Traditional backup has a structural limit that scheduling more jobs will not solve: you always lose the data generated between the last copy and the moment of the disaster. If your most recent copy is three hours old and storage fails at 2:55 pm, you have lost almost three hours of transactions. For many workloads that is acceptable; for a financial database or an ERP in full operation, it is not.
You can shrink that window by running copies every hour, every 15 minutes or every 5, but each snapshot has an I/O cost on the production infrastructure, and there comes a point where you cannot go any lower without degrading performance. Veeam Continuous Data Protection (CDP) tackles the problem from a different angle: instead of taking periodic snapshots, it intercepts every write and replicates it almost continuously to a secondary site.
In this article we explain what CDP is and how it works on top of VMware's I/O filter, how it differs from snapshot-based replication and periodic backup, the key concepts you need to know (short-term and long-term restore points, failover, failback), which workloads justify continuous replication and how CDP fits inside a layered disaster recovery strategy.
What CDP Is and How It Works in Veeam
Continuous Data Protection is a data-protection technology that records every block-level change the moment it happens, instead of capturing the full state of the machine at fixed intervals. Veeam CDP relies on the VMware I/O filter (VMware vSphere APIs for I/O Filtering, VAIO), a component that sits in the data path of the virtual machine, between the guest operating system and the storage.
Thanks to VAIO, every time the VM writes a block, that change is duplicated in real time and sent to the replica server at the secondary site. There are no VMware snapshots to create and consolidate, no VM stun and no backup window: the stream of writes is replicated continuously. The result is an RPO (Recovery Point Objective) measured in seconds rather than in minutes or hours.
CDP does not remove the need for consistent points: on top of the continuous stream of changes, it generates restore points at regular intervals that let you roll back to a known state. This combination of a continuous stream and periodic points is what makes it possible to recover the workload both to the second before the incident and to an application-consistent state several hours earlier.
CDP versus Snapshot Replication and Periodic Backup
It is easy to confuse CDP with classic Veeam replication, but they are different mechanisms. Periodic backup creates full or incremental copies in a repository at scheduled intervals (every few hours or once a day); its goal is retention and granular recovery, not immediate availability. Its RPO is measured in hours.
Snapshot-based replication keeps a bootable copy of the VM at the secondary site and updates it by taking a VMware snapshot every so often (for example, every 15 minutes or every hour) to send only the blocks that changed. It is far better than backup for failover, but its RPO is limited by the snapshot frequency and by the impact of creating and consolidating those snapshots in production.
CDP removes the snapshot from the equation. By intercepting every write through VAIO, it does not depend on a capture frequency: the change travels to the destination almost the instant it happens. That is why its RPO drops to seconds and its production impact is lower than taking very frequent snapshots. The trade-off is that it demands more sustained bandwidth and more resources at the destination, because the data flow is constant.
Key Concepts: Restore Points, Failover and Failback
To design a CDP policy properly you need to be clear on several concepts that define how and how far back you can recover:
- check_circle Short-term restore points: those generated almost continuously by the CDP stream, with second-level granularity. They let you roll back to the instant immediately before the incident, but they are retained for a short period (usually hours).
- check_circle Long-term consistent points: periodic points, with application-aware processing, that guarantee the consistency of applications such as databases or mail servers. They are retained for days or weeks and are what you fall back on when the most recent state is corrupt.
- check_circle Failover: the process of switching to the replica at the secondary site when the production workload becomes unavailable. With CDP you can choose the exact restore point you want to boot from.
- check_circle Failback: the reverse operation. Once the primary site is restored, you return the workload to its original location, replicating back only the changes that occurred during the failover to minimise the sync time.
- check_circle Secondary site: the location (another datacenter or a DRaaS provider) where the ready-to-boot replica lives. It must have enough compute and storage to take over the critical workloads during the contingency.
RPO, RTO and Which Workloads Justify CDP
Before deciding where to apply CDP it helps to recall the two metrics that govern any recovery plan. The RPO (Recovery Point Objective) is the maximum amount of data you can afford to lose, measured in time: an RPO of 5 minutes means you will lose, at most, the last 5 minutes of data. The RTO (Recovery Time Objective) is the maximum time you can take to be operational again. If you want to dig deeper into both, we cover them in our guide on disaster recovery, RPO and RTO.
Rule of thumb:
CDP attacks the RPO directly, driving it down to seconds. Combined with a replica ready for failover at the secondary site, it also cuts the RTO to minutes. Not every workload needs that level: reserve CDP for the ones where losing minutes of data has a real business cost.
