Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re[4]: [equinox-dev] About the benchmarks of OSGi

Isnt benchmarking wonderful! :-)

Kind regards,

     Peter Kriens

JM> Yes but the resolver etc is more complicated and takes more
JM> work. �What you are calling esoteric may well be main line for
JM> Eclipse (and vice versa). �Pascal'point is very apt. �If you are
JM> measuring startup and footprint the R4 changes likely play a role.
JM> �In any event, implementations should stand on their
JM> merits/behaviour. �The scale factor should not be underestimated
JM> also. �Eclipse has been optimized to run 1000s (yes, 3 zeros) of
JM> bundles. �Our footprint might be larger to start but in larger
JM> systems we are smaller and faster.

JM> Jeff

JM> Peter Kriens <Peter.Kriens@xxxxxxxx>
JM> Sent by: equinox-dev-bounces@xxxxxxxxxxx
JM> 04/13/2005 10:09 AM
JM> Please respond to
JM>  pkriens and Equinox development mailing list

JM> To
JM> Pascal Rapicault/Ottawa/IBM@IBMCAcc
JM> Equinox development mailing list <equinox-dev@xxxxxxxxxxx>Subject
JM> Re[2]: [equinox-dev] About the benchmarks of OSGi

JM> R3 are still compatible with R4. R4 added some new
JM>  class loading features but they are mostly in the more esoteric range
JM>  for large deployments that do not want to refactor their code.

JM>  R3 bundles are still more or less the preferred mode and run unchanged
JM>  on R4 frameworks.

JM>  Kind regards,

JM>  � � Peter kriens

 PR>> First you would have to compare OSGi implementations that are
 PR>> functionnaly equivalents. For example eclipse' osgi implements
 PR>> R4.0 behavior which is a lot different than the R3.0.
 PR>> Then depending on which aspects you want to measure you would
 PR>> probably have to instrument the code base in order to get
 PR>> comparable numbers.
 PR>> Some numbers that could be interesting:
 PR>> ????- memory consumption of a bundle
 PR>> ????- time to load a class
 PR>> ????- startup time
 PR>> ????- how many bundles can the system handle
 PR>> ????- service lookup time

 PR>> PaScaL

 PR>> "rvm2004" <rvm2004@xxxxxxx>
 PR>> Sent by: equinox-dev-bounces@xxxxxxxxxxx
 PR>> 04/12/2005 11:19 PM
 PR>> Please respond to
 PR>> �Equinox development mailing list

 PR>> To
 PR>> equinox-dev@eclipse.orgcc
 PR>> Subject
 PR>> [equinox-dev] About the benchmarks of OSGi

 PR>> Hi all
 PR>> �I am very interested in the Eclipse and OSGi technology. Now
 PR>> I am doing the research about OSGi, and I want to know whether
 PR>> there are some key points or benchmarks that can measure the
 PR>> performance of the OSGi platform(such as Eclipse OSGi,SMF etc.)
 PR>> ,for example the usage of virtual memory, physics memory, basic
 PR>> startup time, and so on. That is if I want to compare several OSGi
 PR>> platform’s performance, and optimize one of them, which aspects
 PR>> should I consider? Which aspects could demonstrate the performance
 PR>> of a OSGi platform?And if I want to do some optimization for OSGi
 PR>> platform,which aspects can I optimize?what is the Future
 PR>> directions?Any response will be appreciated, thanks very much.

 PR>> �Best wishes, Zhoujing

 PR>> ?网易188财富邮--财富的象征;全新开放, 丰富用户名等您注册
 PR>> ��最�?G空间,附件40兆收发;免费注册体验
 PR>> �equinox-dev mailing list
 PR>> �equinox-dev@xxxxxxxxxxx
 PR>> �

JM>  �

JM>  -- 
JM>  Peter Kriens � � � � � � � � � � � � � � �Mob +33633746480
JM>  9C, Avenue St. Dr�z�ry � � � � � � � � � �Tel +33467542167
JM>  34160 Beaulieu, France � � � � � � � � � �Tel +15123514821
JM>  AOL,Yahoo pkriens � � � � � � � � � � � � ICQ 255570717

JM>  _______________________________________________
JM>  equinox-dev mailing list
JM>  equinox-dev@xxxxxxxxxxx


Peter Kriens                              Mob +33633746480
9C, Avenue St. Dr�z�ry                    Tel +33467542167
34160 Beaulieu, France                    Tel +15123514821
AOL,Yahoo pkriens                         ICQ 255570717

Back to the top