The Evolution of Web-Caching Markets by liwenting


									    INTERNET WATCH

                                                                                               relationship with its suppliers may sat-

    The Evolution                                                                              isfy these requirements, but for making
                                                                                               bread, only the technical specifications
                                                                                               are a concern.
                                                                                                  Historically, base goods are introduced

    of Web-Caching                                                                             and then standardized grades are estab-
                                                                                               lished to let markets, exchange trading,
                                                                                               and derivatives develop. Prices also shift
                                                                                               from cost to market based. A market

    Markets                                                                                    consists of participants who have access
                                                                                               to price information about the goods
                                                                                               they buy and sell. An exchange is a for-
                                                                                               mal marketplace that often provides
    Chris Kenyon, IBM Zürich Research Laboratory                                               extra services connected with the risks of
                                                                                               contract default or noncompliance.
                                                                                               Goods can exist as a commodity regard-

         nternet businesses lose billions of
         dollars each year because of slow
         and failed Web services, and new
                                                                                  Web-caching companies must
         companies and coalitions have                                            outsource, aggregate, and
         rushed to provide the worldwide
    content caching that can help recapture                                       customize to thrive in the face
    this lost revenue. Although the media
                                                                                  of online commoditization.
    has often characterized these companies
    as trailblazers in the e-commerce revo-
    lution, they are merely exploiting a tem-
    porary niche along the path to Internet       clients, the first of which was Yahoo.       less of whether or not a formal, orga-
    infrastructure commoditization and dis-       Recently acquired by Cable & Wireless,       nized market or trading exchange exists.
    tributed e-markets. The current players       Digital Island provides Web site and            Forwards and futures derivatives
    have a chance to succeed, but the future      application hosting via a private network    address the specific needs of consumers
    belongs to innovators in the distributed      that covers 35 countries with 2,550          or producers. A forward contract is an
    storage and compute online commodi-           enterprise servers, a 70 percent increase    agreement to buy or sell a specific
    ties markets.                                 in servers from a year ago.                  amount of a commodity at a given time
                                                     Distance still matters in the Infor-      in the future for an agreed price. A
    WEB CACHING                                   mation Age. Geographically distributed       futures contract is a standardized for-
       Web caching—the temporary storage          caching servers exist because the Internet   ward contract traded on an exchange.
    of Web objects for later retrieval—makes      alone cannot satisfy the desire to provide   Derivatives (options) on futures address
    using the Web less expensive and              content quickly and economically over        uncertainties about actual supply or
    improves performance because it reduces       large geographic areas. Web-caching          demand compared with expectations.
    bandwidth consumption, server load,           companies have succeeded by providing           Goods that have become exchange-
    and latency.                                  this service. However, the conditions that   traded-commodities include oil (1983),
       Akamai and Digital Island, the two ear-    led to the formation of these companies      electricity (1999), and—most recently—
    liest Web-caching companies, represent        are rapidly changing, and now they must      bandwidth (1998-2001). Technical and
    significant operations aimed at control-       adapt their business methods to compete      geographical variations are not a barrier
    ling a vital choke point in the Internet      successfully in the online storage and       to commoditization. Oil is produced at
    market. Even after the dot-com correction     computing commodities markets.               thousands of places around the world,
    in 2000, both generated record-breaking                                                    each with its own unique composition.
    IPOs and billion-dollar valuations. The       COMMODITIES EVOLUTION                        However, oil is graded and priced accord-
    companies’ growth and profitability con-         A commodity is a widely used good          ing to its departure in composition from
    tinued during the past year’s general tele-   that, once it meets established technical    three main benchmarks—West Texas,
    com meltdown.                                 specifications, is supplier neutral. For     Brent, and Dubai—and their locations.
       Today, Akamai operates in 62 coun-         example, if flour has certain features, the   Electricity is pooled over large areas and
    tries and has 11,600 servers—more than        bakery does not care where the flour         then location is factored into the price.
    double its count in late 2000—within          comes from—only its price and avail-         Similarly, bandwidth is a point-to-point
    820 networks and more than 2,000              ability matter. The bakery’s long-term       offering with developing city-pair price

