we have a wiki with 30K generated pages (from an UML model). This is a test
instance, so no current CPU load.

JSPWiki Engine Version 2.10.1

Total Number of Pages 29079

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
 1270 wiki      20   0 3635920 1.167g   6848 S   0.0 15.3 509:18.11 java


Am 29.03.2018 14:18 schrieb "Juan Pablo Santos Rodríguez" <
juanpablo.san...@gmail.com>:

Hi,

a little late to chime in, but did wanted to note a couple things:

- there's a contributed jdbc provider ([#1], referenced at [#2]), so it is
possible to store your wiki pages on a database

- if I'm not mistaken, all pages aren't stored on memory, references, both
ingoing and outgoing, to them are ([#3]). Those references are Strings
(page names), so references should be on the order of hundreds of bytes, as
opposed to whole pages. These references are used to perform searches like
top ten lists, unreferenced pages, etc. which otherwise would be expensive
to make. There's also the lucene index which IIRC is backed by a file, and
ehcache if you're have caching turned on, which you can customize too. I've
never tried to see where the upper limit on pages would be, so I'm not
really sure how many pages JSPWiki can handle on an standard installation.

Out of curiosity, I've done an mvn tomcat7:run-war with latest JSPWiki and
the default set of pages and the amount of memory it takes is ~60MB, on a
JDK 1.8.0_121, no memory parametrization at all. Would anyone mind sharing
the amount of pages vs JDK - memory settings vs memory consumed on JSPWiki
instances?

Last, @V.Fedorov, ELWiki looks really interesting! is there any link we
could use to refer to it at
https://jspwiki-wiki.apache.org/Wiki.jsp?page=JSPWikiApplications ?


best regards,
juan pablo


#1]: https://github.com/djemerson01/jspwiki-jdbcprovider
[#2]: https://jspwiki-wiki.apache.org/Wiki.jsp?page=ContributedProviders
[#3 <https://jspwiki-wiki.apache.org/Wiki.jsp?page=ContributedProviders[#3>
]:
https://github.com/apache/jspwiki/blob/master/jspwiki-war/src/main/java/org/apache/wiki/ReferenceManager.java

On Thu, Mar 22, 2018 at 7:16 PM, V. Fedorov <fedorov...@yandex.ru> wrote:

> Believe it or not, but the JSPwiki code is already ported on components
> - OSGi of a bandles, and...:
>  - the layer of visualization has produced with use of Eclipse RAP;
>  - data are submitted by EMF model;
>  - the persistence of data is carried out by means of CDO.
>
> Screenshot of the functioning wiki - see https://yadi.sk/i/UezpmVo63Tf4
> 2g
> (in the picture: at the left - IDE; on the right - the browser with
> ElWiki).
>
> There is a one change in syntax of a wiki-marking - for references
> between pages it is required to specify the page identifier.
> For example: [Sample page | @218]
> Because of it the name wiki is: ElWiki (Enhanced links wiki).
> The data model of these pages is organized hierarchically, references
> between them are based on identifiers of pages.
> So existence of pages with identical names, for example in different
> branches of wiki is possible.
>
> Used data model is presented here - https://yadi.sk/i/jaljk1m23Tf3zw
>
> In the presented realization there is no need to load all pages, for
> processing of communications between them.
>
> Work on the project is continued. (changes has been started and
> continuing by me)
>
> PS: In general, I use JSPwiki for logging of my practices, notes from
> books, articles.
> Though in my option - the pages of JSPwiki are stored in the database,
> not on a disk,
>  but in general, loading all pages in memory, impossibility to set the
> any name of page, and etc. -- doesn't suit me now.
>
> On Чт, 2018-03-22 at 08:26 +0100, Jürgen Weber wrote:
> > guess not, on startup JSPWiki loads all pages into memory to parse
> > intra-page links.
> >
> > 2018-03-22 5:24 GMT+01:00 dagarwa...@gmail.com <dagarwa...@gmail.com>
> > :
> > >
> > > JspWiki uses filesystem instead of a conventional database. Would
> > > you suggest using jspwiki if it were to cater to 1billion page ?
> > > (Provided, a solid backup mechanism ). Would it scale that much
> > > with both performance and data-integrity perspective ?
>

Reply via email to