Re: [beehive-dev] Scaling existing Beehive System

[Soheil Hassas Yeganeh]

Hi Wintson,

Good question. You can load balance anyway we want using the migration and instrumentation capabilities. It is as simple as implementing a dynamic optimizer.

The existing dynamic optimizer (enabled with -instrument), tries to place bees in a way to reduce inter hive traffic. For your example, if we have new hives to the system and suddenly a few switches migrate to that hive, bees that are managing those switches are migrated accordingly.

This is good for SDN applications, but might not be optimal for other types workloads. For example, when the workload arrives at random on any hive with no predefined mapping. For those cases, the dynamic optimizer is pretty simple to implement. We just need to migrate a few bees to the new hive as soon as the new hive joins the cluster. It shouldn't be more than a few 10 lines of code.

Hope this helps.

Best,

Soheil

On Wed Feb 18 2015 at 11:44:48 AM Winston Yeung <yeung....@gmail.com> wrote:

Hello,

I was wondering: if a system is already established with beehive and the keys are already distributed, what happens when we try to scale the system by adding or removing new machines? How does Beehive handle the distribution of keys and data? Thanks!

Cheers,

Winston

--
You received this message because you are subscribed to the Google Groups "beehive-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beehive-dev+unsubscribe@googlegroups.com.
To post to this group, send email to beehi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/beehive-dev/a70e1de1-1432-44a0-9e2a-6664fa6675ad%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.