LiveCycle Mosaic applications are fundamentally a collection of SWF files that are downloaded to the client from the server and then run within Flash Player which in turn runs within a sandbox in the client browser. From an application user’s perspective, the time taken for that initial download of SWFs is extremely important. Various strategies such as minimizing SWF sizes, caching etc. need to be employed to minimize user wait time for initial application load. Also, the filesystem repository should be hosted on high performance storage.
We performed some tests to gauge the effect of network bandwidth on a sample Mosaic application in our Technical Marketing lab. Using Borland SilkPerformer, we performed multiple one-hour, 5 concurrent user tests on LiveCycle Mosaic 9.5 running on Oracle WebLogic 10.3 on Windows Server 2008. Each test employed throttling to simulate various network bandwidth scenarios such as ADSL 6 Mbps down, ADSL 9 Mbps down, LAN (1 Gbps up and down) etc. Each virtual user only had a maximum of four TCP connections permitted to the server. The results are charted below.
Please note that network bandwidth is not the same as network latency. Typically, traceroute provides this information. For example, the tracert output for this test (load test client to server) is provided. The network latency was less than 1 millisecond.
Tracing route to server.company.com [18.104.22.168]
over a maximum of 30 hops:
1 <1 ms <1 ms <1 ms server.company.com [22.214.171.124]
Based on the trend line equation in the chart, application load performance y (in seconds) can be predicted when effective network bandwidth x (in Mbps) is known.