Disaster recovery planning

Requirements analysis

Data protection and business continuity are vital for data centers. The failure of a critical application or data loss can cause significant business losses. Hybrid Disaster Recovery provides two layers of capabilities to protect data and ensure business continuity.

• Geo-redundancy

  Server images and data backups are uploaded directly to the Alibaba Cloud disaster recovery repository. This provides highly reliable geo-redundancy in the cloud. Stable geo-redundancy ensures that critical data is not lost during extreme events, such as a fire in your on-premises data center. After the on-premises facility is repaired, you can recover the data to your on-premises environment.

• Cloud disaster recovery

  To reduce business losses from application failures, Hybrid Disaster Recovery can quickly recover your applications on Elastic Compute Service (ECS) if your data center experiences a critical failure and cannot be recovered quickly.

RTO and RPO requirements

Application disaster recovery has two core metrics:

• RPO: The maximum amount of data loss that an application can tolerate when a failure occurs. The more important the data, the lower the RPO target should be. A lower RPO often requires more frequent data backups and replication. This increases the load on the production environment and network, and usually results in higher costs.

• RTO: The expected time from the start of a disaster recovery operation to when the application is back online. The greater the business loss per unit of time during a failure, the shorter the RTO target should be.

Business departments usually propose RTO and RPO targets in consultation with the IT department. The final targets are determined based on technical feasibility, impact on existing systems, and cost. The standards for RTO and RPO often have a linear relationship with infrastructure costs.

Hybrid Disaster Recovery provides simple configurations to meet different RPO and RTO requirements. For example, Continuous Data Replication (CDR) can provide RPO and RTO targets ranging from seconds to minutes.

Application analysis

Before you deploy disaster recovery, you must understand the deployment of your critical applications, their environmental dependencies, and how clients connect to them.

• Application deployment

  Before you deploy critical applications, consider the following three factors:

  • The servers that the application includes

  • The network connectivity between the servers

  • The configurations required within the servers

  For example, a simple web application includes the following components:

  • The application includes one database server, one backend server, and one web frontend server.

  • The three servers are on the same network.

  • The backend server has a configuration item that specifies the IP address of the database server. The web frontend server has a configuration item that specifies the IP address of the backend server.

  After you identify these factors, you can make the following plan:

  • Hybrid Disaster Recovery must protect these three servers.

  • When recovering to Alibaba Cloud, restore these three servers within the same VPC.

  • After a full server recovery, you can use the same IP addresses as the on-premises environment to ensure the application can run. Alternatively, you can use an automated script to modify the configuration items after the recovery is complete.

• Environmental dependencies

  Application disaster recovery is a process that requires cooperation across multiple departments. It involves roles such as application administrators, data center administrators, and network administrators. A complete disaster recovery plan that meets business requirements must consider details from multiple perspectives, including the following:

  • The environments that the application depends on, such as Active Directory (AD) and DNS.

  • The network configurations required by the application.

  In many cases, applications have other important environmental dependencies. For example, in a Windows environment, many applications depend on AD to run. When you recover the application to the cloud, your VPC environment must be able to connect to the AD service. DNS services are also a critical requirement in many environments.

  Take AD as an example. There are usually two scenarios:

  • If you have already deployed multiple primary and secondary AD servers in different data centers, you only need to establish an Express Connect or SSL-VPN connection between the data center where the AD servers are located and your cloud VPC.

  • If your AD servers are centrally deployed in a single data center and might go offline simultaneously, do the following:

    • Deploy a secondary AD server in your cloud VPC and keep it connected to the primary on-premises AD server. If a failure occurs on-premises, you can use the AD server in the cloud.

  Similarly, DNS servers must also be configured accordingly to meet the application environment requirements after disaster recovery.

• Client application connections

  After the application is recovered, you must ensure that clients can connect to it. The typical steps are as follows:

  • If the recovered application server has the same IP address as the original and the DNS server is also successfully recovered, clients only need network connectivity to the application. You might need to use an SSL-VPN or Express Connect connection to ensure clients can connect to the recovered application in the cloud. Alternatively, you can assign a public IP address to the recovered application for client access.

  • If the application recovery does not require the original IP address, you can also modify the DNS records to ensure clients can connect to the new service.

  • If both the domain name and the IP address change, you must modify the client configuration.

Network environment

Based on the number of application servers, data volume, RTO and RPO standards, and environmental dependencies, you can select and deploy a suitable network environment. The following are network environment configuration options:

Network between your on-premises data center and Alibaba Cloud

• Hybrid Disaster Recovery uses an optimized data storage and transmission algorithm, so it does not require a leased line connection between your on-premises data center and Alibaba Cloud. However, for scenarios with large data volumes and strict RPO requirements, you should use a leased line connection. This ensures that the disaster recovery service can meet the required metrics.

• After application recovery, based on the connection requirements for clients, AD, DNS, and your Alibaba Cloud VPC, you might need to consider using an SSL-VPN, an Express Connect connection, or assigning a public IP address to the application to ensure normal operation.