by Jelani Harper 16 March 2020
There are few facets of modern business that haven’t been impacted by some form of artificial intelligence. One of the least discussed (yet perhaps most pivotal) applications is for ensuring business continuity in the event of failure, disaster, or data loss.
The capability to issue timely, periodic backups is critical for organizations looking to remain operational in case of any of the aforementioned events—which might also include various infiltration or exfiltration attempts on behalf of cyber attackers.
Consequently, the capacity to expand the variety of backups and their flexibility is of utmost importance to data-driven organizations. Backups are oftentimes the first step in business continuity implementations, which should ideally involve options for maintaining network availability. According to OJ Ngo, CTO at DH2i, “smart high availability technology complements and resolves the issue of making sure the business is always available.”
By leveraging the rules-based side of AI to facilitate intelligent high availability, organizations can reduce the amount of data loss experienced. By infusing the backup process with intelligent algorithms, “part of which involves light-level machine learning,” Retrospect general manager JG Heithcock said, organizations can accelerate disaster recovery in case of any type of data loss.
The combination leverages total AI (its statistical base and knowledge base) to propagate business continuity.
Job-based and policy-based backups
Traditionally, backing up servers, VMs or individual applications was done on a job-based model. “You start at eight in the morning every day of the week, backup this list of sources in this order, and go to this destination,” Heithcock explained. The job-based model is still credible for situations in which sources are regularly present at the defined time for the specified backups. However, the increasing decentralization of the data landscape, in addition to the strong influx of mobile technologies, has rendered this model much less effective for the array of portable devices currently in use.
According to Heithcock, it’s much more practical to perform backups on mobile devices such as laptops by applying AI because “that’s where the job-based thing falls down.” In a jobs-based model, if a mobile device or storage unit isn’t present at a particular time, that device won’t get backed up until the next scheduled job (which could be weeks away). The intelligent algorithms of statistical AI enable users to devise better polices for backups while the system implements them with a degree of license that’s ideal for contemporary computing options.
The rules comprising organizational policies for network availability typify the knowledge base side of AI. Those rules provide the foundation for the automation that takes place for failovers, load balancing, or any other applications where there’s a need to dynamically transfer workloads between environments. According to Ngo, organizations must “define the rules as the administrator or the god of the system, of how the business should be ran or the automation will be done, and the technology carries them out.”
With this approach, organizations can automatically failover from on-premises settings to the cloud, from the cloud to on-premises, or even between clouds, to ensure their networks remain available. Moreover, they can leverage this same approach to simply shift workloads where they’re most computationally viable—which is valuable for applications of machine learning or deep learning. The key to doing so is with dedicated rules so “if a unit of work fails, the technology kicks in automatically without human intervention and starts up another device, and/or transfers or offloads the workload onto another device to make sure business functions are correctly delivered,” Ngo said.
For situations in which business continuity requires backups, policy based-backups incorporate what Heithcock characterized as “a strong focus on algorithms” to facilitate a number of advantages with AI. Instead of rigidly scheduling jobs in a predefined sequence at a set time, this type of back-end intelligence enables organizations to specify the requirements for their backups—which machines to backup, how frequently, and where—then lets the system handle them. Specifically, a combination of machine learning and static algorithms is responsible for:
- Non-sequential backups: Intelligent backup solutions deliver their functionality in different sequences based on the availability of devices, which is bound to fluctuate for laptops, tablets and smartphones. If a certain employee’s laptop isn’t available for a daily backup at eight in the morning, the system will backup another while “still looking to see if [the first machine] has come online,” Heitchcock said.
- Preferential backup hierarchies: Part of the decision-making capabilities of the AI powering these systems is applied to determining which backups exceed others in terms of importance. For instance, if a policy calls for daily backups and one employee has been offline three days, while another has only been offline a day and a half, the system will backup the former as a priority. “That’s the proactive part, or the AI part,” Heithcock noted. “It’s going to jigger its priority list to try and get to people who are most out of policy first.”
- Temporal prioritizations: Another crucial aspect of machine intelligence used by smart backups is the capability to prioritize jobs based on the length of time they take. If two jobs should be done at the same time (meaning there’s little difference in the time since the last backup) but “Able is going to be somebody you can backup in 10 minutes and Fred takes you two hours, then it backs up the [former] first to get done with him and have more time to do the other,” Heithcock said.
Such applications of AI highlight the ability to dynamically learn over time and improve processes which might not necessarily be consistent—particularly when accounting for the large degree of variation mobile computing involves. The underlying algorithms are able to adapt how they process their jobs based on the needs of the end-user. In these use cases, “the machine learning is weighing when these guys tend to come online, how much data it’s going to take, can I get this guy in real quick and move onto people who might take more time,” Heithcock summarized.
When the business continuity capabilities of backups are paired with the portability of intelligent high availability technologies, organizations are able to improve their capacity to keep networks available and data protected. Maintaining business continuity requires no less, and this application simply illustrates another dimension of the criticality of AI to contemporary business processes.
Jelani Harper is an editorial consultant servicing the information technology market, specializing in data-driven applications focused on semantic technologies, data governance and analytics.