inconsistent bandwidth resource limits?

[jonesy]

Hi all!

I just noticed something odd in my application, which is derived from the LampCluster template. What I've noticed is that, while my database has Max bandwidth of 1G, my load balancer and web servers have a Max of 100M. Is there some logic behind having web server and database server bandwidth set to different Max values? Is there any reason I shouldn't branch my class and override that value, so long as I have the resources?

Thanks.

[PeterNic]

Brian,

Currently the bandwidth parameter is used only to make the scheduling decision, it is declarative; there is no actual limiting of the bandwidth at runtime (this is described in the known limitations in the release notes).

The database server having 1Gbps limit seems like an omission; I doubt MySQL (or whatever you use for a database) will actually be able to pull a full Gigabit. 100M would be more realistic, you may want to test what you're getting -- use the MON appliance to monitor the # bytes sent/received on the "in" terminal of MySQL.

So, in summary: you can leave it as is, or feel free to branch and fix the maximum -- probably will not make a difference unless you really get above 100Mbps sustained traffic. And even then, the only difference is that AppLogic will not allow you to schedule more than 2 components that have 1 Gbps bandwidth on the same server, even if there is extra CPU and memory (since the physical node's connection is 1 Gbps full duplex, so you cannot get more than 2 Gbps total). We're working on getting the bandwidth setting more meaningful/easier to choose.

Regards,
-- Peter