[forge-dev] Improving the performance of weld for micro-deployments.

[Jozef Hartinger]

I worked on Weld's bootstrap performance about 6 months ago but the 
focus was mainly on large deployments. I am not aware of any obvious 
bottlenecks that would get into the way of micro deployments. If you 
could do a further analysis that would be great.

As for guava showing up in the stats, a lot of work Weld does is done 
within computing maps (e.g. reading metadata using reflection, etc.) so 
you would need to get more in depth here.

Weld uses its own thread pool for concurrent loading, deployment and 
validation of beans. Furthermore, there is a service that pre-resolves 
extension observer methods in multiple additional threads. The thread 
pool sizes default to a configuration that should utilize the most of 
CPU in cases when a single Weld instance is running. However, in your 
environment where multiple Weld instances are booting at the same time 
this may actually harm performance. As a first step I would suggest 
disabling concurrent deployment and the preloader or playing with thread 
pool sizes to see if it changes anything.

See 
http://docs.jboss.org/weld/reference/2.0.0.Beta3/en-US/html/configure.html#d0e5936 
for how this is configured. Note that you'll need to wait for Beta3 as 
the configuration options have changed recently.

On 01/29/2013 07:19 PM, Lincoln Baxter, III wrote:
> Hi Jozef, Stuart, and Weld-devs,
>
> In Forge 2 we are using Weld extensively, and one of the things we do 
> is start up many instances simultaneously.
>
> We may have anywhere from one to one-hundred or more weld instances. 
> Currently we have only seen around 10-12 instances, and performance is 
> "Okay", but in theory, we could see hundreds of instances, at which 
> point, performance starts to be a concern. We're working around this 
> problem by disabling CDI support on some internal addons, but... it's 
> not really reasonable to expect that everyone will do this.
>
> Which means... we need to figure out how to shave as much time off the 
> bootstrap as possible. Currently each weld instance takes anywhere 
> from 80ms to 450ms to start (not really sure why such variation yet,) 
> and we'd hopefully like to get that down even lower, around 10-20ms. 
> Classloading time only would be optimal, but obviously difficult to 
> achieve.
>
>
> How can we get the most speed out of Weld? Most of our deployments 
> have only ~15 bean classes at most. It seems like a lot of time 
> (~30-40%) is being spent in the Google concurrent collections.
>
> (Screenshot attached.)
>
> Thanks,
>
> -- 
> Lincoln Baxter, III
> http://ocpsoft.org
> "Simpler is better."