2               Computer
indices—for example, New York to Los
Angeles.                                          Resources
   The commodity that companies such as
Akamai, Content Bridge, and their com-             The following resources provide additional information about online com-
petitors supply is Web-content caching.           modities and peer-to-peer networking.
Customers do not care who provides this
commodity as long as it meets their con-          Online commodities
tent’s distributed response time.
   Content caching requires compute                 • IETF Content Distribution Internet Working Group,
power, storage space, and bandwidth.                • J. Oltsik, “The Evolution of Enterprise Content Networking,” Lightwave,
Many suppliers offer hard disks and stor-             Sept. 2001, pp. 149-152.
age area networks to meet storage needs,            • W. Sweldens, “Is Cache Trading the Next Big Thing?” Telecoms Capacity,
and compute power is similarly available.             June/July 2001, p. 30.
Web-content caching firms offer an addi-
tional factor—setup speed that quickly            P2P
supports new users. Provisioning a new              • D. Clark, “Face-to-Face with Peer-to-Peer Networking,” Computer, Jan.
user electronically takes much less time              2001, pp. 18-21.
than contract negotiations that are lim-            • M. Macedonia, “Distributed File Sharing: Barbarians at the Gates?” Com-
ited by the speed of human interaction.               puter, Aug. 2000, pp. 99-101.
Storage and processors are not currently            • “Peer-to-Peer Pressure,” The Economist, Nov. 2, 2000, http://www.
available on a geographically distributed   
basis for providing services at electronic          • “Profit from Peer-to-Peer,” The Economist, June 21, 2001, http://www.

   The term online commodity refers to
goods that can be bought and used with           In the past, commodities moved to mar-        Thus far, Web-caching business mod-
the transaction speed of an electronic         kets as soon as there was a need and some    els have evolved from single companies
agent. Hardware and network designs            reasonable method for meeting it. Online     that own all the pertinent hardware to
have included online commodities for           commodities will make a double move to       alliances in which different companies
some time. For example, Sun Micro-             market as consumers not only buy and sell    provide various parts of the hardware.
systems sells servers with processors that     them online but also use them online.        Letting additional companies join these
it can turn on after consumers purchase        Business models are changing and busi-       alliances has been proposed, but online
them, and IBM sells mainframes that it         ness processes are being outsourced so       commodities will submerge any attempts
can completely reconfigure and audit in         rapidly that this move may keep pace with    to control pricing.
real time.                                     the evolution in IT outsourcing.                The P2P movement pioneered by
   However, regardless of how manufac-           The “Resources” sidebar provides           Napster, Gnutella, Freenet, and Grove
turers make them available, whether as         sources for additional information about     already produces software that supports
a single box or as part of a server farm,      online commodities and the P2P move-         the placement and location of storage,
there currently is no distributed elec-        ment.                                        control, and applications. Major com-
tronic marketplace for online commodi-                                                      panies such as Sun and Microsoft have
ties. Geographically distributed storage       CONSEQUENCES                                 also now entered the P2P fray with JXTA
services and processors cannot currently          When this distributed electronic mar-     and Hailstorm, respectively. All these
be obtained at electronic speed. Com-          ketplace for online storage and compute      companies realize that the Internet offers
panies already recognize that they are         power becomes a reality, current Web-        the advantages of scale in distributing
losing billions of dollars in potential rev-   caching business methods—at least with       resources, but they have been slow to rec-
enue because of this limitation.               regard to hardware—will become a lia-        ognize the value of specific geographical
   Online resources are an attractive          bility. A proprietary hardware system—       distribution.
option for companies that host ASPs, and       even one composed of offline com-                P2P software provides basic distrib-
virtual ISPs will add to the push to make      modities—could not compete with a dis-       uted data access but does not support
content caching available. Although the        tributed liquid market of online com-        reservation, pricing, or location. Cur-
peer-to-peer (P2P) movement initially          modities in terms of price or geographical   rently, consumers cannot request file
found distributed coordination and             distribution. A distributed online com-      space with a specific response-time limit
access problematic, it is now adding sig-      modity market could supply storage and       from a set of locations and receive infor-
nificant pressure from the availability        compute power less expensively and more      mation about price and availability.
side.                                          flexibly than any single organization.        Technically, making this possible would

