Migration glossary

Plain-English definitions of the terms used across OffVendor

Migration planning is full of acronyms. Here's what the ones used on this site actually mean, in the context of deciding whether and how to move off a product. For how we turn these into numbers, see the Methodology.

Lock-in and what it costs

The terms behind every number on this site: what ties you to a vendor, and how we total up the price of staying.

Vendor lock-in: The cost and difficulty of leaving a product, proprietary formats, licensing tied to hardware, integrations, or skills, that lets a vendor raise prices with little risk of you switching. Reducing it is the whole point of having an exit plan.
TCO (Total Cost of Ownership): The full 3-year cost of running a product: licensing or subscription, support, and the one-time migration effort to adopt it, not just the sticker price. OffVendor compares TCO between the source and target products.
Sizing unit: The billable quantity a product is priced on, CPU core, socket, host, TB protected, user, endpoint, vCPU, or token volume. Picking the right unit is what makes a cost comparison fair.
Billable cores: For per-core licensing (common in virtualization and databases), the number of physical cores that actually count toward the license after any vendor minimums or core-factor rules are applied.
Support tier: An optional paid support contract (Community, Standard, or Premium) layered on top of licensing. Matters most for open-source targets, where the software is free but production support usually is not.
Illustrative pricing: Indicative figures compiled from public list prices and street pricing, shown as editable defaults so you can model your own scenario. They are starting points, never guaranteed quotes, see the Methodology page.

Scoping a move

How we describe a single vendor-to-vendor move and estimate the work it takes.

Migration path: A specific 'Product A → Product B' move (for example VMware → Proxmox). Each path has its own cost model, effort estimate, tooling, and step-by-step runbook.
Source / Target: The product you are moving off (source) and the product you are moving to (target). The calculator models both sides over three years.
Migration effort: An estimate of the one-time work to move a path, expressed as a Low / Medium / High label and an indicative number of weeks. Driven by how locked-in the source is and how involved the target's adoption tends to be.

Executing the cutover

The mechanics of actually moving workloads, and the resilience targets that constrain how you do it.

Cutover: The moment you switch production traffic from the source to the target, typically a final data sync, a write freeze on the source, a DNS or load-balancer change, then validation, with the source kept as fallback.
Rollback: The pre-planned path back to the source if cutover validation fails. A migration isn't production-ready until the rollback has been rehearsed.
RPO (Recovery Point Objective): The maximum acceptable amount of data loss, measured in time, e.g. 'no more than 15 minutes of data.' It sets how often you must replicate or back up.
RTO (Recovery Time Objective): The maximum acceptable time to restore service after an outage. RPO and RTO together define your resilience target and shape the migration's DR design.
CDC (Change Data Capture): A replication technique that streams ongoing changes from a source database to a target after an initial bulk load, so the two stay in sync until cutover with minimal downtime. Used by tools like AWS DMS and Azure DMS.
V2V / P2V: Virtual-to-virtual and physical-to-virtual conversion, moving a workload's disk and configuration from one hypervisor (or bare metal) to another, a core step in virtualization and repatriation migrations.
Landing zone: A pre-built, governed foundation in a target cloud, accounts/subscriptions, networking, identity, and guardrails, stood up before any workload is migrated.
Bridging (messaging): Running a temporary replication link between an old and new message broker (e.g. Kafka MirrorMaker 2) so producers and consumers can move across without losing messages.

Which way the workload is going

Cloud moves run in both directions, and the direction changes which costs bite.

Egress: The fee cloud providers charge to move data out of their network. Large cross-cloud or cloud → on-prem transfers can incur significant egress, which must be modeled in the migration's business case.
Repatriation: Moving workloads from a public cloud back to on-prem or a private cloud, increasingly common when steady-state demand turns out cheaper on owned hardware than on metered cloud.
Cloud adoption: Moving on-prem workloads into a public cloud to gain elasticity and managed services, trading capital expense for operating expense. The reverse of repatriation.

Leaving a proprietary AI API

Two terms that do most of the work in the LLM category, where lock-in is about the API surface rather than the licence.

Open-weight model: A large language model whose weights are publicly available (e.g. Llama, Mistral, DeepSeek), so it can be self-hosted on your own inference stack instead of called through a proprietary per-token API.
Gateway (LLM/API): An OpenAI-compatible proxy in front of one or more models, so applications can switch providers, or fall back instantly, by changing a base URL rather than rewriting code.

The security segments

"Cybersecurity" is seven different buying decisions with seven different billing units. These are the acronyms that separate them.

EDR / XDR: Endpoint Detection and Response is the agent on the device that records behaviour and can isolate, kill, or roll back. Extended Detection and Response adds identity, cloud, and network telemetry to the same correlation engine. Both bill per endpoint, usually with per-module SKUs on top.
SIEM: Security Information and Event Management: central collection, correlation, and retention of security events. Almost always billed on ingest volume (GB per day) or events per second, which means cost tracks how much you log rather than how large your organisation is. The migration is never the log data, it is the accumulated detection content.
SOAR: Security Orchestration, Automation and Response: the layer that turns an alert into enrichment, containment, and a case. Billed per analyst seat, per playbook run, or bundled into a SIEM. Playbooks are written in a vendor-specific automation model, so leaving means rewriting logic rather than exporting it.
MDR / MSSP / SOC-as-a-Service: Managed Detection and Response, a Managed Security Service Provider, and SOC-as-a-Service all mean renting a security operations team and the platform it runs on. Leaving is an insourcing decision: the tooling is the cheap half and 24x7 analyst coverage is the expensive half.
NDR: Network Detection and Response: detection from traffic rather than from agents, which is how you cover appliances, OT, and anything that will never run an endpoint agent. Billed per sensor, per Gbps inspected, or per monitored site. Unusually, most commercial products are built on the same open engines you could run yourself.
TIP: Threat Intelligence Platform: where indicators, actors, campaigns, and the relationships between them are stored and enriched. The one security segment with genuine portability, because STIX 2.1 and TAXII are real standards that both commercial and open platforms implement.
Detection content: The rules, signatures, queries, and tuning that make a security platform actually detect anything. It is the real asset in any SIEM or EDR, it rarely exports in a form another product can run, and rebuilding it is what a security migration mostly consists of.
Dual-run (security): Running the incumbent and the replacement side by side so detection coverage never drops. In security this is not optional caution: it is the control that guarantees no gap, and it usually runs for a quarter rather than a fortnight.
Ringed rollout: Deploying a new security agent in waves (pilot, then broad) and removing the incumbent only from rings whose detections have been validated. Never estate-wide at once, because two security agents on one host can conflict and one bad ring should not be a fleet-wide outage.

Getting a real price

Where our estimates stop and the commercial process starts.

RFQ (Request for Quote): A structured summary of your requirements, products, sizing, support needs, that you send to an authorized distributor or partner to obtain binding pricing. OffVendor's quote wizard generates one for you.
Reseller / Distributor / Partner: The authorized channel that actually sells you a product and provides a binding quote. OffVendor helps you assemble the requirements to take to one; it does not sell licenses itself.

Missing a term you'd like defined? Email hello@offvendor.com.