Statement:
You are welcome to forward this blog, but please keep the original author information!
Blog: http://blog.csdn.net/halcyonbaby
It is a pleasure to study, study, and summarize the content!
Disaster Tolerance in the cloud environment
What is disaster tolerance?
Simply put, it is a strategy to respond to disasters. Such as fire, theft, man-made damage, volcano, earthquake, flood, war, hurricane, and other natural or man-made disasters.
RTO/RPO
RPO (recovery point objective): the time point that may be restored after a disaster. The amount of data involved in the loss of business data.
RTO (recovery point time): the time required for business recovery after a disaster occurs.
Disaster Tolerance Classification
By RTO: cold, warm, standby
By RPO: synchronous, asynchronous, and offline
Based on business data synchronization technology: Host-Based replication, array-based replication, storage network-based, virtual machine-based proxy, and application-based capabilities (such as database replication capabilities)
Differences between HA and Disaster Tolerance
Ha mainly handles Single Component faults, while dr is responsible for large-scale faults.
There are also some differences between the two from the network perspective. The LAN scale is considered as the HA category, and the WAN scale task is the Dr scope.
From the cloud perspective, HA is a mechanism in the cloud environment to ensure business continuity. Dr is a mechanism for ensuring business continuity among multiple cloud environments.
AWS Disaster Tolerance Solution
AWS's solutions refer to the following types in user scenarios:
- Cold
The cost is the lowest among the three solutions, and the RTO is the longest (> 1 day.
Use S3 for data backup. in case of a disaster, apply for a new virtual machine and use the backup data to recover the data.
Data backup can use common HTTP, VPN, AWS DirectConnect and other links, Snapshot/backup technology for business data synchronization.
- Pilot Light
RTO time (<4hrs) is a relatively economical Disaster Tolerance solution.
Use the replicate/mirror method to synchronize business data.
The disaster recovery virtual machine is started after a disaster occurs.
- Standby
RTO time (<1hrs) is the best for a relatively expensive disaster recovery solution.
Use the replicate/mirror method to synchronize business data.
The disaster recovery virtual machine is always running, but does not provide services.
This solution is divided into two categories. One is that the disaster recovery virtual machine is the same as the production virtual machine, and the service capacity provided after the switchover is the same. The other is that the disaster recovery end maintains a small capacity. After the switchover, the disaster recovery end can provide the business capability but the business capacity is small and needs to be expanded.
Openstack Disaster Tolerance
- Overall Architecture
The overall Dr architecture of openstack is as follows.
There is no plan for whether a new project will be created. At present, we are mainly concerned with the Nova/cinder/completion function, and the orchestration is mainly implemented through heat.
It may become an independent project or even an openstack project in the future.
- Function
Fail over (switching slave nodes after disaster recovery)
Fail back)
Test)
- Solution Introduction
No detailed solution is available. There is only one hight level design.
It is still in the gap identification and completion phase.
- Status quo
Currently, it is mainly concentrated in the case analysis and overall framework design stages.
The specific implementation focuses on the cinder-side metadata and business data synchronization. But the progress is not optimistic.
Reference
- Https://wiki.openstack.org/wiki/DisasterRecovery
- Https://wiki.openstack.org/w/images/4/49/Openstack_disaster_recovery_-_openstack_meetup.pdf
- Http://redhatstackblog.redhat.com/2013/11/26/disaster-recovery-enablement-in-openstack/
- Http://blog.celingest.com/en/2013/03/05/disaster-recovery-in-aws/
- Http://blog.celingest.com/en/2013/03/19/disaster-recovery-aws-high-availability-architectures/
