# Custom sharding logic **URL:** \ **Category:** Dgraph\ **Tags:** kind:question\ **Created:** [August 29, 2020, 1:16pm UTC](https://discuss.dgraph.io/t/custom-sharding-logic/9899 "2020-08-29T13:16:05Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![aleclofabbro](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/aleclofabbro/32/4500_2.png) [@aleclofabbro](https://discuss.dgraph.io/u/aleclofabbro)\ **Post date:** [August 29, 2020, 1:16pm UTC](https://discuss.dgraph.io/t/custom-sharding-logic/9899/1 "2020-08-29T13:16:05Z")
I’m developing a platform that let users publish stuff, stored in Dgraph, in a social-network flavour. Instances of this platform can be deployed independently by any organization and delivering service as a standalone platform. Users can search instance contents using proper Dgraph query and indexes Organizations have also the option to join a federation. Federated instances allow users to seamlessly search for content through all federated-instances. This may be achieved to some extent by configuring each federated Dgraph as (1+)Group in a federated shard. anyhow a bunch of requirements shall be guaranteed: A : content generated in an instance must be stored in its relative group (in their own storage) B : query/read is allowed for all instances C : write is allowed for generating group only D : when federated querying the query should succeed with available groups, even if a are down I can imagine a couple of approaches for this: 1 : having a central single Zero, representing “The Federation” and adding some custom logic to a central single Zero properly tweaked to obtain these behaviors, maybe by implemnting some plugins? 2 : having a dedicate Dgraph for search functionalities in all instances, in this case federated organization may share the whole searchable dataset as replicas option 1 looks the most interesting, if Zero logic can be tweaked option 2 may be more straightforward, and would make requirement A not relevant. I wonder if it would be possible to store indexes only for search purposes, and not the whole datasets (as they would become a heavy redundance) this question is kind of a followup to [this topic](http://discuss.hypermode.com/t/shard-zero-and-alpha-nodes-in-a-federated-flavour/9709) ---
**Author:** ![chewxy](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/chewxy/32/4339_2.png) [@chewxy](https://discuss.dgraph.io/u/chewxy)\ **Post date:** [September 7, 2020, 4:31am UTC](https://discuss.dgraph.io/t/custom-sharding-logic/9899/2 "2020-09-07T04:31:15Z")
Hi @pawan do we know of any way to do this?