Overview

Clusters refer to groups of Arx nodes that collaborate to execute confidential computations in the Arcium Network.

They're created by Computation Customers who define a set of Arx nodes based on specific requirements, such as computational capacity, security features, and the nodes' reputations.

During the creation process, customers specify key attributes, such as:

• Computational Load Capacity: The maximum computational volume a Cluster will be required to handle at any given time. Since different computations vary in size, a single large computation could demand significantly more resources than multiple smaller computations.

• Active Node Requirements: Defines the minimum number of active nodes required within the Cluster, ensuring consistent computational capacity. This includes the Node Priority List, which determines the activation order of nodes. While most Cluster properties, such as the Node Priority List, are immutable after creation, computational limits can be increased if supported by all participating nodes.

• Security Requirements: Outlines the security standards and protocols for the Cluster.

And more.

Customers can also reuse existing Clusters that fit their requirements, reducing setup costs and improving resource utilization. For example, a project might reuse a high-capacity Cluster for private trading algorithms instead of setting up a new one.

Participation & Admission of MXEs

Once created, a Cluster’s Arx nodes must approve their participation.

This approval process allows nodes to consider various factors, including their own workload, trust in other nodes, and geographical jurisdiction. Clusters can be further configured to include random nodes selected by the network’s sybil-resistance mechanisms, strengthening security by reducing the risk of collusion or centralized control.

Apart from the requirement for node approval to add nodes to Clusters, each Cluster must admit MXEs, which contain computation definitions. MXEs act as the environment where computations are executed, defining the specific types of tasks that can be processed. This mechanism for MXE admission is entirely distinct from the Cluster invitations system.

Admission is managed through configurable behaviors: some Clusters may require unanimous approval from all nodes for new MXEs, while others may have a streamlined process that automatically admits requests. If a single Arx node rejects the admission of a given MXE to their Cluster, then the MXE simply fails.

This flexibility ensures that Computation Customers can quickly find a Cluster suitable for their needs, balancing speed and security.

Parallelization and Scalability

Clusters in the Arcium Network are structured to optimize computational efficiency by distributing tasks across multiple nodes and Clusters, enabling high-performance processing and scalability.

Rather than parallelizing a single computation across many nodes within a Cluster, Arcium focuses on handling different computations concurrently across the network, ensuring reliability and high availability.