FR2815212A1 - Loading of a multi-page file from Internet server to client, uses page sequence profiles and loads pages in anticipation on basis of profile and current page - Google Patents

Abstract

The file resides on a server and has multiple pages. The server provides a database of page sequence types (Si), and a sequence type is chosen on the basis of the current page (Pv). Pages following the current page are then loaded to the client. The execution of this procedure is on the client machine.

Description

<Desc / Clms Page number 1>

The present invention relates to a device for loading a file comprising a plurality of pages.

 The device is arranged in a workstation, a personal computer, a workstation or any equivalent equipment commonly referred to as "client".

 The file resides in a server equipment connected to the workstation by a transmission network, the Internet for example. The term server must be understood in its broadest sense and includes in particular the websites.

 For example, the file, such as text, graphic, or image, can not be viewed in its entirety on the workstation screen. Moreover, its consultation is not always sequential, that is to say that in certain points several options are possible for the continuation of this consultation. It follows that the file is subdivided into pages which are independent entities and which can generally be fully visualized on the screen.

When the user has finished viewing a page, the current page, he sends a request to the server in which appears the next page he wants to obtain.

Due to network congestion or server overload, the wait time required to transmit the request and then load the next page in the workstation can be relatively long.

 Several solutions have therefore been proposed to load into a cache one or more pages that the user is supposed to require after the current page. To limit the waiting time, the loading takes place during the consultation of this current page.

 A first solution is to load all the pages that could possibly be requested by the user. This is not satisfactory because it leads to a considerable increase in traffic for the server. In addition, if the number of pages is large, the time required to load these pages can be very long and even exceed the consultation time of the current page plus the loading time of a page.

 A second solution implemented in the server is to define the profile of the user by memorizing the various consultations he has already done. A program then makes a prediction on the page that is going

<Desc / Clms Page number 2>

 ask the user based on his previous behavior. Again, the server is under an extra load since it must handle all file lookups.

chargement de la page courante. L'indication des pages prédéfinies figure dans la page courante si bien qu'il faut modifier tout le fichier. A third solution disclosed in the patent application FR 2 779 594 consists in loading certain predefined pages following the

loading the current page. The predefined pages are indicated in the current page so that the entire file must be modified.

la suite de la consultation de la page courante, ceci en limitant au maximum la surcharge du serveur. The present invention thus relates to a means of loading in advance one or more pages that the user is likely to require to

following the consultation of the current page, this limiting the server overload as much as possible.

According to the invention, a method of loading a file residing in a server, this file comprising a plurality of pages, the last of which is loaded on the current page, comprises the following steps: - acquisition from the server of a database which lists the standard sequences of pages of the file, - selection of at least one standard sequence which includes the current page, and - loading of the page which follows the current in at least one of the selected standard sequences as soon as the current page has been loaded.

Advantageously, an initial sequence of pages which ends with the current page having already been loaded, the method comprises a step for calculating the length of a sequence of pages which precedes this current page in the selected standard sequences and which is identical to a Final sequence of pages of the initial sequence, and it includes a step to load the page that follows the current page into at least one of the sequences that lead to the longest end run.

 According to an additional characteristic, the method comprises a step for loading the page that appears most frequently following the current page in the standard sequences presenting the longest final sequence.

Preferably, the method comprises a step for rejecting the standard sequences whose page following the current page belongs to the initial sequence.

Optionally, the method allows to load several pages from the selected type sequences.

 The invention also relates to a device arranged in a workstation comprising means for implementing the method above.