The workloads that do justify CDP are those combining a high change rate with intolerance to data loss: transactional databases (OLTP), ERP systems, e-commerce platforms with orders arriving continuously and financial systems where every unrecorded transaction is lost money or a compliance problem. In these cases, the extra network and storage cost is justified by the value of the protected data.
The workloads that do not justify CDP are those with a low change rate or low criticality: file servers with sporadic access, development and test environments, static web servers or archival workloads. For these, a periodic backup or a snapshot-based replica provides a more than sufficient RPO without the overhead of continuous replication.
Requirements to Deploy CDP
CDP is not a box you simply tick: it requires the right infrastructure at the source, across the network and at the destination. The three fundamental requirements are:
- check_circle VMware with VAIO: Veeam CDP depends on VMware's I/O filter, so it requires compatible VMware vSphere environments. This capability is not available in the same way across all hypervisors, because it is based on that specific storage-level integration.
- check_circle Sufficient, stable bandwidth: because the stream of writes is replicated continuously, the link to the secondary site must sustain the peak write throughput of the protected workloads. Insufficient bandwidth causes the real RPO to degrade below what you configured.
- check_circle A properly sized destination site: the replica needs compute and storage at the secondary site, plus extra space to retain the short-term and long-term restore points. A DRaaS provider solves this requirement without you having to build and maintain a second datacenter of your own.
CDP in a Layered DR Strategy
CDP does not replace backup: it complements it. A mature disaster recovery strategy is layered, applying to each workload the level of protection its criticality justifies. You reserve continuous replication for what cannot afford data loss, and use periodic backup for everything else and for long-term retention, thereby honouring the 3-2-1 rule.
Continuous replication
Transactional databases, ERP, e-commerce, financial systems. Seconds-level RPO, failover in minutes.
Snapshot replication
Application servers and business services with an RPO of minutes to an hour and fast failover.
Periodic offsite backup
Files, dev/test, archival and long-term retention. The foundation of the 3-2-1 rule.
At EasyDataHost we offer DRaaS with Veeam and a secondary site in our datacenter in Spain, so you can deploy CDP for your critical workloads without investing in a second physical location. For the standard layers, we complement the strategy with offsite backup with Veeam Cloud Connect. If you have not set up replication yet, our guide Veeam Cloud Connect step by step explains how to get started.
Comparison Table: Backup vs Snapshot Replication vs CDP
The following table sums up the key differences between the three mechanisms across the parameters that matter most when designing a continuity plan:
| Criterion | Periodic backup | Snapshot replication | CDP |
|---|---|---|---|
| Typical RPO | Hours | Minutes to 1 hour | Seconds |
| Typical RTO | Hours | Minutes | Minutes |
| Mechanism | Scheduled copies | Periodic snapshots | Every write captured (VAIO) |
| Production impact | Low (window) | Medium (snapshots) | Low and continuous |
| Network usage | Low | Medium | High (continuous) |
| Relative cost | Low | Medium | High |
| Long-term retention | Yes (days/years) | Limited | Short + long-term points |
| Ideal use | Retention / 3-2-1 | Important services | Critical workloads |
Frequently Asked Questions
What RPO can you achieve with Veeam CDP?
A seconds-level RPO. CDP captures every write through VMware's I/O filter (VAIO) and replicates it almost continuously to the secondary site. Compared with a periodic backup with an RPO of hours, the window of data at risk shrinks to a configurable threshold that can drop to just a few seconds, depending on available bandwidth.
Does CDP replace traditional backup?
No. CDP and backup are complementary. CDP protects critical workloads with a very low RPO, but its network and storage costs are higher and its retention is usually shorter. The recommended approach is layered: CDP for what is critical, and periodic backup for everything else, for long-term retention and to satisfy the 3-2-1 rule.
Which hypervisors support Veeam CDP?
Veeam CDP relies on VMware's I/O filter (VAIO), so it requires compatible VMware vSphere environments. It is not available in the same way across all hypervisors, because it depends on that specific VMware storage-level integration.
Conclusion
Periodic backup protects your information, but it always leaves a window of data at risk between copies. For workloads that cannot afford that loss, CDP changes the game by replicating every write almost continuously:
- arrow_right CDP achieves a seconds-level RPO by capturing every write through VMware's I/O filter (VAIO), without relying on periodic snapshots.
- arrow_right It combines short-term restore points (seconds) with long-term application-aware consistent points, and supports swift failover and failback to and from the secondary site.
- arrow_right It is justified for transactional databases, ERP, e-commerce and financial systems; for low-criticality workloads, the network and storage cost does not pay off.
- arrow_right It is a layer within a DR strategy: CDP for what is critical and periodic backup for everything else and long-term retention. EasyDataHost offers DRaaS with Veeam and a secondary site in its datacenter in Spain.