Deviseur Azure

Turn a plain hardware spec into an Azure quote. Live prices come from the
Azure Retail Prices API (https://prices.azure.com/api/retail/prices); the
spec→flavor mapping uses the local catalog in references/vm-catalog.json.

Defaults: region francecentral, currency EUR. Override with --region / --currency.

Scripts are under scripts/. Reference data is under references/.

Sizing rule (when the spec doesn't match an Azure standard size)

Customer specs often don't line up with Azure's fixed vCPU grid (1, 2, 4, 8,
16, 32 …). When the spec falls between sizes, apply this rule:

• RAM must meet-or-exceed the customer's RAM — never give them less memory.
• vCPU may sit just below the customer's vCPU — round the core count down
  to the nearest Azure size rather than up.
• Prefer the D family (general-purpose — Azure's "Standard" tier) when a
  D-series flavor satisfies the RAM floor; fall back to E (memory) when only E
  can meet the RAM at that core count, then F/B.

Example: a 3U4G request (3 vCPU, 4 GiB) → recommend D2s_v5 style sizing
(2 vCPU, RAM ≥ 4 GiB, D-series). The 3rd core is dropped (no 3-vCPU Azure size);
RAM is honored; D-series wins because it's the Standard general-purpose family.

This rule is implemented in pick_flavor / target_vcpu (scripts/azure_lib.py)
and drives both the interactive proposal and the RVTools batch mapping.

Disk-sizing rule (when no tier matches exactly)

Azure managed-disk tiers are discrete (4, 8, 16, 32, 64, 128, 256 … GiB). When a
requested size has no exact tier, minimize cost by minimizing the disk size
within technical tolerance:

• Consider tiers within 80%–120% of the requested size and pick the
  smallest one (cheapest — this may sit just below the request, e.g. an
  80 GiB request → a 64 GiB tier, since 64 = 80% of 80).
• If no tier falls in 80%–120%, step up to the smallest tier above 120%.

Examples: 80 GiB → 64 GiB tier (in-band, cost down); 100 GiB → 128 GiB
(64 is below 80%, 128 is the smallest above 120%); 300 GiB → 256 GiB (in-band).

Implemented in select_disk_tier / disk_tier_for_size (scripts/azure_lib.py);
a tier below the request is flagged undersized. Drives every disk price across
the quote, RVTools, and fleet-export paths.

Workflow (TWO steps — always interactive)

Step 1 — Propose flavors

Parse the spec into vCPU and RAM (GiB). Common shorthands:

• 4U8G / 4C8G / 4 vCPU 8GB → --vcpu 4 --ram 8
• 8U64G → --vcpu 8 --ram 64

Run:

python3 scripts/propose_flavors.py --vcpu 4 --ram 8 --region francecentral

Display the full proposal table, then ask the user to pick a flavor (#).
Do NOT skip ahead to the quote — the user chooses first.

The table shows candidates across families (Burstable / General / Memory /
Compute), the recommended D-series first (per the Sizing rule above — RAM
meets-or-exceeds, vCPU floored to the nearest size just below). Every row lists
the Linux monthly price across all four commitment options — PAYG, Spot, 1yr
Reserved, 3yr Reserved (N/A where a family has no such rate).

Step 2 — Quote the chosen flavor

Once the user picks a flavor, capture the disk requirement from the original
spec (e.g. "25G SSD" → --disk-size 25 --disk-type premium-ssd) and run:

python3 scripts/query_quote.py --sku Standard_F4s_v2 --region francecentral \
    --disk-size 25 --disk-type premium-ssd --os linux --qty 1

Display the full output: VM compute table, managed disk, and combined total
(with the 1yr-reserved saving note).

Batch mode — RVTools import

When the user provides a VMware RVTools export (.xlsx) and wants the whole
estate sized, skip the interactive flavor pick and run the batch analyzer. It
reads the vInfo sheet and, per VM, maps the allocated vCPU/RAM/disk to a
D-preferred Azure flavor following the Sizing rule above (RAM meets-or-
exceeds; vCPU floors to the nearest size just below), then prints per-VM mapping
plus rollup totals (PAYG + 1yr Reserved) for a target region.

python3 scripts/analyze_rvtools.py inventory.xlsx --region francecentral
# include powered-off VMs / write a Markdown report:
python3 scripts/analyze_rvtools.py inventory.xlsx --include-poweredoff --output
# try it with the bundled sample:
python3 scripts/analyze_rvtools.py examples/rvtools-sample.xlsx

OS is auto-detected per VM from the sheet's OS column (Windows rows are
priced with the Windows rate; --os is only the fallback when the column is
absent). Tell the user up front: RVTools has no Azure region (region is a
target you pick, default francecentral) and is a point-in-time allocation
snapshot, not performance data — for true right-sizing use Azure Migrate.
Powered-off VMs and templates are excluded by default. Requires openpyxl
(pip install openpyxl). A sample export lives at examples/rvtools-sample.xlsx.