<Desc / Clms Page number 3>

 The present invention will now appear in greater detail in the context of the following description of exemplary embodiments given by way of illustration with reference to the appended figures which represent: FIG. 1, the elements necessary for the device according to the invention, and - Figure 2, a flowchart of the method according to the invention.

types qui est réalisée au niveau du serveur, un site Internet le cas présent. Lorsqu'un utilisateur vient se connecter à ce site pour consulter le fichier, le serveur réalise un enregistrement de la séquence des pages demandées ou log de visite. On accumule ainsi dans un répertoire un nombre important de logs de visite qui sont créés à l'occasion de chaque connexion au site. Le nombre retenu peut aller de quelques milliers à quelques million. In the first place, it is necessary to define the sequence database

types that is performed at the server level, a website this case. When a user connects to this site to view the file, the server makes a record of the sequence of the requested pages or visit log. Thus, a large number of visit logs are accumulated in a directory, which are created during each connection to the site. The number retained can range from a few thousand to a few million.

The next step is to reduce the size of the directory to keep only typical sequences so that its transmission to the workstation is ensured within a reasonable time. We will only remember a few hundred logs. To make this reduction, several possibilities are offered. For example, it is possible to make a selection at random or to retain a block of consecutive logs. It is also possible to start from an initial selection then to make it evolve by testing the effect of the addition of a new sequence on the performance of the device described below.

With reference to FIG. 1, the workstation essentially comprises a microprocessor CPU, a random access memory RAM and a cache memory CA, each connected to a first controller CONT. This first controller intended for bus and memory management is further connected to a PCI peripheral interface which is itself connected to a second MASTER controller in charge of managing the peripherals, among which is the SCR screen. The second MASTER controller is also connected to an internal INT interface to which is connected a mass storage unit such as a disk (not shown) and an external EXT interface connected to the network. Access to the SERV website is via a first S1 and a second S2 server called proximity.

requière le chargement de la base de données des séquences types dans la mémoire de masse du poste de travail. Avantageusement, le chargement est réalisé globalement ou par parties durant une ou plusieurs périodes d'inactivité de la liaison de transmission entre le poste de travail PC et le serveur SERV. La During the first consultation of the site, the user of the workstation

requires the loading of the database of the sequences in the mass memory of the workstation. Advantageously, the loading is performed globally or in parts during one or more periods of inactivity of the transmission link between the PC workstation and the SERV server. The

<Desc / Clms Page number 4>

 database may possibly be retained after consultation for further use. Each type sequence Si is indicated by an index i varying from 1 to t which is its rank in the directory. The page which appears in the jth position in the type sequence Si is marked Sjj, j varying from 1 to n.

 We now sit during a consultation conducted by the user of the workstation. This one has already required an initial sequence IS composed of the pages Pk, k varying from 1 to v, Pv thus appearing the current page of visualization or current page.

 Upon loading the current page detected by the first controller CONT, it triggers the following operations.

dans la séquence type Sjj-1. En cas d'identité, on réitère la comparaison en prenant la page précédente dans la séquence initiale Pv-2 et dans la séquence type Si, j-2 et ainsi de suite jusqu'à obtenir deux pages différentes. On identifie ainsi une suite finale de pages de longueur Li commune à la séquence initiale IS et à la séquence type Si, cette suite se terminant par la page courante Pv. With reference to FIG. 2, the initial sequence IS is compared with each of the standard sequences Si. If the current page Pv belongs to the type sequence Si and if it does not belong to the initial sequence IS, then the pages preceding the current page in the initial sequence Pv-1 and

in the sequence Sjj-1. In case of identity, the comparison is repeated by taking the previous page in the initial sequence Pv-2 and in the type sequence Si, j-2 and so on until two different pages are obtained. Thus, a final sequence of pages of length Li, which is common to the initial sequence IS and to the type sequence Si, is identified, this sequence ending in the current page Pv.

We then look for standard sequences that have the final sequence of maximum length. If a single sequence has the longest sequence, the device of the invention issues a request to cause the loading in the CA cache of the page following the current page in this sequence. If, on the other hand, several standard sequences comprise the final sequence of maximum length, we examine the page following that follows the current page Pv in each of these sequences. The frequency of appearance of each separate page is defined as the number of standard sequences in which it is present. The device of the invention then triggers the loading of the page followed by the greatest frequency of appearance if this page is unique.

In the opposite case, a continuation page is chosen at random, the first of the directory for example.

