BCP AND DRP
Business continuity and disaster recovery: BCP and DRP
Choose, VM by VM, the level of protection for your applications: automatic failover with no data loss if a server goes down, and recovery on a remote site in the event of a disaster.
BCP or DRP: what are we talking about?
The business continuity plan (BCP) aims to keep the service running despite a failure: a server, a disk, a power supply. The disaster recovery plan (DRP) organizes the restart after a disaster that affects an entire site: fire, flood, extended outage.
Two indicators describe them: the RPO, the amount of data you are willing to lose, and the RTO, the time needed to restore the service.
BCP: synchronous replication, with no data loss
In synchronous mode, every write from the VM is copied to a second server on the same site before being acknowledged to the application. Losing the primary server loses no acknowledged write: the RPO is zero.
In the event of a failure, the platform isolates the faulty server (fencing), then restarts the VM on the server that holds its replica. You choose whether this failover is automatic, or proposed to your team for approval.
DRP: asynchronous replication to a remote site
In asynchronous mode, writes are sent continuously, through an encrypted tunnel, to a distant recovery site. The VM does not slow down; the replica lags a few seconds to a few minutes behind, depending on the objective set for the VM.
Failover to the recovery site is decided by a human: it is a business decision, taken with full knowledge of the facts. The two modes can be combined: synchronous on site, asynchronous at a distance.
Never a “split brain”
A VM must never run in two places at once. The platform only fails a VM over automatically to an up-to-date replica; if the replica is lagging, the decision is left to a human, who then knows what they risk losing.
The VM comes before its replication
A network outage or an unreachable replica must not stop your application. The VM carries on alone, the replica is marked as lagging and an alert is sent. When the link comes back, resynchronization happens automatically, with the VM running, without a full copy.
Building your DRP
A DRP is more than a replica: it requires a recovery site, objectives (RPO, RTO), a failover procedure and regular drills. Our engineers build it with you, application by application.
Our commitments: RPO, RTO and availability
Our commitments depend on the protection chosen for each VM; the table below summarizes them. With a synchronous replica and automatic failover, we commit to a monthly availability of 99.9%, i.e. at most 43 minutes of downtime per month, with no data loss (zero RPO) and recovery in under 3 minutes.
During an outage of the replication link, the VM keeps running but the RPO is no longer met: it grows with the duration of the outage, and an alert is sent after 5 minutes. Excluded are announced maintenance, the simultaneous loss of a server and its replica, the loss of a site for a VM without a replica on another site, and causes internal to the VM. The detailed terms are set out in the contract.
Our commitments by protection level
| VM protection | RPO (data loss) | RTO (recovery) | Monthly availability |
|---|---|---|---|
| Synchronous replica, automatic failover (BCP) | Zero | Under 3 min | 99.9% (at most 43 min of downtime per month) |
| Synchronous replica, failover subject to your approval | Zero | Under 3 min after your approval | 99.5% (at most 3 h 39 min of downtime per month) |
| Asynchronous replica on a recovery site (DRP) | A few seconds: objective set per VM, alert beyond 5 min | Under 1 min after the failover decision | – |
| Continuous data protection (CDP) | A few seconds, within the chosen window (1 h to 14 days) | Depending on the volume restored | – |
| Scheduled backups | The interval between two backups | Depending on the volume restored | – |
| No replica | – | Server repair | No numeric commitment |
BCP AND DRP
At a glance
🔁
Synchronous replication
Zero RPO: no acknowledged write is lost.
🌍
Asynchronous replication
To a remote site, without slowing down the VM.
⚡
Automatic failover
Or proposed to your team, as you choose, VM by VM.
🛡️
No split brain
Automatic failover only to an up-to-date replica.
📴
Resilient to outages
The VM keeps running, and replication catches up on its own.
🎚️
Your choice, per VM
Synchronous, asynchronous, both, or no replica.
PROCESS
How it works
Step 1 · Classify your applications
For each one, set the acceptable data loss (RPO) and downtime (RTO).
Step 2 · Choose the protection
Synchronous for the BCP, asynchronous for the DRP, both for critical applications.
Step 3 · Practice the failover
Test recovery regularly: a DRP that has never been exercised remains a hypothesis.
FREQUENTLY ASKED QUESTIONS
Good to know
What is the difference between a BCP and a DRP?
The BCP (business continuity plan) keeps the service running during a failure, for example a server failure, thanks to synchronous replication and automatic failover. The DRP (disaster recovery plan) organizes the restart on another site after a disaster, based on asynchronous replication.
What do RPO and RTO mean?
The RPO (Recovery Point Objective) is the maximum amount of data you are willing to lose, expressed as time. The RTO (Recovery Time Objective) is the maximum acceptable time to restore the service.
Does my VM stop if replication is interrupted?
No. The VM keeps running, the replica is marked as lagging and an alert is sent. Replication catches up automatically when the link comes back.
Do I need to replicate all my VMs?
No. VMs with no data to protect, such as the identical servers of an autoscaling group or test environments, can run without a replica, which saves space and network bandwidth.
What SLA do you offer?
With a synchronous replica and automatic failover: 99.9% monthly availability (at most 43 minutes of downtime per month), a zero RPO and an RTO of under 3 minutes. With failover subject to your approval: 99.5% (3 h 39 min). A VM without a replica is brought back into service after its server has been repaired, with no numeric commitment.
What RPO and RTO for a DRP?
Asynchronous replication to the recovery site follows an RPO objective set for each VM, typically a few seconds, with an alert beyond 5 minutes. Once failover has been decided, the VM resumes on the recovery site in under one minute.
GRIDALE CLOUD
The other services in your VDC
How the Gridale cloud works
The Gridale cloud combines proven building blocks (OpenBSD, KVM, etcd, pf, relayd) around simple, auditable code. …
Your virtual data center (VDC) in France
A VDC brings your entire infrastructure together in an isolated space: servers, networks, firewall, VPN, storage …
Virtual servers hosted in France
Choose a predefined configuration or build your VM by the vCPU, by the GB of RAM and by the GB of disk. Linux and …
Encrypted VMs, from disk to memory
Encrypt each VM’s disk with AES-256 using a key unique to that VM, and its memory with the AMD processor. A …
Appstore: your applications ready in minutes
Choose an application and a VM size: the platform takes care of the rest, from installation to HTTPS, including …
Autoscaling: your servers follow the load
An instance group adds VMs when the load goes up and removes them when it goes back down. You pay for the capacity …
Routed SDN network: private networks and public IPs
Each organization has its own network, isolated from the others at the routing level. Your VMs communicate with …
Firewall as a Service (FWaaS) based on pf
Every VM has its own firewall, applied on its interface by pf, OpenBSD’s firewall. You get the fine-grained …
Ready to launch your project?
Estimate your infrastructure in two minutes or talk it over with an engineer.
