Glossary
Hybrid cloud
Hybrid cloud joins a private environment with one or more public clouds so that data and applications can move between them. Steady, sensitive workloads run on private infrastructure, bursty projects and managed services run in the cloud, and the two sides operate as one.
What has to be shared for two environments to act as one?
A private data center and a public cloud connected by a network are not yet a hybrid cloud. Four layers make them work together: a common identity system, so one account and one policy apply on both sides; networking that links them with predictable bandwidth; monitoring that shows both in one place; and a common storage interface, often S3, so an application does not care which side holds its data. Without the last layer, every workload that crosses the boundary needs custom code, and the arrangement is two separate estates joined by a link. The private side is typically a private cloud.
What decides where a workload runs?
Data gravity decides, more than compute does. Copying a petabyte over a 10 Gb/s link takes more than nine days at full use. A job that has to read that petabyte repeatedly cannot sit on the far side of the link from it. The practical result is that compute follows the data: the analysis moves to where the data lives, and data crosses the boundary only when it has to. The idea has its own entry under data gravity, and the related concept of data locality covers keeping processing near storage.
One authoritative copy of each data set keeps the arrangement coherent. Duplicates in both environments invite arguments about which one is current.
Which patterns are common?
| Pattern | What runs where | What it relies on |
|---|---|---|
| Burst | Primary data on premises; peak compute in the cloud | Fast link, data staged in advance |
| Archive | Active data on premises; aged data moved to a cloud tier | Lifecycle rules, acceptable retrieval delay |
| Cloud disaster recovery | Production on premises; a recoverable copy in the cloud | Replication, tested failover |
What does a hybrid cloud look like when it starts to drift?
Three signs appear gradually. Transfer charges rise month over month, usually from egress fees on data that crosses the boundary on a schedule. Several copies each claim to be the master, so changes are made in one place and missed in another. And policies diverge: retention, encryption and access rules differ between the two sides because they were set at different times by different teams. Regular review of those three catches the drift before it becomes a migration project. Moving data between sides on purpose is the subject of data mobility and, when a workload changes home, cloud migration.
Frequently asked questions
What is the difference between hybrid cloud and multi-cloud?
Hybrid cloud mixes deployment models, such as on-premises with public cloud. Multi-cloud uses more than one provider. An estate can be both.
Why is a common storage interface important?
It lets applications read and write the same way wherever the data sits, so workloads can change location without code changes.
Does hybrid cloud require a dedicated network link?
Not strictly, but data-heavy hybrid designs usually use one, because internet paths vary in bandwidth and latency.
Related terms
- Private cloud: the owned side of a hybrid design.
- Multi-cloud: using more than one provider.
- Data gravity: the rule that places workloads.
- Data mobility: moving data across the boundary.
- Egress fees: the charge on crossing it.