Hyper Automation, Artificial intelligence, Cognitive IT are some of the hot topics in the industry today. In a recent discussion with an industry veteran, I was directed to the HBR article by Julia Kirby - titled - "An Inside Look at Facebook's Approach to Automation and Human Work". The article, which primarily touches upon the need to remove the human element, not just for the cost benefit but also to eliminate other uncertainties around humans, also speaks about the challenges - which indeed resonates well with enterprise data center operators. So can other enterprises learn best practices from different organizations and replicate the same kind of automation in-house? Yes, but with a tweak.
From an enterprise standpoint, it is important to identify specific scenarios and see how we can leverage and apply the learnings from different organizations, such as Facebook. For instance, the 25000:1 server operator ratio, as mentioned in the article, is highly ambitious from an Enterprise Data Center perspective to the extent that there are hardly any enterprises who even have that kind of physical servers globally. Therefore, it is important to chart down some of the key differences in an Enterprise Data Center set-up vs Facebook's Approach, or for that matter Google, AWS, Azure etc. and determine how differently and efficiently this can be done.
Let's look at the standard enterprise infrastructure set-up and see what stacks it up:
- Large variation of hardware models (Rack, Blade, Cartridge), type (RISC Unix, CISC) of servers due to purchase history
- Similar to servers (SAN, NAS, JBODS), there are a variety of storages catering to specific performance, availability, cost and scalability needs
- Unique Network configurations, SAN, LAN, WAN, Internet, Clustering, Management, DR replication etc.
- Existence of multiple operating systems (OS) due to northbound dependencies from application and platform and southbound due to hardware variation
- Backup and archival set-up due to the nature of data and business compliance requirements
- Many custom appliances to deliver certain functionality
- Tailored security software/hardware/process to address specific threats and business requirements
- Broad range of BSPOKE and COTS application having unique configuration and implementation, ISV dictated prerequisites
- Application operation requirements and limitations
- Enterprise specific integration architectures around Technology, Information and Application
- Range of tools due to historical reasons and best fit adoption
- Cabling and rack layout based on the variety of the commissioned infrastructure
- A variety of inflight projects that demands operations and change alignments
A good analogy of managing the different data center set-ups is similar to managing 1000 sheep's for a year vs managing 1000 humans. A brief chat with a kindergarten teacher is all it takes to bring in the appreciation. 1000s of replicas of identical server with few variation is a different scenario when compared to an Enterprise Data Center where every server is virtually unique.
For instance, at Wipro, the data centers we deal with have servers ranging from 10s of 1000s in capacity at the higher end and few hundreds on the lower end. A globally distributed delivery centers manned by experts, a comprehensive tools framework, automated process and demarcated organization structure defines our overall service delivery system for the data center operations. ServiceNXT - our delivery platform forms the corner stone of our operations.
The few things in my opinion that has helped us consistently move the needle of efficiency are:
- Analytics on data: Gain insights from the wealth of data and global patterns using different data science techniques
- A dedicated automation team: Investment in dedicated development team with sharpened infrastructure programming skills
- Pollinate skills: Impart training to administrators, client specific team, get them thinking to contribute
- Tools foundation: A delivery platform and framework with high standardization and extensibility
- Opportunity across the data center stack/layer: Bucketing the opportunities of automation across the stack, create expert evaluation groups, prioritize based on effort vs benefit
- Mimic Human action: Get broad collaboration, crowd source to design BOTS that mimic human response
- Agile development: Adopt an agile SCRUM based approach for the development lifecycle of BOTS, event correlations logics, skill based routing, learning algorithms etc.
- Encourage shared delivery: Educate customers the benefit of shared delivery vs exclusive ring fenced models
- Automation catalogue/marketplace: Advertise automation assets, bring transparency, encourage/incentivize usage
- Accept variety is a reality: Push standardization but keep also keep the automation pipeline across
The results from these actions have been very encouraging where we have been able to drive better satisfaction for our customers. I personally believe the Data Centers of the new-age enterprise need to be born intelligent, acquire intelligence, and cultivate IQ.