Optimizing Regional Support: What the Data Says, and the Plan Ahead

Pocket Network currently supports 20 regions around the world via two cloud providers, in the name of providing speedy, unstoppable RPC access.

In conjunction with our community partner, POKTscan, and in relation to our recent announcement and efforts to reduce costs across the network, we are evaluating the viability, cost-efficiency, and performance impact of our regions across the network. Historically, our approach has been to add more Portals in more regions, in an effort to reduce latency.

But what if adding more regions could actually increase latency?

The Numbers

The POKTscan team recently shared some interesting and compelling data with us regarding the locations of our Portals. The data offers valuable insights as to how many Portals we should realistically have in order to optimize latency.

Please see the POKTscan team’s forum post discussing their findings in greater depth. Our overview today will be high-level, and highlight our plan to reduce our number of regions, and the goal of doing so.

In the above graph, we note that after the third Portal region, we see exponentially diminishing returns with regard to latency. The three regions that POKTscan’s analysis highlighted as optimal for coverage are:

  • US East (Virginia)
  • EU Central (Frankfurt)
  • AP South (Singapore)

Coincidentally, with the analysis provided by POKTscan, these are the regions Pocket is already deploying in with our new cloud provider.

The Plan

In our recent Update to the DAO, we indicated our pursuit of a new cloud provider, which will help us with our high-priority goal to cut costs. We’ll have a subsequent update to properly announce which provider that is and provide more information as to how we came to that decision, but for now we can share that we have executed on an agreement, and we will be consolidating our entire cloud presence on our new provider.

During this consolidation, we will be deprecating and migrating Portal Regions to move towards greater network efficiency and cost-efficiency. As discussed above, we will first reduce our current AWS presence incrementally from the current 20 supported regions to our final three:

  • US East (Virginia)
  • EU Central (Frankfurt)
  • AP South (Singapore)

This change represents an opportunity to reduce burn by up to $960,000 annually, and increases the Network’s quality of service at the same time.

Our goal is to test the reduction in each region incrementally to ensure that the reality of reducing the regions doesn’t negatively affect latency or the network. Over the next two quarters, we will aim to complete both our migration and our region reduction. As each region comes up for deprecation, we will make an announcement in Discord at least 24 hours prior to deprecating it. We will then monitor for a time (approximately 72 hours) before looking to retire the next region.

We’re really thankful for the analysis conducted by POKTscan on Portal placement across regions, and we’re glad to use sound data in making this decision. Please stay tuned for additional communication as we start rolling out this plan.