Since 2006 I have been fortunate to work with leading global and regional banks, financial services, and insurance firms (BFSIs) during IT consulting engagements. I remember working directly in the bank’s data center, mounting the keyboard and monitor to rack-mounted servers to install and configure IT solutions in the initial years. Due to security protocols, not even Secure Shell/Remote Desktop Protocol were allowed to access those mission-critical servers from internal staff networks. Gradually, I witnessed a shift to Virtual Desktop Infrastructure to implement and run IT solutions.
BFSIs Now
In recent years, banking digital transformation has become a mandate, and BFSIs embrace the identity-centric security paradigm. Depending on technology maturity, almost all BFSIs now have a multi-cloud strategy at play.
A multi-cloud strategy offers various benefits, including easing development processes and ensuring high-quality, agile, secure, and scalable software solutions. As a result, many BFSIs are adopting market-leading public clouds (AWS, Azure, and GCP) and fostering private cloud offerings (e.g., based on projects CNCF, serverless, open source, and more).
The Choice of Cloud
For each service or application, BFSIs choose to go for public, private, hybrid, or multi-cloud based upon their enterprise architecture principles. In my view, the security needs of the specific service still should play a significant role.
A few example scenarios:
- The BFSI wants to max out traditional on-premise infrastructure and only wants elasticity from the public cloud resulting in hybrid cloud.
- The BFSI cannot host some application data outside the perimeter, and the public cloud is just used for processing but not for storing the information, which also results in hybrid cloud.
- Some new initiative service has to leverage the best of each cloud demanding huge horizontal and vertical scale with multiple 9s of availability and top-notch performance resulting in multi-cloud. If this blends a private cloud, too, then it’s multi-hybrid cloud.
The technical fact is that possibilities with the cloud are endless: policies, blueprints for consistent deployments, elasticity, scale, security, availability, auto-failover groups with SQL, multi-region writes with NoSQL, the choice of APIs for the different needs, MongoDB API, Gremlin API, Cassandra API, Cosmos DB API, guaranteed throughput and I/O, storage redundancy, access tiers, performance tiers, network flexibility VPC, site-to-site, point-to-Site VPN, encryption at all levels in transit or at rest, container orchestrators, and many more from SaaS to PaaS to IaaS.
The Evolving Challenge
Taking an example of one of the leading BFSIs, that uses a multi-cloud for AI and machine learning (ML) for the following financial use cases:
- Trading decisions using ML in the interest rate market
- Natural language processing in equity investing
- Clustering data to enhance returns in the equity market
- And more...
The typical initiatives are:
- Anomaly Detection – unusual patterns
- News Analytics – sentiment summarization and trading signals
- Quantitative Client Intelligence – insights from multi-channel client communications to improve service
- Smart Documents – extraction of meaning out of lengthy text sources to reduce manual operations
- Virtual Assistants – automated responses to client queries to gain operational efficiency
- And more...
Figure 1: An example from leading BFSI for an AI and ML ecosystem over multi-cloud
It’s evident that BFSIs use AI and ML for financial data. General models don’t work, and therefore the BFSIs need a skilled ML workforce.
Additionally, AI / ML is an evolving space; up-skilling engineering practices is an ongoing job. BFSIs are supportive of approaches to automate the evaluation of these models and make them more robust.
While BFSIs are busy exploring and using these specialized areas with multi-clouds, they see the 3Vs challenge (Volume, Variety, and Velocity of data). Every cloud provider has monitoring, analytics, and orchestration services but achieving end-to-end observability is not easy with a DIY approach.
An Observability Pattern for Multi-Cloud
Figure 2: An example architecture pattern for observability in multi-cloud environments
BFSI services that don’t have sensitive data should be scoped for a pattern that provides three-dimensional observability:
- Observability of machine’s data
- Observability of automation platform data
- Observability of IT service management (ITSM) processes data
Observability of Machines Data
Each public cloud provider has a set of monitoring and analytics capabilities. Striking a balance between cloud-specific monitoring and developing domain-agnostic AIOps is not simple. It’s a journey.
Implementing AIOps for each step provides deeper observability:
- Outside-In Experience - For mobile channel user experience insights, wrap the mobile apps for iOS or Android. For web channel experience, inject snippets in the headers. For global responsiveness and full-page performance insights, rely on synthetic checks for web and APIs.
- Apps Inside-Out - The monolith has been built over the years, and the additions of microservices keep enlarging the footprint. Modern APM has monitoring agents for applications running on traditional and modern platforms like Docker Swarm, Openshift, and Kubernetes across multi-cloud environments. The benefit is the application of topology and machine learning to find the root cause within seconds.
- Infrastructure - This space has a broader scope of consolidation for many point monitoring tools via one omni-capable infrastructure monitoring component, which works across traditional infrastructure and multi-cloud environments.
- Networks - This includes managing a complex maze of networks comprised of software-defined modern architectures, legacy devices, and connected multi-protocol end-points. A comprehensive and scalable network monitoring approach helps a move from reactive to preemptive network management. Key use cases are the network topology for fault isolation, network performance management, flow and packet analysis, and more.
- Domain-Agnostic AIOps - This layer sums up the four above and provides business line service analytics, service alarming, situations, alarm noise reduction before ticketing into ITSM, centralized dashboards, predictive analytics, capacity analytics, and more.
Observability of Automation Platform Data
An automation platform plays a significant role for BFSIs. Typically, BFSIs have many fragmented automations for database, file transfers, system tasks, batch processing, dynamic workloads, and more. To comply with business continuity, BFSIs have to prove their fitness via disaster recovery drills at a regular frequency, requiring another set of orchestration.
On top of these, multi-cloud has its own set of orchestration capabilities. An overarching automation platform orchestrates across all these islands of isolated automation and provides observability per line of business, forecasting, critical path, and more.
Observability of ITSM Processes Data
BFSIs are highly detail-oriented and careful with making changes into production environments, and many leverage topology (network, app, service topologies) for change impact analysis and approvals. This scenario is an example of observability of one dimension (machine data) helping another (process).
The new additions to production are evaluated via production transition design gates. Predictive analytics and capacity analytics both play a role here. Incident management analytics (canceled, closed without action, etc.) helps in governing monitoring rules. The massive chunk of repetitive requests serves as an input for improving automation maturity in new areas. These are more examples of observability dimensions complementing each other.
End-to-End Unified Observability with AIOps
Traditionally, organizations have adopted various point monitoring tools to gain insights into the multiple data sources mentioned above. This approach no longer works when faced with the complexity of multi-cloud enterprise environments and can become extremely costly to maintain. To gain unified visibility and manage the chaos, adopting an AIOps solution, like Broadcom’s DX Operational Intelligence, is the first step to success.
Visit Broadcom Enterprise Software Academy’s AIOPs resource center to learn more about the value of AIOps.
Tag(s):
AIOps