Riak Core: VNodes of different flavors
russell.brown at icloud.com
Fri Jun 9 15:04:14 EDT 2017
On 9 Jun 2017, at 19:16, Lloyd R. Prentice <lloyd at writersglen.com> wrote:
> A number of definitions and explanations of VNodes can be found in various sites on the web. I think I have a somewhat fuzzy understanding of VNodes and how they are replicated around the ring in a Riak Core cluster.
> Slide 11 in the slide deck below shows "Your App" resting on top of Riak Core and around Riak KV:
> What's not clear to me is, if "Your App" has multiple functionalities or services, how are these different functionalities mapped to VNodes?
> For instance, say one functionality requires user input, another obtains a value from a sensor, another computes and stores a second value, and yet another graphs and displays those values to the user.
> How is a system such as this mapped to VNodes in Riak Core--- all wrapped into a single VNode implementation or parsed across VNodes of different type or functionality and all replicated in the same ring?
It’s really up to you. It makes sense if you have multiple separate units of work to have multiple vnodes. RiakKV (the database) has vnodes for reading/writing data, and other vnodes (riak_pipe vnodes) for map/reduce, for example. If you look at Ryan Zezeski’s seminal and still relevant TryTryTry blog series, you’ll see it has different vnodes for different tasks (https://github.com/rzezeski/try-try-try.)
> Many thanks,
> Sent from my iPad
> riak-users mailing list
> riak-users at lists.basho.com
More information about the riak-users