Recipes

A recipe is the set of parameters for the transfer engine that uses a configuration—compression, chunk size, transport, timeouts, post-deployment actions, etc. The Recipes page (sidebar → Recipes) is where you create, refine, and reuse them.

Provision

A folder tree on the left organizes the recipes just like Images and Infrastructure — create, rename, move, lock, search, and Delete folders by dragging them or right-clicking them. The table in the The right-hand side lists all the recipes with their key parameters at a glance, and a set of actions per row.

Action	What does it do?
★ / ☆	Mark a recipe as default — the one that is preselected when you open the drop-down menu. Only one recipe can be set as the default at a time.
Duplicate	Open the editor that's already loaded with this recipe, ready to save as a new one.
Edit	Open the editor for this recipe.
Export	Download this recipe as JSON.
Delete	Delete the recipe (with confirmation).

A bulk action bar appears as soon as you select more than one row, allowing you to export or delete multiple recipes at once. New recipe opens the full editor (below); its drop-down menu does too offers New Recipe with Assistant — a guided alternative questions, described below — and a way to create a new folder.

The Recipe Editor, Tab by Tab

Basic

• Name (required) and Description (optional).

• Base image — No image (the master uses whatever is on its disk) live — no need to choose in advance) or With image (an image of the (The catalog is selected at the time of deployment, or specified below.)

  • When using an image, image decompression controls how it is served The master file is a compressed image: on the fly (it decompresses while transmits, maintaining a configurable chunk cache in RAM — 2048 chunks by default) or pre-extracted (extracted to a (temporary raw file before serving it).

• Rendering Resource — choose the image or master from which to clone it when you expand it (by default), or set it now in the recipe itself so that all deployments that use it always use the same one source.

• Deployment Transport — P2P (the default; peers are redistribute blocks among themselves, BitTorrent-style) or Multicast (UDP: a single transmission reaches all listeners on the segment — see the "Multicast" tab below).

• Incremental diff — compares each client's current disk with the It performs a hash-based master check and only transfers the blocks that have actually changed; ideal for redeploying an operating system patch instead of a complete image.

• Simulation (dry run) — it receives and unpacks everything normally, but never writes to disk, to test a recipe without altering the data in no customers.

• Advanced Options (collapsed by default):

  Field	Values	What is it used for?
  Compression	lz4 (fast, default) / zstd (best ratio)	Balance between CPU and network/disk usage.
  Level	It depends on the algorithm (e.g., lz4 1–12, zstd 1–22)	Higher = smaller output, more CPU.
  Hash	blake3 (fast, default) / sha256 (compatible)	Integrity-checking algorithm for blocks and disk verification.
  Chunk size	Number + unit (4M by default)	Smaller chunks are distributed more quickly among P2P peers; larger chunks work better on fast, stable networks.
  Chunk Selection	random (default) / rarest / sequential	Which block does a client request next—maximum diversity, the least available first (good for unstable networks), or in order (gentler on slow disks)?
  Master's Program Overview	mdns (default) / static	Clients either automatically connect to the master or connect directly to a fixed address.
  Diagnostic Metrics	on/off	It collects detailed telemetry by node (disk latency percentiles, CPU/RAM, RTT/TCP retransmissions, NIC errors) and displays it in real time during deployment.
  Save telemetry to disk	disabled / .jsonl / .csv	Save that telemetry for later analysis.

  The exact list of compression algorithms, hash algorithms, and The chunk selection strategies provided here always match compared to what the installed netdd binary actually supports — Stranola does It checks once at startup and adjusts the form accordingly.

Master's Degree

Settings for the machine (or the server itself) that acts as the source of the transfer.

Field	By default	What is it used for?
Max. Mbps	0 (unlimited)	Limits the total output bandwidth of the master.
Max. Concurrency	0 (unlimited)	It limits the number of customers the master can serve directly at one time.
Fair-share distribution	disabled	Distribute the master's bandwidth evenly so that no fast client monopolizes the rest.
Fair penalty (ms)	10	A delay applied to a customer who exceeds their fair allocation quota.
Super-seed	disabled	The master sends each block only once in total, relying on P2P to propagate it—this reduces disk I/O from N× to approximately 1×.
Super-seed reserve(s)	30	How long to wait before resending a block that no customer has confirmed receiving; 0 = never resend it.
Scattered (sparse)	Enabled	Skip empty disk blocks instead of transferring them—this typically saves 30–60% in traffic on partially empty disks.
Upon completion (master's degree)	None	It does nothing, reboots, shuts down, or runs a custom shell command when the master process finishes.
Timeout with no clients (s)	0 (wait forever)	Terminate the deployment if no client has connected during this time.
Max. concurrent clients	0 (unlimited)	Limits the number of clients served simultaneously.