 The succession of operations described above can be performed for example by means of a computer program in C language represented in the appendix.

également dans un mode dégradé. Ainsi, la recherche de la suite finale la plus This is the preferred embodiment but the invention applies

also in a degraded mode. So, looking for the most final sequel

<Desc / Clms Page number 5>

 long is not essential and it is possible to load in the cache CA the page that follows the current page in the first sequence that includes this page. Similarly, it is not imperative to calculate the frequency of appearance of the various pages continued in the standard sequences comprising a final sequence of maximum length. It is indeed possible to retain the page following the first selected sequence.

sélection exposés ci-dessus, le dispositif de l'invention permet de charger dans la mémoire cache plusieurs d'entre elles. Moreover, when different pages meet the criteria of

selection exposed above, the device of the invention allows to load in the cache several of them.

 After consulting the current page, the user of the workstation specifies the next page he wants to consult. If this page has been loaded into the cache, it is immediately displayed. If this is not the case, a request is transmitted to the server as in the state of the art, but the invention does not lead to an increase in the waiting period.

 The embodiments of the invention presented above have been chosen for their concrete nature. It would not be possible, however, to exhaustively list all the embodiments covered by this invention. In particular, any means described may be replaced by equivalent means without departing from the scope of the present invention.

<Desc / Clms Page number 6>

Annexe

#define M 10/* length of the sequence */ #define NBP 1500/* # of html pages */ char tab) e [NBP] [255] ;/*) D and string of each page */ int database [1000] [M] ;/* max # of seq in Learning Base : 1000*/ int cache [NBP] ;/* content of the navigator cache */ int history [NBP] ;/*userbehaviororhistory*/ char URL [255] ;/* Page en cours de lecture par l'internaute */ int date=0, N1 ;/* # of seq. */ int cacheindx, predictindx ; int tsize ;/* # of pages */ /*******************************/ void Prediction (void) { char predict [255] ;/* La page predite par le systeme */ int q [NBP] ;/* stats per page */ int i, j, k, l, a, onmax ; int Ion, dejavu, maxq, maxindx ; /* get current URL asked by the user, determine its ID and increment the history sequence */ i=0 ; dejavu = 0 ; while ( {i < Csize) & & (dejavu == 0)) {i++ ; dejavu = 1 ; for (k=O ; k < =str) en (URL) ; k++) if (tabie [i] [k]'=URL [k]) {dejavu = 0 ; break ;} }/* end while */ if ( (i==t~size) & & (dejavu == 0)) printf ("Not in Database\n") ; history [date] = i ; date++ ; cache [cacheindx] =i ; cacheindx++ ; ion marx = 0 ; for (a=0 ; a < NB~P ; a++) {q [a] = 0 ;} for (k=O ; k < N1 ; k++) for (I=0 ; 1 < (M-1) ; 1++) {if (database[k][l] == history [date-1]) {dejavu = 0 ; for (a= (cacheindx-1) ; a > =O ; a--) {if (database [k] [I+1] == cache [a]) {dejavu=1 ; break ;} } if (dejavu == 0)

Annex

#define M 10 / * length of the sequence * / #define NBP 1500 / * # of html pages * / char tab) e [NBP] [255]; / *) D and string of each page * / int database [1000 ] [M]; / * max # of seq in Learning Base: 1000 * / int cache [NBP]; / * content of the browser cache * / int history [NBP]; / * userbehaviororhistory * / char URL [255]; / * Page being read by the user * / int date = 0, N1; / * # of seq. * / int cacheindx, predictindx; int tsize; / * # of pages * / / ******************************* / void Prediction (void) {char predict [255]; / * The page predicted by the system * / int q [NBP]; / * stats per page * / int i, j, k, l, a, onmax; int Ion, dejavu, maxq, maxindx; / * get current URL asked by the user, determine its ID and increment the history sequence * / i = 0; dejavu = 0; while ({i <Csize) && (dejavu == 0)) {i ++; dejavu = 1; for (k = 0; k <= str) in (URL); k ++) if (tabie [i] [k] '= URL [k]) {dejavu = 0; break;}} / * end while * / if ((i == t ~ size) && (dejavu == 0)) printf ("Not in Database \ n"); history [date] = i; date ++; cache [cacheindx] = i; cacheindx ++; marx ion = 0; for (a = 0; a <NB ~ P; a ++) {q [a] = 0;} for (k = 0; k <N1; k ++) for (I = 0; 1 <(M-1); ++) {if (database [k] [l] == history [date-1]) {dejavu = 0; for (a = (cacheindx-1); a> = O; a--) {if (database [k] [I + 1] == cache [a]) {dejavu = 1; break;}} if (dejavu == 0)

