Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Fault tolerant event service

1 view
Skip to first unread message

Bjørn

unread,
Jan 23, 2001, 10:35:30 AM1/23/01
to
Hi everyone...

Struggling in the domain of telecom, my employer would expect the services I
partake in implementing to have a certain degree of fault tolerance.

When I'm basing my work on Visibroker, will using the Event Service always
introduce a single point of failure, or is possible to configure/federate
the Event Service in such a way that it is fault tolerant?

The Naming Service can at least be replicated, is there such a mechanism in
any Event Services??

Thanks for your advice!

Bjørn W. Bjanger, ConsultIT AS


Lev Korostyshevsky

unread,
Jan 23, 2001, 1:37:02 PM1/23/01
to
I think you can arrange something based on OAD and / or triggers, but if
you need a clean solution, I suggest you write your own event service.
You see, the event service has many major drawbacks.

* It is not scalable, because VisiBroker only allows one instance of a
channel. Even if they had allowed more than one instance, they would
have had to support persistent storage of messages, otherwise the
instances would have to be syncronized, which is expensive.

* It is not fault tolerant, precisely for the same reason.

* It has no control whatsoever over the quality of service; this also
means that messages may not arrive, or arrive in an incorrect order.

* It forces all consumers and suppliers to be activated as objects. I
believe the spec says that push suppliers may be pure clients, but it
doesn't really work with VB. This causes a performance penalty, as well
as unnecessary life cycle problems.

* Multiple messages cannot be sent.


Unfortunately, VBC does not support neither the Notification Service nor
the Messanging Service, which solve some of the problems described
above.


Lev

Bjørn <bja...@postkassa.no> wrote in message news:3a6da47b_1@dnews...

malcol...@my-deja.com

unread,
Jan 23, 2001, 5:42:59 PM1/23/01
to
In article <3a6da47b_1@dnews>,
Hi Bjorn,

the other respondent, Lev, sums up the issues quite well. If switching
ORBs is an option you may want to look at TAO. The last few months we
have been adding fault tolerance elements into the real time event
service of TAO and other areas. TAO is open source and has zero cost
license. I know there are other switching costs but before the project
goes too far you may want to explore that option.

regards Malcolm Spence

--
Business Development Manager
Object Computing Inc. (OCI)
St. Louis MO 63141
(314) 579-0066 ext.226


Sent via Deja.com
http://www.deja.com/

0 new messages