You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The AWS Cloud Development Kit (CDK) is used to deploy a single Allora Worker Node. The CDK application deploys the following infrastructure:
28
-
28
+
29
29
- Virtual Private Cloud (VPC)
30
30
- Internet Gateway (IGW) to allow inbound requests for inferences from consumers and outbound responses from the worker node revealing inferences
31
31
- Public subnet that has a direct route to the IGW
32
32
- Security Group (SG) with TCP Port 9010 open inbound allowing requests for inferences to be routed to the Allora Worker Node
33
33
- Single Amazon Elastic Compute Cloud (EC2) instance (the Allora Worker Node) assigned to the public subnet
34
34
35
-
The Allora Worker Node is accessed by the user internally and is not exposed to the Internet to protect the node from unauthorized access. A user can gain access to the EC2 Instance using AWS Session Manager.
35
+
The Allora Worker Node is accessed by the user internally and is not exposed to the Internet to protect the node from unauthorized access. A user can gain access to the EC2 Instance using AWS Session Manager.
36
36
37
37
Multiple processes run on the Allora Worker Node (EC2 instance):
38
38
@@ -41,7 +41,7 @@ Multiple processes run on the Allora Worker Node (EC2 instance):
41
41
42
42
Allora Public Head Nodes publish the Allora chain requests (requests for inferences from consumers) to Allora worker nodes. When a worker node is initialized, it starts with an environment variable called BOOT_NODES, which helps handle the connection and communications between worker nodes and the head nodes.
43
43
44
-
The worker node (docker container) will call the function that invokes custom logic that handles the actual inference. The request-response is a bidirectional flow from the Allora chain (inference requests from consumers) to the public head nodes to the worker node and finally to the model server that reveals inferences.
44
+
The worker node (docker container) will call the function that invokes custom logic that handles the actual inference. The request-response is a bidirectional flow from the Allora chain (inference requests from consumers) to the public head nodes to the worker node and finally to the model server that reveals inferences.
45
45
46
46
## Additional materials
47
47
@@ -68,7 +68,7 @@ This is the Well-Architected checklist for the Allora worker nodes implementatio
68
68
| Performance efficiency | Compute selection | How is compute solution selected? | Compute solution is selected based on best price-performance, i.e. AWS EC2 T3 Medium instance suitable for bursty workloads. |
69
69
|| Storage selection | How is storage solution selected? | Storage solution is selected based on best price-performance, i.e. Amazon EBS volumes with optimal IOPS and throughput. |
70
70
|| Architecture selection | How is the best performance architecture selected? | A combination of recommendations from the Allora Network community and Allora Labs testing. |
71
-
| Sustainability | Hardware & services | Select most efficient hardware for your workload | The solution uses AMD-powered instances. There is a potential to use AWS Graviton-based Amazon EC2 instances which offer the best performance per watt of energy use in Amazon EC2.
71
+
| Sustainability | Hardware & services | Select most efficient hardware for your workload | The solution uses AMD-powered instances. There is a potential to use AWS Graviton-based Amazon EC2 instances which offer the best performance per watt of energy use in Amazon EC2.
0 commit comments