Network

Field	By default	What is it used for?
Receive Limit per Client (MB/s)	0 (unlimited)	Limits a client's total download speed (master + P2P combined), which is useful for protecting slow uploads.
TCP Timeout (ms)	10000	How long can a connection remain idle before it is considered dead?
DSCP/QoS	0 (unmarked)	Mark outgoing packets for network-level prioritization (46 = highest priority, EF).
Monitoring interval (s)	2	How often is the network's health monitored during deployment?
Error threshold (%)	1.0	Error rate above which a link is marked as degraded.
Collision threshold (%)	0.5	Collision rate above which a link is marked as degraded.

Client

Field	By default	What is it used for?
Check after typing	Enabled	It calculates the hash of the entire disk after cloning and compares it to the original.
Tit-for-tat	disabled	Prioritize uploading data to the peers who are sending you the most data in return, thereby discouraging those who only receive data.
No resumption	disabled	Ignore any previous resume status and start the transfer from scratch.
Seed time(s)	0 (disappears when finished)	How long does a terminated client continue to provide data to other peers before disconnecting?
Local RTT (ms)	0.5	A round-trip time below which two peers are treated as if they were on the same switch, enabling a deeper pipeline between them.
Master's program enrollment delay (ms)	50	Grace period granted to P2P peers before they have to request data directly from the master.
Pipeline depth	16	How many chunk requests a client keeps active at any given time.
Stagnation timeout (s)	0 (internal default: 60 seconds)	Abort the transfer if no chunks arrive within this time.
Location of the master's program	automatic discovery	Optionally, set the master's host:puerto directly instead of revealing it.
Upon completion (client)	None	Do nothing, restart, shut down, check the disk, or run a custom shell command when a client terminates.

Preflight

An optional network test that runs on all clients Before the actual cloning begins, to check the wiring faulty, half-duplex connections, or insufficient bandwidth soon in rather than halfway through the deployment.

Field	By default
Enable Preflight	disabled
Trial duration per customer(s)	10
Maximum error rate (%)	2.0
Exclude slow clients below (Mbps)	0 (do not exclude)

Multicast

This is displayed when you select Multicast as the deployment transport in the "Basic" tab. Instead of each customer requesting the data individually, The master sends each block once to a multicast group, and the switch The network repeats it to everyone listening — the master's bandwidth It ceases to be a bottleneck as you add more customers.

Field	By default	What is it used for?
Multicast group	—	e.g., 239.0.0.1 (the range 239.x.x.x is recommended).
Data port	7100
Port NACK	7101	Customers use it to request the retransmission of missed blocks.
TTL	1	1 keeps traffic within the local network segment.
Max. Speed (Mbps)	400	0 removes the limit (not recommended — a warning is displayed).
Multicast interface	automatic
FEC Error Correction	disabled	Reed-Solomon error correction (forward error correction) — repairs lost packets without requesting a retransmission.
FEC Redundancy (%)	10	Displayed only when FEC is enabled; higher values allow for more error tolerance at the cost of increased bandwidth.
Hybrid Mode	Enabled	Keeps the P2P/TCP connection alive alongside multicast, as a recovery mechanism and for clients that join late or are experiencing issues—it is recommended to keep this enabled at all times.
RX Buffer (MB)	32	Kernel receive buffer size; increase it for faster networks (10GbE → ~96, 25GbE+ → ~192).
RX Lot	64	Package Receipt Batch Size.
Lot TX	64	Batch size for package shipments.

The Smart Assistant

New Recipe Assistant creates a recipe for you by asking questions simple ones instead of adjusting the parameters directly — a good a good place to start if you're not sure what to change.

1. Hardware — client disk type (SSD / fast HDD / slow HDD) / unknown), RAM per machine, environment (physical / virtual / hybrid), and Approximately how many client machines are there?
2. Network — link speed, stability, whether there are multiple switches which files are involved, whether they are managed, and how much free space is available on the disk of the master's program.
3. Objective — first clone, incremental update, simulation, or just check; and what to do when a client disconnects (nothing, restart, shut down, check disk).
4. Fine-tuning — a set of optional switches: keep the Stable speed on SATA SSDs; apply counterpressure in the event of congestion disk/link, abort if SMART reports a failing disk, diagnostics in live, encrypt traffic, run a preflight test, reciprocity tit-for-tat, check after writing — plus a choice of algorithm hash and an auto-shutdown timer for the master unit.
5. Summary — name the recipe, provide an explanation in plain language a clear explanation of what was chosen and why; optionally, it displays the exact resulting parameters, and generates the recipe.

The assistant is a standalone deterministic rule engine—it does not call no AI service, and it generates a standard recipe that you can then Continue editing and refining in the full editor.

Export and Import

Export (from the toolbar, for all recipes, or by row/selection (for some only) downloads a flat JSON file with all the recipe parameters — everything specific to this facility Specific details (internal IDs, image references) are intentionally omitted so that the file is portable.

Import reads that file and creates new recipes from it. Any installation-specific information is always regenerated when the import (it is never read from the file), and a name collision occurs automatically handles it — if "Windows 11 base" already exists, the copy The imported name becomes "Windows 11 base (2)," and so on — like this Importing never accidentally overwrites an existing recipe.