the next few years(cid:12) It is becoming possible to store books and articles electronically and deliver
`them over high(cid:19)speed networks to consumers(cid:12) Further(cid:2) the notion of a publishable document is
`broadening to include audio and video(cid:2) graphic images(cid:2) lectures(cid:2) annotations(cid:2) and other elements
`that convey information(cid:12) The total amount of information available to be put on line dwarfs the
`petabyte(cid:19)sized EOSDIS database(cid:2) and the amount that is likely to become available in the near
`future is an order of magnitude greater(cid:12)
`A corollary of the changes to come in publishing is that the education industry draws much
`closer to publishing(cid:12)
`Instead of live lectures to small groups of students(cid:2) educational products
`that are part text(cid:2) part video(cid:19)lecture(cid:2) and part interactive training can serve much of the needs of
`thousands of students(cid:12) These prospects motivate the following research directions(cid:20)
`
`(cid:0) Management and delivery of extremely large bodies of data at very high rates(cid:12) Typical data
`consists of very large objects (cid:22) in the megabyte to gigabyte range (cid:22) and may require delivery
`with real(cid:19)time constraints(cid:12)
`
`(cid:0) Protection of intellectual property(cid:2) including cost(cid:19)e(cid:3)ective collection of small payments and
`inhibitions against reselling of information(cid:12)
`
`(cid:0) Organization of and access to overwhelming amounts of information(cid:12)
`
` (cid:2) Collaborative Design
`
`Large projects such as aircraft are now being designed and built by consortia of independent com(cid:19)
`panies(cid:12)
`Information regarding such projects often has a multidecade lifetime(cid:2) supporting both
`maintenance and subsequent modi(cid:18)cations(cid:12) Designs may be simulated on a computer (cid:22) for per(cid:19)
`formance(cid:2) feasibility of assembly(cid:2) and correctness (cid:22) before becoming physical reality(cid:12) Designs
`evolve both before and after (cid:18)rst manufacture(cid:2) causing a proliferation of current(cid:2) tentative(cid:2) and
`historical con(cid:18)gurations of related information(cid:12) Di(cid:3)erent design disciplines tend to use distinct
`design tools with di(cid:3)erent underlying models and notations(cid:12) Moreover(cid:2) designs often last longer
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 5 of 17
`
`

`

`than the tools that produced them(cid:2) so di(cid:3)erent components of one design may end up having been
`developed by di(cid:3)erent versions of the same tool(cid:12) We are thus led to consider problems such as(cid:20)
`
`(cid:0) As with several of the other paradigm problems suggested above(cid:2) we must deal with integra(cid:19)
`tion of varied sources of information(cid:2) including legacy sources(cid:12)
`
`(cid:0) Collaborative design requires support for new forms of concurrency control and sharing mech(cid:19)
`anisms in the underlying databases(cid:12)
`
`(cid:0) Interacting collections of processes(cid:2) such as design and simulation(cid:2) require (cid:23)work(cid:21)ow(cid:24) man(cid:19)
`agement in which long(cid:19)term transactions interact in sound ways(cid:12)
`
`(cid:0) Support for versions of the design for a single component and support for con(cid:18)gurations
`combining versions of many components is essential(cid:25) see Section (cid:12) (cid:12) (cid:12)
`
` Trends That A(cid:5)ect Database Research
`
`Before presenting our agenda for database research(cid:2) we present some observations about trends
`that impact this research(cid:12) In this section(cid:2) we discuss in turn hardware trends(cid:2) the directions of the
`DBMS vendors(cid:2) research and the business climate(cid:2) and the World(cid:19)Wide(cid:19)Web(cid:12)
`
`(cid:2) Technology Trends
`
`For (cid:18)fty years there has been an exponential improvement in a number of parameters that measure
`our ability to compute(cid:12) Each of the following have improved by a factor of or more every ten
`years(cid:12)
`
` (cid:12) The number of machine instructions executable in a second(cid:12)
`
`(cid:12) The cost of a typical processor(cid:12)
`
` (cid:12) The amount of secondary storage per unit cost(cid:12)
`
`(cid:12) The amount of main memory per unit cost(cid:12)
`
`These improvements in price(cid:29)performance of critical components make it possible(cid:2) each few years(cid:2)
`to provide solutions to new classes of problems and to create new kinds of products and services
`that were previously beyond our reach(cid:12)
`We expect these trends to continue unabated into the next millenium(cid:12) Moreover(cid:2) in the last few
`years(cid:2) these trends have been joined on their upward spiral by two new parameters that heretofore
`had improved(cid:2) but not as explosively as the others(cid:2) namely(cid:20)
`
`(cid:0) The number of bits transmitted per unit cost(cid:12)
`
`(cid:0) The number of bits transmitted in a second(cid:12)
`
`These new trends result in an environment where it is possible to deal with terabytes of data and
`complex queries in a cost e(cid:3)ective manner(cid:12)
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 6 of 17
`
`

