Architecture and message delays

modus_ponens

Hi,

We have a doubt concerning how to organize a Livelink web service architecture. In our project, we have created several web services which interact with Livelink web services to operate. An idea of the information flow could be this:

     Our web services <----->  Livelink web services <-------> Livelink

Now we have to install this software on production environment. This environment is composed of several machines. A diagram could be the following:

Machine (with Websphere Application Server)
       |
       |
       V
    Machine (Balancer)
       |
    ------------------------------------------------------------
    |                                                                              |
    V                                                                            V
 Machine (Tomcat + Livelink instance)       Machine (Tomcat + Livelink instance)
 

Concerning this architecture, we have as requisite that our web services have to be installed on Websphere machine, but we are discussing where to install Livelink web services, either on Websphere or in a Tomcat instance. The idea is to minimize the information flow delay, taking into account that information interchange between different machines would always be slower than interchange over different applications on the same machine. Installing Livelink web services on Websphere would minimize the delay on Livelink web services calls and responses; installing them on Tomcat would minimize the delay on Livelink API calls and responses.

Can you give us some advice in order to ascertain which option is preferable?

Thanks in advance.
 

sgrasley

Another aspect to consider is the authentication tokens.  By default they are tied to a particular client machine IP address (i.e. the application server in the case of web services).  If the web services are behind a load balancer, and your token is stored outside the load balancer, then you could run into problems with the token being considered invalid.  This functionality is configurable in the Livelink security settings.

Other than that, I can't really direct you as to specific benefits of one architecture over the other.  I suspect that the communication between the web services and Livelink will be a little lighter weight than the communication between a consumer and the webservices, because the latter has additional overhead of SOAP, etc.  That might be another point that suggests putting the web services on the system outside of the load balancer.

Thanks,

Scott