Popular posts

Elastic is a trendy keyword that can means a lot of things. We are elastic in several ways:

• we can add more indexing capacity (more clusters)
• we can add more search capacity (more machines with copy of data, our biggest user perform around 150M queries per day)
• we can distribute your search worldwide while ensuring a fast search everywhere (adding DSN replicates in a different region)
• we have all data in memory in 99% of cases but we have few use users where over-booking make sense to reduce costs (over-booking of memory)

At the end, elasticity depends a lot of the use case

We host ou starter to pro plans on high-end clusters, we usually host up to 100 customers on a cluster. When we install a new cluster, it is empty. Of course users can grow above the limit of the cluster, in this case we migrate some users on the new cluster, this is an operation which is transparent for our users and let them upgrade or downgrade. We ensure we always have more resource than the current usage (we keep around 25% of free resource on our global infra for scaling)

For our big enterprise customer, we have a dedicated infrastructure for them and they have the control on upgrade/downgrade. We also have a migration procedure to move from a pro plan to an dedicated infrastructure, you don't need to reimport your data.

With the growth we have, we add new machines in production close to every day. So we scale up the infra very often. There is no technical limitation to scale down (same principle than scaling up), we hope it will never be needed :)

Julien's comment answers it well :+1: We definitely plan to keep improving our abilities to scale up & down as fast (dynamically?automatically?) as possible -> it's always a matter of focus & priorities. Right now, our current elasticity capabilities covers 99% of the use-cases we're targeting.