<Desc / Clms Page number 7>

Ion = 1 ; for (a= (date-2) ; a > =O ; a--) {if ( (I+a-date+1) > =0) {if (database [k] [) +a-date+1] ==history [a]) Ion++ ; else break ; } } if (ton > ) onmax) ion marx = Ion ; for (a=0 ; a < NBP ; a++) {q [a] = 0 ;} } if (Ion == onmax) {q [database [k] [I+l]] ++ ; }/* end if dejavu */ }/* end if database */ }/* end for 1 */ if (lon~max ! =O) {maxq=0 ;/* find the most interesting */ for (k=O ; k < NBP ; k++) if (q [k] > max~q) {max~q = q [k] ; maxjndx = k ;} } q [maxindx] = 0 ; predictindx = max~indx ; for (k=0 ; k < =strlen (table [predictindx]) ; k++) {predict [k] =tab) e [pred) ctjndx] [k] ;} predict ]='\0' ;/* generate a string */ }

Ion = 1; for (a = (date-2); a> = O; a--) {if ((I + a-date + 1)> = 0) {if (database [k] [) + a-date + 1 ] == history [a]) Ion ++; else break; }} if (ton>) onmax) ion marx = Ion; for (a = 0; a <NBP; a ++) {q [a] = 0;}} if (Ion == onmax) {q [database [k] [I + 1]] ++; } / * end if dejavu * /} / * end if database * /} / * end for 1 * / if (lon ~ max! = O) {maxq = 0; / * find the most interesting * / for (k = O; k <NBP; k ++) if (q [k]> max ~ q) {max ~ q = q [k]; maxjndx = k;}} q [maxindx] = 0; predictindx = max ~ indx; for (k = 0; k <= strlen (table [predictindx]); k ++) {predict [k] = tab) e [pred) ctjndx] [k];} predict] = '\ 0'; / * generate a string * /}

Claims (6)

 CLAIMS 1) Method of loading a file resident in a server (SERV), this file comprising a plurality of pages, the last loaded is the current page (Pv), characterized in that it comprises the following steps: - acquisition from said server of a database which lists the standard sequences (Si) of pages of said file, - selection of at least one standard sequence (Si) which comprises said current page (Pv) and loading of the page which following said current (Pv) in at least one of said selected sequences as soon as said current page has been loaded. 2) Method according to claim 1 characterized in that, an initial sequence (IS) of pages which ends with said current page (Pv) having already been loaded, it comprises a step for calculating the length (Li) of a sequence of pages preceding said current page (Pv) in said selected type sequences (Si) and which is identical to a final sequence of pages of said initial sequence (ins), and includes a step for loading the page that follows said current page (Pv) in at least one of the standard sequences that lead to the longest end run. 3) Method according to claim 2, characterized in that it comprises a step for loading the page which appears most frequently as a result of

 said current page (Pv) in said type sequences (Si) having the longest end run. 4) Method according to any one of claims 1 to 3, characterized in that it comprises a step for rejecting said standard sequences (Si) whose page following said current page (Pv) belongs to said initial sequence (ins) . 5) Method according to any one of the preceding claims, characterized in that it allows to load several pages from said selected sequences (Si). 6) Device arranged in a workstation (PC) comprising means for implementing the method according to any one of the preceding claims.