Cloud Instance Recommender

🧩

Custom Rules

📄

Sample CSV Template

▲

View and download the standard CSV format for optimal auto-mapping:

VM Name,App Name,CPU Count,Memory (GB),CPU Utilization,Memory Utilization,GCP Region,ENV,OS,Workload,Compliance,Min Gen,Exclude,Include Only,Current Instance Type
web-server-01,Storefront,4,16,45,60,us-central1,Production,Linux,Web Server,,,,,n2-standard-4
db-server-02,Billing,8,32,70,80,us-west1,Production,Windows,Database,PCI,,"Burstable,GPU",,n2-standard-8
app-server-03,Billing,2,8,35,45,europe-west1,Dev,Linux,General,,,,,e2-standard-2

Optional columns — click to see rule descriptions

Column	Values	Rules triggered
ENV	Production / Staging / Dev / Test	1a f1/g1/e2-shared burstable excluded (Prod/Staging) · 1b Current-gen only (Prod) · 1c ≥2 vCPUs (Prod) · 1d ≥4 vCPU for DB/Web workloads (Prod)
OS	Linux (default) / Windows	Windows → ARM (T2A series) excluded
Workload	General · Database · SQL Server · Web Server · Cache · ML/AI · Batch · HPC · SAP	Preferred series sorted first: Database→M1/M2/M3/M4 · SQL Server→M1/M2/M3/M4 · Web→N2/E2/N4 · Cache→M1/M2/M3 · ML/AI→A2/A3/G2 · Batch→C2/C3 · HPC→H3/C2 · SAP→M1/M2/M3/M4
Compliance	PCI / HIPAA / SOC2 / FIPS	1b Current-gen only · No burstable
Min Gen	n2 / n4 (family name)	Excludes machine families older than the specified generation (e.g. n4 → only N4/C4 and newer)
Exclude	Comma-separated type names (e.g. Burstable,GPU)	Excludes those instance types for this row only — quote the cell when it names more than one

All columns are optional — blank cells use safe defaults.

Current Instance Type fires no rule and does not affect sizing by default — only in cloud-to-cloud mode does a row without CPU Count / Memory (GB) get sized from it. It records what the VM runs on today (e.g. n2-standard-4) and is carried through to the preview and every export, beside the recommendation that would replace it.

📊 Download Excel Template

📁

Upload CSV File

📤

Click to upload CSV or Excel (.xlsx) file

or drag and drop your file here

☁️

Cloud Provider Selection

Select Cloud Providers:

🔴 GCP

⚙️

Recommendation Type

Like-to-Like Recommendation
Recommends instances with equal or higher CPU and memory (cheapest option)

Optimized Recommendation
Recommends instances based on actual utilization data

Both
Generates both like-to-like and optimized recommendations

🧠 Optimization Settings

Size against: Averages hide bursts: a VM averaging 20% with a p95 of 85% is not a downsize candidate. p90, p95, p99 and Peak use the matching CSV columns; a row without them falls back to what it has, and the basis used is reported per row in the Sized On column.

Optimization Mode:

CPU-Based

Memory-Based

💻 CPU Utilization Ranges:

CPU Downsizing (N-1): % to %

CPU Keep Same (N): % to %

CPU Upsizing (N+1): % to 100%

🧠 Memory Utilization Ranges:

Memory Downsizing (N-1): % to %

Memory Keep Same (N): % to %

Memory Upsizing (N+1): % to 100%

📐

Sizing Source

🔀 Cloud-to-cloud sizing Size a row from its Current Instance Type when it has no CPU Count / Memory (GB). Explicit CPU and memory always take precedence.

🔧

Advanced Instance Filtering (Optional)

▼

✨ Current Generation Only Include only latest generation instances (recommended for better performance)

🏷️ Restrict to specific instance family names Choose instance categories (General purpose, Compute optimized, and so on)

🔧 Restrict to specific processor types Choose preferred processor architectures

📋 Restrict to specific instance families Choose main instance families (E2, N1, N2, C2, etc.)

❌ Exclude specific instance types Exclude certain categories of instances

🧠 Rule Engine Defaults

Set page-level defaults for all rows. Per-row CSV columns override these. Rules merge with the filters above — conflicts are flagged in red.

ENV

OS

Workload

Compliance

Current-Generation Hardware Confidential Computing

Min Instance Gen

Rule descriptions ▸

Rule	Effect
ENV=Production	Excludes shared-core (e2-micro/small), enforces current-gen, min 4 vCPU for web/DB
OS=Windows	De-prioritises T2A (ARM) families; prefers Intel or AMD
Workload=Web Server	n2/c2 first; Database → n2-highmem/m1 first; ML/AI → a2/g2 first
Compliance=HIPAA/PCI	Excludes shared-core; enforces current-gen
Min Gen=N2+	Excludes N1 (1st gen); keeps N2, C2, N2D and newer only
Workload=SQL Server	Enforces a 4-vCPU licence floor when it can — SQL Server is licensed per core (minimum four), so a smaller box saves no licence cost. If no eligible candidate is that large, the fallback pick is kept and the rule is reported as not applied.
Workload=ML/AI (GPU)	Requires a GPU/accelerator instance; every other workload excludes accelerators, so a GPU box is not recommended by accident. If the pool has no accelerator to offer, the fallback pick is kept and the rule is reported as not applied.

🚀

Generate Recommendations

📊

Usage Statistics

0

🔧 Tool Uses

Total recommendations generated

0

🖥️ VMs Processed

Total VM instances analyzed

💾

Cloud Instance Recommender - GCP