Scripts

propose_flavors.py

Option	Default	Notes
--vcpu / -v	required	vCPU count
--ram / -m	required	RAM in GiB
--region / -r	francecentral	Azure region
--currency / -c	EUR	Currency code
--max	6	Max candidates

query_quote.py

Option	Default	Notes
--sku / -s	required	Chosen flavor (Standard_ prefix auto-added)
--region / -r	francecentral	Azure region
--currency / -c	EUR	Currency code
--os	linux	linux or windows (headline total)
--qty / -q	1	Number of identical VMs
--disk-size	none	Disk size in GiB (rounded up to next tier)
--disk-type	premium-ssd	premium-ssd, standard-ssd, standard-hdd
--output / -o	none	Write the quote to a Markdown file. Bare flag auto-names it quotes/quote-<sku>-<region>-<date>.md; pass a PATH to choose the location. Output is still echoed to the screen.

When the user asks to save/export/deliver the quote as a file, add --output
(bare for an auto-named file under quotes/, or with a PATH).

For a Windows headline (--os windows), the total table splits each
reserved term into no AHB (Windows Server licence stays at PAYG on top of
the reserved compute) and with AHB (Azure Hybrid Benefit — only the reserved
compute is billed). Always quote both so the customer sees the licence upside.

export_fleet_xlsx.py

Export a multi-group fleet quote to a multi-sheet Excel workbook (Selection,
Per-Group, Fleet Total, TCO). Use when the user wants several different specs /
OSes / quantities priced together, or asks for Excel/.xlsx output. Pulls
live prices, carries the Windows AHB / no-AHB split, and builds a 1/2/3-year TCO.
Define the fleet with one repeatable --group "SPEC,OS,QTY[,SKU]" per distinct
group; the SKU is auto-picked from the sizing rule unless pinned as a 4th field.

python3 scripts/export_fleet_xlsx.py \
    --group 4U8G,windows,10 --group 3U6G,linux,10 --group 5U11G,linux,10
# pin a SKU, set region / output path:
python3 scripts/export_fleet_xlsx.py --group 8U64G,linux,5,E8s_v5 \
    --region westeurope --output /tmp/quote.xlsx

Option	Default	Notes
--group / -g	required	SPEC,OS,QTY[,SKU]; repeat per group. SKU optional (auto-picked)
--region / -r	francecentral	Azure region
--currency / -c	EUR	Currency code
--output / -o	auto	.xlsx path; default auto-named under quotes/

The 2-year TCO column is a time horizon (a 1yr reservation renewed) — Azure
VM reservations exist only in 1yr and 3yr terms.

analyze_rvtools.py

Option	Default	Notes
file	required	Path to the RVTools .xlsx export
--region / -r	francecentral	Target Azure region
--currency / -c	EUR	Currency code
--os	linux	Fallback OS when the sheet has no OS column (OS is auto-detected per VM otherwise)
--disk-type	premium-ssd	Disk type applied to every VM
--sheet	auto	Worksheet name (auto-detects vInfo)
--include-poweredoff	off	Include powered-off VMs
--include-templates	off	Include VM templates
--output / -o	none	Write Markdown report; bare flag auto-names under quotes/, or pass a PATH

Pricing notes (important for correct quotes)

• Spot is returned by the API as type=Consumption with "Spot" in the
  meter name (not type=Spot). The scripts handle this.
• Reservation retailPrice is the TOTAL for the whole term; the scripts
  convert it to an effective hourly/monthly rate.
• Linux and Windows+AHB share the same compute price.
• Windows licence = Windows PAYG − Linux PAYG. Reservations discount compute
  only, so a reserved Windows VM is reserved compute + licence at PAYG without
  AHB, or just reserved compute with Azure Hybrid Benefit. Quote both.
• Reservations come only in 1yr and 3yr terms — there is no 2-year RI. A
  "2-year" figure is a time horizon (1yr reservation renewed), not a product.
• Disk is mapped to a tier by the disk-sizing rule below (not always a
  round-up). Disks have no reservation discount.
• Monthly = hourly × 730; annual = hourly × 8760.

Reference data

• references/vm-catalog.json — SKU → family / vCPU / RAM. Extend this to add
  flavors (e.g. GPU N-series, constrained-core sizes).
• references/disk-tiers.json — managed-disk size tiers and API sku names.
• references/service-mapping.md — user term → Azure serviceName and region
  aliases, for ad-hoc queries beyond VMs/disks.