Short update after the discussion with Azeez. - The need to re-write the throttle core is still at large, so the best was to see how we can decouple the persistence logic from the throttle core (at least as much as possible). - A cluster updatable global counter will be included to the ThrottleContext. The idea is that each node will periodically broadcast the local counter info to the members in the cluster and the ThrottleConfiguration will update the value of the Global counter summing up the local counter values. - The ThrottleConfiguration will also push the global counter values to the Axis2 Configuration Context; a K, V pairs identified by the ThrottleContext ID. - A new platform component needs to be written to read the throttle related Axis2 Config Context list and persist them periodically (duration configurable). The throttle core will have no visibility into this persistence logic, so this will be completely decoupled. - So who should do the persistence? We can start with letting all nodes to persist first, but later (or in parallel) we can improve the Hazlecast's leader election (if that's not already there), so that the leader takes the responsibility of persisting. - The counters will be read off the persistence store at the time of Hazlecast Leader election takes place? (An alternative is to load the global counters at the init of ThrottleConfiguration but that means coupling throttle core with persistence.)
I will update the design accordingly. Any more thoughts or suggestions? Regards, Manoj On Thu, Dec 19, 2013 at 12:30 PM, Manoj Fernando <[email protected]> wrote: > +1. Let me setup a time. > > Regards, > Manoj > > > On Thursday, December 19, 2013, Srinath Perera wrote: > >> We need Azeez's feedback. Shall you, myself, and Azeez chat sometime and >> decide on the first Arch design? >> >> >> On Thu, Dec 19, 2013 at 11:55 AM, Manoj Fernando <[email protected]> wrote: >> >> Hi Srinath, >> >> That sounds like a much cleaner solution. We can perhaps use the native >> map-store declarative [1] which I think does something similar. It may >> sound a little silly to ask... but are we keeping Hazlecast active on a >> single node environment as well? :) Otherwise we will have to handle >> persistence on a single node in a different way. This is with the >> assumption of needing to persist throttle data on a single node environment >> as well (but questioning if we really need to do that is not totally >> invalid IMO). >> >> Shall we go ahead with the Hazlecast option targeting cluster deployments >> then? >> >> - Manoj >> >> [1] https://code.google.com/p/hazelcast/wiki/MapPersistence >> >> >> On Thu, Dec 19, 2013 at 10:51 AM, Srinath Perera <[email protected]>wrote: >> >> One another way to do this use Hazelcast and then use "though cache" or >> "Change listener's" in Hazecast for persistence. >> >> --Srinath >> >> >> On Tue, Dec 17, 2013 at 4:49 PM, Manoj Fernando <[email protected]> wrote: >> >> +1 for persisting through a single (elected?) node, and let Hazlecast do >> the replication. >> >> I took into consideration the need to persist periodically instead of at >> each and every request (by spawning a separate thread that has access to >> the callerContext map)... so yes... we should think in the same way for >> replicating the counters across the cluster as well. >> >> Instead of using a global counter, can we perhaps use the last updated >> timestamp of each CallerContext? It's actually not a single counter we >> need to deal with, and each CallerContext instance will have separate >> counters mapped to their throttling policy AFAIK. Therefore, I think its >> probably better to update CallerContext instances based on the last update >> timestamp. >> >> WDYT? >> >> If agree, then I need to figure out how to make delayed replication on >> hazlecast (is it through the hazelcast.heartbeat.interval.seconds config >> item?) >> >> Regards, >> Manoj >> >> >> On Tue, Dec 17, 2013 at 4:22 PM, Srinath Perera <[email protected]> wrote: >> >> We need to think it a cluster setup do we need persistence as well? As we >> can have replication using Hazelcast? >> >> If we need persistence, I think it is a good if a single node persists >> the current throttling values, and if that node fails, someone else takes >> it place? >> >> Current implementation sync the values across the cluster per each >> message, which introduce significant overhead. I think we should go to a >> model where each node collects and update the values once few seconds. >> >> idea is >> 1) there is a global counter, that we use to throttle >> 2) Each node keep a global counter, and periodically it update the global >> counter using value in location counter and reset the counter and read the >> current global counter. >> 3) Until next update, each node make decisions based on local global >> counter values it has read already >> >> This will mean that the throttling will throttle close to the limit, not >> exactly at the limit. However, IMHO, that is not a problem for throttling >> usecase. >> >> --Srinath >> >> >> >> >> On Mon, Dec 16, 2013 at 7:20 PM, Manoj Fernando <[email protected]> wrot >> >> > > -- > Manoj Fernando > Director - Solutions Architecture > > Contact: > LK - +94 112 145345 > Mob: +94 773 759340 > www.wso2.com > > -- Manoj Fernando Director - Solutions Architecture Contact: LK - +94 112 145345 Mob: +94 773 759340 www.wso2.com
_______________________________________________ Architecture mailing list [email protected] https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