`

`(cid:2) Database Architectural Trends
`
`While less profound than the exponential growth of computing and communication(cid:2) there have
`been a number of important changes in the way databases are structured and used over the past
`(cid:18)ve years(cid:12) We note the following(cid:20)
`
`(cid:0) While relational systems were well on their way to supplanting earlier approaches in (cid:2) we
`now observe that the relational approach is today ubiquitous(cid:12) Relational databases are today
`used in applications ranging from those running on top of very large parallel architectures
`(cid:13)e(cid:12)g(cid:12)(cid:2) NCR (cid:16) to those running on home computers(cid:12)
`
`(cid:0) Client(cid:19)server architectures have migrated from (cid:18)le systems to database systems over the last
`(cid:18)ve years(cid:12) We expect it will become progressively more common for database servers to be
`accessed remotely over networks(cid:12)
`
`(cid:0) The traditional(cid:2) record(cid:19)based data that populated relational databases of (cid:18)ve years ago has
`been joined by various kinds of (cid:23)multimedia(cid:24) data(cid:12) This trend is fueling the success of ORDB
`and is causing relational vendors to dramatically enhance their engines to deal with a richer
`data model(cid:12) In short(cid:2) systems that are merely relational DBMSs will be the legacy systems
`of the next decade(cid:12)
`
`(cid:2) Research and the Business Climate
`
`The economics of research and development has changed signi(cid:18)cantly in the past (cid:18)ve years(cid:12) In
`general(cid:2) corporations that traditionally supported substantial basic research have had to cut back(cid:2)
`as pro(cid:18)t margins on many product and service lines have shrunk(cid:12) Their research has been reoriented
`toward short(cid:19)term projects intended not as prototypes but as frontier work intended directly for
`market(cid:12) Moreover(cid:2) government research agencies have in the recent past felt a similar need to
`reorient from long(cid:19)term or basic research to short(cid:19)term(cid:2) closely managed projects(cid:12)
`However(cid:2) there are indications that database research is regarded positively by government and
`industry(cid:12) For example(cid:2) Toole and Young (cid:14) (cid:15) gives a prominent place to information systems in
`future strategic research(cid:2) unlike some earlier documents of this type(cid:12) The growth of information(cid:19)
`intensive industries and the importance of increasing the e ciency of business practices have each
`put database technologies at or near the top of corporate concerns(cid:12)
`The trend in industrial database research(cid:2) while mixed(cid:2) has many positive elements(cid:12) Companies
`that have had to restructure their research enterprise have frequently maintained or increased
`their database component(cid:12) While many of the companies that are most closely identi(cid:18)ed with
`database systems as products have remained substantially outside the research community(cid:2) some
`other companies that (cid:18)ve years ago did not regard database systems as central to their mission
`have moved aggressively to build a database research component(cid:12)
`
`(cid:2) The Information Superhighway Just Rolled Through Your Living Room
`
`No futuristic discussion can be called complete without a discussion of the World(cid:19)Wide Web (cid:13)WWW
`or Web for short(cid:16)(cid:12) While people were debating the nature of the (cid:23)Information Superhighway(cid:24) or
`(cid:23)National Information Infrastructure(cid:2)(cid:24) the Web (cid:22) an informal collection of connected (cid:23)documents(cid:24)
`based on HTML (cid:13)Hypertext Markup Language(cid:16) (cid:22) started growing at an astronomical rate(cid:12) Re(cid:19)
`cently(cid:2) the number of Web bits carried by the Internet has grown (cid:6)(cid:30) per month(cid:2) or a factor
`of growth per year(cid:12)
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 7 of 17
`
`

