Create the AWS EC2 Instances

Log in to the AWS Management Console and navigate to the EC2 service.

1. Configure Security Groups

Create a security group for your MemVerge.ai cluster.

• Navigate to EC2 Dashboard > Security Groups > Create Security Group
• Name: MemVerge.ai-sg
• Description: Security Group for MemVerge Memory Machine for AI (MemVerge.ai)
• VPC:
• Add the following inbound rules:
• SSH (22/TCP) from your IP
  • Type: SSH
  • Protocol: TCP
  • Port Range: 22
  • Source: Anywhere-IPv4 or your public IP address
  • Description: None
• Kubernetes API (6443/TCP) from 0.0.0.0/0
  • Type: Custom TCP
  • Protocol: TCP
  • Port Range: 6443
  • Source: Anywhere-IPv4
  • Description: Kubernetes API
• NodePort Services (30000-32767/TCP) from 0.0.0.0/0
  • Type: Custom TCP
  • Protocol: TCP
  • Port Range: 30000-32767
  • Source: Anywhere-IPv4
  • Description: NodePort Services
• HTTP Services (80/TCP) from 0.0.0.0/0
  • Type: HTTP
  • Protocol: TCP
  • Port Range: 80
  • Source: Anywhere-IPv4
  • Description: HTTP Web Service
• HTTPS Services (443/TCP) from 0.0.0.0/0
  • Type: HTTPS
  • Protocol: TCP
  • Port Range: 443
  • Source: Anywhere-IPv4
  • Description: HTTPS Web Service
• KubeFlow Service (30323/TCP) from 0.0.0.0/0
  • Type: Custom TCP
  • Protocol: TCP
  • Port Range: 30323
  • Source: Your public Internet IP Address (Anywhere-IPv4 works, but is insecure)
  • Description: Kubeflow Web Service
• Add the following outbound rules:
• Type: All traffic
• Protocol: All
• Port range: All
• Destination: Anywhere-IPv4 (0.0.0.0/0)
• Description: Allow all outbound traffic
• Click "Create security group"
• After creation, edit the security group to add internal communication ports:
• Select the newly created security group
• Click "Edit inbound rules"
• Add the following rules:
  • Type: Custom TCP, Port Range: 2379-2380, Source: Select the security group itself, Description: etcd
  • Type: Custom TCP, Port Range: 10250, Source: Select the security group itself, Description: Kubelet API
  • Type: Custom TCP, Port Range: 10259, Source: Select the security group itself, Description: kube-scheduler
  • Type: All Traffic, Source: Select the security group itself
  • Type: All ICMP - IPv4, Port Range: All, Source: Select the security group itself, Description: Allow Ping
  • Type: Custom TCP, Port Range: 10257: Source: Select the security group itself, Description: kube-controller-manager
  • Type: Custom TCP, Port Range: 8080: Source: Select the security group itself, Description: memcached
  • Type: HTTP, Port Range: 80: Source: Select the security group itself, Description: HTTP Web Service
  • Type: HTTPS, Port Range: 443: Source: Select the security group itself, Description: HTTPS Web Service

• Click "Save rules"

• Create an SSH Key pair for all cluster instances

• Navigate to EC2 Dashboard > Network & Security > Key Pairs

• Click "Create Key Pair"
• Name: MemVerge.ai-cluster-key
• Key pair type: RSA
• Private key file format: .pem
• Click "Create key pair"
• Download and securely store the .pem file

• Change the permissions on the .pem file using chmod 400 <ssh-key.pem>

• Launch EC2 Instances

• Navigate to EC2 Dashboard > Instances

• Click "Launch Instances"
• For the management host:
• Name: MemVerge.ai-mgmt
• Amazon Machine Image (AMI): Ubuntu Server 22.04 LTS (HVM), SSD Volume Type
• Instance Type: m5.xlarge (4 vCPU, 16 GiB Memory)
• Key pair: MemVerge.ai-cluster-key
• Network Settings:
  • Firewall (security groups): Select existing security group > mai-sg
• Configure Storage:
  • 1 x 60 GiB gp2 Root volume (Not encrypted)
• Click "Launch instances"
• For the GPU worker:
• Name: MemVerge.ai-worker001
• Amazon Machine Image (AMI): Ubuntu Server 22.04 LTS (HVM), SSD Volume Type
• Instance Type: g5.2xlarge (8 vCPU, 32 GiB Memory, NVidia A10 GPU)
• Key pair: MemVerge.ai-cluster-key
• Network Settings:
  • Firewall (security groups): Select existing security group > mai-sg
• Configure Storage:
  • 1 x 100 GiB gp2 Root volume (Not encrypted)
• Click "Launch instances"

Wait for the instances to be fully provisioned. This may take several minutes.