# DeepWaste™ Technology | PointFive Canonical: https://www.pointfive.co/products/pointfive-os/deepwaste Hundreds of optimization types across cloud providers and data platforms. Five layers of waste detection, from surface-level to deep architectural inefficiencies. See how DeepWaste finds cloud and AI waste others miss. # Hundreds of ways to cut your cloud and AI bill. PointFive's DeepWaste Detection Engine finds cloud waste that no other tool can, from quick wins to deep architectural inefficiencies, across AWS, Azure, GCP, OCI, Snowflake, Databricks, Kubernetes, and AI platforms. Evaluate opportunities in the context of your workloads and verify the outcome after remediation. [Book a demo](https://www.pointfive.co/request-demo) A cross-section through limestone terrain reveals the layers and relationships within. Hundreds Optimization Types Multi-cloud Coverage Dozens Services Covered 5 Layers of Waste ## Five layers of infrastructure waste. Explore five layers of waste, from common resource issues to architectural inefficiencies. Use detection findings and available context to investigate opportunities, review remediation, and measure savings after changes. - ### Layer 1 / Surface-Layer Waste Orphaned, inactive, and abandoned resources still incurring charges. Investigate usage and dependencies before deciding which resources to change, then measure the effect of the implemented remediation. - Common detection depth - Unused Compute Disk Attached to Stopped VM - Inactive Kinesis Stream - Orphan RDS Instance Snapshot - ### Layer 2 / Configuration Optimization Misaligned settings versus real usage, storage tiers, disk types, capacity modes, instance sizes, and caching configurations. Validating these at scale demands correlation of performance metrics, usage patterns, and workload context to produce recommendations engineers actually trust and act on. - High Volume detection depth - Inefficient FSx Volume Configuration - Bedrock Prompt Caching Underutilization - Suboptimal Standard Storage for DynamoDB Table - ### Layer 3 / Application Driven Waste Waste created by how applications use cloud services: API call patterns, model selection, transfer routing. Invisible without deep application-level analysis and multi-source data correlation. - Deep detection depth - Suboptimal ListBucket API Usage in Spark S3 Buckets - Expensive Bedrock Model Used for Simple Tasks - Redundant S3 Transfer Acceleration from Nearby Locations - ### Layer 4 / Data & Lifecycle Drift Data that has aged out of its original purpose but stays in expensive storage tiers. Retention policies that no longer reflect actual access patterns. Requires continuous monitoring of data access frequency and intelligent lifecycle recommendations. - Deep detection depth - Archival S3 Bucket Storing Objects in Non-Archival Classes - Long-Retained RDS Cluster Manual Snapshot - Suboptimal Retention for CloudWatch Log Group - ### Layer 5 / Architectural Inefficiencies Suboptimal infrastructure design causing structural cost overhead, routing decisions, region placement, cluster sizing. Requires VPC Flow Logs, snapshot side-scanning, and virtual cost allocation techniques unique to PointFive. The hardest to find, often the most impactful to fix. - Deepest detection depth - Suboptimal Region for an EC2 Instance - Expensive SQS Traffic Traversing Through a NAT Gateway - Excessive DaemonSet Overhead in EKS Cluster ## Every layer matters. Every layer compounds. Real savings come from continuous detection across all five layers, with validated context, root-cause analysis, and engineering-grade remediation at every depth. Surface findings fund the deeper work. Deeper layers prevent waste from recurring. What sets PointFive apart isn't just finding waste, it's the validated context, multi-source correlation, and remediation tooling that turns findings into action across every layer. Deeper layers require VPC Flow Logs, snapshot side-scanning, and virtual cost allocation, techniques unique to PointFive. ## The broadest and deepest coverage. DeepWaste analyzes dozens of cloud and AI services across all major providers, with new services added continuously. ### Compute Services include EC2, Azure VMs, GCP Compute ### Kubernetes Services include EKS, AKS, GKE, ECS ### Databases Services include RDS, DynamoDB, Cosmos DB ### AI & ML Services include Bedrock, SageMaker, Azure OpenAI ### Data & Analytics Services include Snowflake, Databricks, BigQuery ### Storage Services include EBS, S3, Blob Storage ### Serverless Services include Lambda, Functions, Cloud Run ### Networking Services include VPC, CloudFront, Load Balancers ### Observability Services include CloudWatch, Datadog, Logging ### Streaming Services include MSK, Kinesis, Pub/Sub ### Security Services include Secrets Mgr, App Config ## A taste of hundreds of optimizations. A curated sample of what our DeepWaste engine detects, each backed by usage patterns, configuration analysis, and validated remediation playbooks. The real catalog runs much deeper. ### Expensive VPC NAT Gateway Deployment AWS Detects workloads routing massive data through NAT Gateways when free alternatives like Gateway Endpoints exist. Estimate the impact using applicable routing, endpoint, and data-transfer charges. Easy Fix Networking Validate Rollout ### Schedulable EC2 Instances AWS Identifies instances with low off-hours activity that can be automatically started and stopped on a schedule. Eliminates waste from resources running 24/7 when only needed during business hours. Automated Compute ### Serverless-Ready OpenSearch Domains AWS Identifies low-usage provisioned OpenSearch domains for a serverless comparison. Evaluate capacity charges, supported features, workload patterns, and migration requirements before assuming a saving. Quick Migration Search ### Suboptimal Azure Disk Types Azure Investigates Azure managed-disk configuration and usage to identify potential tier