`

`As a result(cid:2) the number of active users of the Internet has passed (cid:30) of the US population by
`some estimates(cid:12) Following similar technology(cid:19)based trends such as VCRs or audio CDs(cid:2) we expect
`that internet access will reach the great majority of the population just a few years after breaking
`the (cid:30) barrier(cid:12) As a consequence(cid:2) we expect over the next few years several orders of magnitude
`of growth in the availability and use of information over the Internet(cid:12) We expect that the provision
`and use of such information will become a concern of every individual(cid:12)
`Databases and database technology will play a critical role in this information explosion(cid:12) Al(cid:19)
`ready Webmasters (cid:13)administrators of World(cid:19)Wide(cid:19)Web sites(cid:16) are realizing that they are database
`administrators(cid:2) with a di(cid:3)erent title(cid:12) Many large Web sites are turning to DBMS technology to
`keep track of the ever increasing numbers of stored objects(cid:12) In addition(cid:2) innovative sites are al(cid:19)
`ready prototyping traditional DBMS applications such as electronic catalogs using the Web as an
`infrastructure(cid:12)
`
` New Research Directions
`
`Against this backdrop(cid:2) our workshop participants suggested the following research agenda as the
`most important way to enable the new generation of applications mentioned in Section (cid:12) Indicated
`research directions are grouped into (cid:18)ve major categories(cid:20)
`
`(cid:0) Problems associated with putting multimedia objects into DBMSs(cid:12)
`
`(cid:0) Problems involving new paradigms for distribution of information(cid:12)
`
`(cid:0) New uses of databases(cid:12)
`
`(cid:0) New transaction models(cid:12)
`
`(cid:0) Problems involving ease of use and management of databases(cid:12)
`
`(cid:2) Support for Multimedia Objects
`
`The explosive growth of the Web(cid:2) as well as the challenges of EOSDIS(cid:2) electronic commerce(cid:2) and
`digital publishing(cid:2) creates a collection of challenges which must be addressed for future database
`systems(cid:12) Below are the principal areas for research involving multimedia data(cid:12)
`
`(cid:2) (cid:2) Tertiary Storage
`
`Indisputably(cid:2) multimedia data is very large(cid:2) and its very bulk presents us with new challenges(cid:12) As
`mentioned in Section (cid:2) new applications like EOSDIS or electronic libraries involve data volumes
`in the petabyte range(cid:12) Despite the exponential growth in disk sizes(cid:2) these volumes are unlikely to
`be maintainable solely on magnetic or magneto(cid:19)optical disk in the foreseeable future(cid:12) We are thus
`presented with problems of managing a new level in the storage hierarchy(cid:2) called tertiary storage(cid:12)
`This third level uses storage devices that are orders of magnitude slower than (cid:23)secondary storage(cid:24)
`(cid:13)disks(cid:16)(cid:2) yet also of vastly greater capacity(cid:12) Tertiary storage devices include compact(cid:19)disk juke boxes
`or tape silos(cid:2) and typically use a mechanical arm to physically move the desired tape cassette or
`CD to a reader(cid:12)
`In a sense(cid:2) access to tertiary storage is by bu(cid:3)ering selected data items on secondary storage(cid:2)
`just as access to secondary storage is by bu(cid:3)ering selected data into main memory from disk(cid:12)
`However(cid:2) the raw numbers and the proportions make the optimization of the tertiary(cid:19)to(cid:19)secondary
`transition rather di(cid:3)erent from the secondary(cid:19)to(cid:19)main transition(cid:12) As just one example(cid:2) today much
`
`
`
`Enfish, LLC; IPR2014-00574
`Exhibit 2226
`Page 8 of 17
`
`

`

`tertiary information is found on tape cassettes(cid:12) Not only are tape cassettes slower to access than
`blocks on a disk by three orders of magnitude (cid:13)seconds vs(cid:12) milliseconds(cid:16)(cid:2) but (cid:18)nding data not at
`the beginning of a tape cassette can increase the access time by (cid:6) orders of magnitude beyond
`that(cid:2) while access to the middle of a disk block does not increase access time substantially(cid:12) Thus(cid:2)
`the question of where in a tape cassette information is placed is signi(cid:18)cant today(cid:2) while the position
`of data within a disk block or track is relatively unimportant(cid:12)
`
`(cid:2) (cid:2) New Data Types
`
`Each form of multimedia information (cid:13)data type(cid:16) requires its own collection of (cid:18)rst(cid:19)class concepts
`(cid:13)operations and functions(cid:16)(cid:2) along with a high performance implementation involving appropriate
`data structures and access methods(cid:12) As a simple example of the challenge(cid:2) a recent experiment
`benchmarking current object(cid:19)oriented systems noted a vast di(cid:3)erence in the performance of such
`systems on a large text object(cid:12) When asked to (cid:18)nd the last character of a megabyte(cid:19)long text
`string(cid:2) some systems retrieved the whole string and then applied the (cid:23)(cid:18)nd last(cid:24) operation(cid:2) while
`others were able to retrieve only the last byte or some short tail of the string(cid:12) We need to think
`carefully about(cid:20)
`
`(cid:0) The operations available for each type of multimedia data(cid:2) and the resulting implementation
`tradeo(cid:3)s(cid:12)
`
`(cid:0) The integration of data involving several of these new types(cid:12)
`
`(cid:2) (cid:2) Quality of Service
`
`In general(cid:2)
`Delivering multimedia data to many users presents several new research problems(cid:12)
`when data is large(cid:2) access and delivery can easily be a bottleneck(cid:12) However(cid:2) large data objects are
`frequently accessed in a very predictable fashion(cid:12) For instance(cid:2) a video server delivering movies to
`many households at once can assume that each request will remain in force and must be delivered
`at the standard rate until a (cid:23)stop(cid:24) button is pushed(cid:12) Thus(cid:2) there are opportunities to optimize
`access based on predicted use(cid:2) and those predictions will normally be quite accurate(cid:12)
`Moreover(cid:2) some multimedia information comes with severe delivery constraints(cid:12) For example(cid:2)
`video generally must be delivered at a (cid:18)xed rate(cid:2) and there is a penalty for late delivery(cid:2) since the
`movie