                                                                                                             November 2001               3
    Internet Watch

    not be a big step for P2P software            NEW BUSINESS ROLES                             radius of influence of such a hypermar-
    because it already provides access. In           In Internet years, Web-caching com-         ket. On the Web, we can expect the same
    fact, startups such as Mojo Nation            panies may have about a decade before          effect from concentrations of storage and
    already have pricing schemes in place,        this division into online commodity com-       compute capacity. Aggregation also pro-
    but they do not offer distributed control     ponents and open-source software and           vides opportunities for statistical multi-
    nor is this present in various Grid initia-   market makers takes place, after the           plexing and bulk provisioning of the
    tives (         breathing space occasioned by the recent       underlying hardware. A geographically
       To duplicate the access that Web-          reduced interest of venture capital. This is   dispersed Web-caching business would
    caching companies provide, the next           about 18 months of calendar time,              prosper by outsourcing rather than own-
    development in P2P software should let        although the changes have already              ing this infrastructure.
    a consumer submit multiple response-          started. How can current Web-caching              Customization involves putting to-
    time-to-location requests for given stor-     firms evolve to take advantage of the           gether a geographically distributed offer-
    age sizes and receive grouped replies. The    changing environment rather than suffer        ing, composed of online commodities,
    next step after that is to provide content    from it? This commoditization will cre-        that exactly matches a customer’s needs.
    management. Duplicating the business          ate business roles that support cus-           Instead of outsourcing a network to a vir-
    relationships of Web-caching companies        tomization and aggregation or economies        tual private network, succeeding in the
    is more difficult, but a network of elec-      of scale.                                      future market will require Web-caching
    tronic agents tasked with requesting and         Wal-Mart is a prime example of the          companies to outsource content to virtual-
    coordinating distributed resources would      advantage of aggregation—few similar           content response-time providers.
    leave no alternative.                         competitors can prosper within the

                                                                                                       mphasizing geographical dispersion

                                                                                                 E     is a new concept, compared with the
                                                                                                       typical activities of large financial
                                                                                                 players such as Enron and MSDW, but
                                                                                                 structuring commodities into a sophisti-
                                                                                                 cated package is not. Linked with com-
                                                                                                 panies offering distributed control and
                                                                                                 application placement software, these
                                                                                                 will be some of the new competitors for
                                                                                                 existing Web-caching companies. They
                                                                                                 will also have the advantage of a back-
                                                                                                 ground in using packaged flexibility—
                                                                                                 commodity derivatives—to manage
                                                                                                 uncertainties about price and availability.
                                                                                                    Some current old-style Web-caching
                                                                                                 players have the domain expertise they
                                                                                                 need to become major players in the dis-
                                                                                                 tributed online storage and computing
                                                                                                 commodities markets. Their business
                                                                                                 models must change, but these compa-
                                                                                                 nies still have a chance to recapture their
                                                                                                 role in this evolving market place. ✸

                                                                                                 Chris Kenyon is a research staff member
                                                                                                 at IBM Zürich Research Laboratory,
                                                                                                 Switzerland. Contact him at chk@zurich.

                                                                                                  Editor: Richard G. Mathieu, Dept. of
                                                                                                  Decision Sciences and MIS, St. Louis
                                                                                                  University, St. Louis, MO 63108;

4                Computer

To top