Exploring the leading appliance-based data backup systems
Learn how leading integrated backup appliance systems address key considerations, including capacity and performance, and what operational benefits they offer.
Appliance-based integrated data backup systems offer a range of benefits, including the ability to run virtual workloads and long-term data preservation. Selecting the right appliance to meet your needs depends on a number of factors, including capacity and performance requirements and secondary data usage.
Here, we examine integrated backup appliance offerings from eight leading vendors: Arcserve, Barracuda Networks, Cohesity, Commvault, Dell EMC, Rubrik, Unitrends and Veritas Technologies.
The vendors profiled here include those with a long legacy of backup and data protection knowledge and intellectual property mixed with a couple of younger startups that have made an impact.
This division creates a need for a deeper dive into the technology to explore the architectural differences between the old guard and the new.
The benefits of a scale-out vs. scale-up architecture
As the amount of data stored in backup systems increases, backup platforms need a way to grow in capacity and performance. The newcomers on our list -- Cohesity and Rubrik -- use an underlying scale-out file system to store backup content. The technology for both platforms has been developed as part of the core intellectual property. The design goals have been set around handling large numbers of backup images, usually snapshots, with high levels of reliability. Commvault took the same approach, incorporating a scale-out file system into its HyperScale architecture. This implementation is based on Red Hat's Gluster file system. Commvault decided to partner with Red Hat rather than build its own file system.
The vendors profiled here include those with a long legacy of backup and data protection knowledge and intellectual property mixed with a couple of younger startups that have made an impact.
A scale-out architecture provides resiliency for the data stored within the file system. Cohesity integrated backup appliances that start at a minimum of three nodes, although the C2000 model is sold as a four-node chassis. Commvault data backup systems also start at three nodes and increase in sets of three. Rubrik offers two-, three- and four-node configurations.
The remaining data backup appliance vendors examined here generally implement a scale-up architecture. This means capacity increases are delivered by implementing bigger servers that have more CPU, memory and disk drives. The result is a range of offerings that start small and could be used for remote or branch offices, as well as enterprise-class platforms that support hundreds of terabytes. For example, Arcserve offers four configurations of its data protection appliance. Barracuda provides 12 standard appliance models and another four with additional encryption. Unitrends also offers 12 different integrated backup appliances. The exception here is Veritas; its appliances can be deployed specifically as media or master servers to enable scaling of specific functionality. This also enables a design where customers could choose to mix bespoke backup products, such as master servers, with appliances in the same configuration.
Implementing a scale-out architecture provides operational benefits similar to those offered for primary storage by the hyper-converged vendors. Capacity can be added over time in smaller increments into one single backup cluster. Scale-up requires a bit more capacity planning, as the increments are typically bigger and more expensive, although some vendors scale by adding extra disk shelves to their appliances. However, it really comes down to understanding your organization's growth and distribution of backup requirements. For example, if your organization has many sites, it may be more cost-effective to manage several smaller integrated backup appliances.
Editor's note
Using extensive research into the data backup market, TechTarget editors focused on integrated backup appliances that include hardware and software that are provided by the same vendor. Our research included data from TechTarget surveys, as well as reports from other respected research firms, including Gartner.
How leading integrated backup appliances address secondary storage
With disk as the dominant backup medium, vendors have added the ability to use the backup appliance as a storage target or to run virtual machines (VMs) from the backup hardware as a temporary data store. Here, the architecture of the platform is crucial because accessing data within the file system that comes from many backup points requires a file system built for that process. Arcserve Unified Data Protection's Instant Virtual Machine feature, Barracuda's LiveBoot, Cohesity, Rubrik and Unitrends all offer the ability to boot a VM directly from a virtual data store on the backup appliance. This restored image can be used for testing or even live-migrated back to primary storage using tools such as VMware vMotion.
Spinning VMs up in this manner is a powerful feature and can be extended to other uses, such as validating the integrity of backups or seeding test and development environments.
Features and support offered by appliance-based data backup systems
Vendors that have been in the market longer will typically have greater support for backing up a range of OSes, hypervisors and applications. Cohesity and Rubrik initially supported the VMware vSphere hypervisor platform, but over time, they added support for other hypervisors and common applications to their products. These vendors targeted the hypervisor because it was easy to use backup APIs provided by the hypervisor vendor to take backups without deploying agent software into each VM guest. Providing support for additional platforms and applications becomes more challenging because some might require agents. And, of course, supporting more platforms means more testing with each software version release.
Legacy backup vendors generally have expertise and support for common applications, such as SAP, Oracle, SQL and Microsoft Exchange. There is also a great heritage of backup innovation in older platforms. Dell EMC, for example, uses Avamar and Networker technology within its range of Integrated Data Protection Appliance products. Data efficiency is delivered using Dell EMC Data Domain technology, widely recognized as a leader in data deduplication. Veritas, which has years of experience in highly scalable backup products, uses NetBackup as the basis of its range of integrated data backup systems.
An increasing trend in the backup market, including appliances, is to natively support primary storage platforms. Integrating directly with storage provides better access to snapshot data, enabling this data to be stored more efficiently. Array-based snapshots have less impact on the application, compared to those from the hypervisor. Array-based support is available with backup appliances from Cohesity, Commvault, Rubrik and Veritas.
Transforming and migrating backups to an appliance model
Unlike primary storage platforms, backup systems typically retain many versions of data over a long period of time. Many organizations even use backup as a form of long-term archive. Because of this, backup systems tend to stay in place for many years, complicating the process of keeping track of when and where a system was backed up. Having the ability to easily transition backups from a legacy platform into an appliance model is one area in which the longer-established vendors have an advantage over the newer vendors. Arcserve, Commvault and Veritas offer seamless interoperability between their traditional and appliance platforms. Their integrated backup appliances essentially run the same software, and moving data between backup environments is an existing feature.
Data backup systems that support cloud
public cloud -- and, in particular, cloud storage -- offers great potential as a long-term storage medium for backup data. Older backups are quickly moved to an inactive and infrequently accessed state. This data can be offloaded to public cloud providers and recalled as needed. Accessing this data over distances is not a problem, as the access profile is sequential and is not particularly affected by latency.
Arcserve, Barracuda, Dell EMC, Cohesity, Rubik, Unitrends and Veritas all enable the cloud to be used as a backup target. Both Barracuda and Unitrends provide their own cloud service. It is important to understand exactly how this is implemented by the provider. Ideally, cloud storage would store only unique, deduplicated data at a granular level. Remember that cloud providers do not offer levels of data deduplication, so all data that is written is charged for, regardless of any redundancy there may be.
Traditional backup cycles, using periodic full backups that are supplemented by incremental or differential backups in between the full backup operations, will not work well with cloud storage unless there is deduplication at the appliance layer. A better option is to use incremental-forever backups. It is important to validate how cloud storage is used to ensure that charges do not mount up unnecessarily.
The next step in cloud usage is to use public cloud as a recovery point for applications. Many vendors offer virtual appliances that run in the public cloud and can access data written by on-premises appliances. This includes Arcserve, Cohesity CloudReplicate, Rubrik and Unitrends. Reinstancing images to the cloud provides disaster recovery functionality without having to invest in a secondary data center. It also enables the cloud to be used for test/dev with backups seeding data.
Integrated appliances that automate the backup process
Finally, we should highlight the use of APIs. Rubrik focuses heavily on the use of PowerShell and REST API features to enable backup, restore and testing processes to be automated. Cohesity offers similar features. Remember this is an API to manage the platform and is separate from the APIs used to back up data in the first instance. APIs are also available for Commvault HyperScale and Unitrends.