changes. Validate latency, IOPS, throughput, and platform compatibility before a change. High Impact Storage ### GPU Instance Rightsizing AWS Identifies GPU instances (P4d, P5, G5, Inf2) running ML training and inference workloads at low GPU utilization, recommending smaller instance types or spot-based alternatives without impacting throughput. AI Compute GPU ### Vertex AI Training Job Optimization GCP Detects Vertex AI custom training jobs using oversized machine types or running without preemptible/spot instances, common in ML experimentation workflows where cost discipline is often overlooked. ML Training Spot-Ready Showing 6 of 10 detections in this category. Want to see what's hiding in your infrastructure? ## Real customer impact Turn detection findings into prioritized engineering work, then measure the effect of each implemented change against actual usage and billing. [See your waste](https://www.pointfive.co/request-demo) ## Finding waste is step one. Fixing it is the whole point. Investigate findings, prepare fixes for review, and delegate approved remediation to Coworkers in PointFive OS. Measure savings after changes are implemented. ### AI Coding Agents AI coding agents use available infrastructure context to generate proposed remediation for engineering review. ### Built-In Workflows Connect directly to Jira, Slack, ServiceNow, and your existing ticketing systems. Findings flow into your team's natural workflow, no context switching required. ### Infrastructure-as-Code Terraform and CloudFormation fixes ready to merge. Every remediation maps to your IaC stack, so changes go through your existing review and deploy pipeline. ### Human Oversight at Every Step Root-cause analysis, impact assessment, and safe rollback paths, all included. AI generates the fix. Your engineers review, approve, and deploy with full confidence. ## From finding to fix in minutes, not months. Traditional cost optimization ends with a spreadsheet. PointFive delivers validated findings with actionable remediation, so engineering teams ship savings instead of triaging alerts. [See it in action](https://www.pointfive.co/request-demo) ## Dozens of services. Deep analysis on every one. Not just visibility: DeepWaste analyzes usage patterns, access frequencies, and workload characteristics across dozens of services organized by infrastructure domain. - ### Compute - Amazon EC2 - Azure VMs - Compute Engine - OCI Compute - WorkSpaces - ### Kubernetes & Containers - Amazon EKS - Azure AKS - Google GKE - OpenShift - Amazon ECS - AWS Fargate - Azure Container Instances - Azure Container Apps - Cloud Run - ### Databases - Amazon RDS - DynamoDB - ElastiCache - OpenSearch - Azure SQL - Azure PostgreSQL - Azure Cosmos DB - Azure Cache for Redis - Cloud SQL - Memorystore - Bigtable - OCI Autonomous DB - MongoDB Atlas - ### Data Warehouses & Analytics - Snowflake - Databricks - Amazon Redshift - BigQuery - Azure Synapse - Azure Fabric - ### Storage - Amazon EBS - Amazon S3 - FSx / EFS - Azure Disks - Azure Blob Storage - Azure NetApp Files - Persistent Disk - Cloud Storage - Filestore - OCI Block Volume - OCI Object Storage - ### Serverless & Functions - AWS Lambda - Azure Functions - Cloud Functions - OCI Functions - ### Networking & CDN - VPC / NAT - Elastic Load Balancing - CloudFront - Route 53 - API Gateway - Azure Networking - Azure Front Door - Cloud CDN - OCI Networking - ### Monitoring & Observability - CloudWatch - CloudTrail - VPC Flow Logs - Azure Monitor - Log Analytics - Cloud Logging - Datadog - ### Streaming & Messaging - Amazon MSK - Kinesis - SQS / SNS - Amazon MQ - Azure Event Hubs - Azure Service Bus - Pub/Sub - ### AI & Machine Learning - Amazon Bedrock - SageMaker - EC2 GPU Instances - Azure OpenAI - Azure Machine Learning - Azure GPU VMs - Vertex AI - GCP GPU Instances - Anthropic Claude - OpenAI API - ### Security & Configuration - Secrets Manager - Azure App Config - App Service - Elastic Cloud ## Built different. Detects different. ### Agentless architecture Agentless discovery uses read-only access. Infrastructure changes require customer-authorized write permissions and approval before execution. ### Research-driven detection PointFive Labs researches infrastructure inefficiencies and develops detections using configuration, usage, and cost evidence. ### Multi-source data correlation We don't just read your bill. We correlate billing data with CloudWatch metrics, VPC Flow Logs, CloudTrail activity, and direct API state, finding waste that single-source tools miss. ### Agentic Remediation AI coding agents deliver contextual recommendations with human-curated remediation playbooks. Teams can investigate causes, assess estimated impact, and review remediation plans before approving execution. ## Frequently asked questions ### PointFive's DeepWaste™ detection engine identifies over 500 types of cloud and AI waste across compute, storage, databases, networking, serverless, AI/ML, data & analytics, observability, streaming, security, and containers. Each detection is backed by usage patterns, configuration analysis, and validated remediation playbooks, not just billing thresholds. ### Opportunities are prioritized, actionable savings recommendations surfaced by the DeepWaste engine. Each opportunity includes the resource affected, estimated savings, risk level, remediation steps, and can be assigned to a team member or pushed to Jira/ServiceNow. Opportunities are continuously updated as your environment changes. ### PointFive's anomaly detection uses machine learning to establish baseline spending patterns and alerts you when costs deviate significantly. Each anomaly includes the affected service, cost impact, and detailed analysis showing actual vs. expected spending, so you can investigate and resolve cost spikes before they compound. --- Source: the public page above. Product screenshots and illustrative interfaces are examples, not live customer data.