Download the authoritative guide: Cloud Computing 2018: Using the Cloud to Transform Your Business
Containers are virtualized but not by hypervisors. They can be deployed to a VM but are not VMs.
Both containers and VMs use server/host OS as the bottom two layers of the stack. In VM environments, the next level is the hypervisor followed by VMs containing guest OS, libraries (div/lib in Linux), and applications. A single VM runs two full operating systems: the host and guest OS.
In contrast, containers do not have a hypervisor layer. A container shares the host OS, housing only the libraries and application code and data. Container benefits include greater portability, less operational overhead, lower OS licensing and maintenance/support costs, and less expensive application development.
Orchestration tools and application support services expand container functionality.
Orchestration platforms manage multiple containers with scheduling, cluster management, and resource provisioning. The most widely used orchestration platform is Google-developed Kubernetes, open source since 2014. The toolset enables large numbers of containers to co-exist peacefully without a large operational investment. The orchestration platform allows container owners to:
· Run containers across different computing devices.
· Easily create, move, deploy and retire containers.
· Keep data storage consistent across multiple application instances.
· Balance workloads between containers.
Google Cloud closely integrates Kubernetes with Google Container Engine for cloud-based containers.
Microsoft centralizes its container technologies on Azure with Azure Kubernetes Service, which replaced Azure Container Service. AWS supports containers with Amazon Elastic Container Service (ECS). Docker makes its own orchestration platform, Docker Swarm. Apache Mesos is a large-scale container cluster technology popular with container uses like Twitter and Apple Siri.
Application support tools improve container application performance. For example, many of these toolsets improve container application networking. Docker Networking virtually connects containers, while open source Project Calico uses IP routing to connect containers with VMs and distributes network security policies.
The $164,000 Question: Do You Have to Backup Containers?
Not every container needs to be backed up. Container data consists of two types of data: the application code/image such as a database, and the application data (if any).
There is no need to back up a container image; it can simply be restarted. When a container launches, the runtime engine checks for a local instance first. If it doesn’t find it – let’s say the container was accidentally deleted off the local machine – it simply recreates it. Moreover, if the container does not create application data, then there is no need to backup. These types of containers are ephemeral/temporary, create no data, and have identical configurations.
But if a container creates persistent application data, then IT needs to back up that data to protect it against data loss, errors, and corruption.
Backing up container data to offsite cloud storage, as opposed to transient hosts, makes for the best data persistence in containers.
· Reliability. Backup software automatically validates container data backup in the background, which ensures full recoverability.
· Manageability. A central management console allows IT to treat container backup as part of a holistic backup strategy.
· Flexibility. Look for container backup that can use policies and automated operations. This gives you the flexibility to assign different backup schedules and storage targets to container data.
· Fair pricing. Understand the backup application’s pricing structure including purchase, licensing, maintenance and support, and minimum hardware requirements. If you go with a cloud subscription model, understand how the provider calculates monthly fees.
Types of Container Backup
There are several ways of doing container data protection including mounting, plug-ins, traditional backup applications, volumes, and scripting. They are not necessarily all good ways; but ways they are. Here are a few of the more common methods:
* Follow the 3-2-1 rule. This older rule still applies in modern backup strategies.
* Maintain at least 3 copies of back up data – active data and 2 backups.
* Store your backups on at least 2 different media types. It’s unlikely that nearline disk and off-site tape or cloud will go down at the same time.
* Keep 1 of the backups in a different location. If all your backups are stored in the data center, then good luck to you should the data center go down.
* For containers running database applications, store the database outside of the container on the same server/host. Direct the application’s writes and reads to the database and use the database backup software you normally do.
* Store your database in a Docker Volume, which enables snapshots and replication. Be wary though: this method results in a crash-consistent backup and not an application-consistent one. For that, you would have to quiesce the database to take the snapshot.
* Deploy traditional backup software into the container and back up application data to persistent storage. This method has the advantage of using the same backup software across the enterprise but can consume a lot of resources and compromise container efficiency.
Container Backup Vendors
Asigra, which developed first commercial Docker backup, is an exception to the heavy resource usage of other traditional backup applications. Asigra has supported Docker container backup since 2014. As of version v14, Asigra added its agentless DS-Client that runs directly from a container and is available for download from Docker Hub.
DS-Client is an orchestration application. Deploying its container to an unprotected host that is running Docker Engine, and Asigra Cloud Backup is immediately available. The software backs up data changes-only in a compressed and deduped format, encrypts the data and transports it to the storage platform. Admins can set policies and schedules for container backup including continuous data protection for high-value data. Backup runs transparently in the background with minimal resource consumption—maximizing cloud consumption and making for a cost-efficient backup solution.
The software is tuned to backup Docker container application data to holistic "as-a-Service" cloud providers like KeepItSafe or to any public megacloud vendor.
The KeepItSafe custom-tuned cloud specializes in protecting über-regulated, long-term retention, and high-value databases. This type of comprehensive data availability solution improves data protection across enterprise data with simplified management, automated backup, verified backup, and rapid restore.
Other vendors also offer container backup. Blockbridge plug-in for Docker Volume backs up and restores container data. Users need to install Blockbridge Elastic Programmable Storage as the storage backend. Commvault backs up container images and data by adding a virtualization client to Docker hosts.
NetApp makes a plugin that works on NetApp hardware and software to create container snapshots and clones. HPE Nimble Storage uses a container plug-in on its flash arrays, while Pure Storage offers Docker and Apache Mesos container backup on its storage systems.
Backing up containers doesn’t end with the backup vendor. You also need to trust the backup target including availability, durability, verification, cost-effectiveness, and a secure cloud environment for browser-based restore from anywhere.
Forward-looking organizations are adopting containers for efficient application development and deployment. These businesses understand that backup is critical to protecting enterprise data no matter where it resides.