Managing persistent data workloads in a multi-cloud architecture is critical for organizations hosting their apps on-premise and in public cloud environments. Aditya Tripathi and Judy Wang from VMware spoke at SpringOne 2020 Conference on Wednesday about architectural best practices to help simplify a multi-cloud strategy.
Several organizations are leveraging multi-cloud architectures to build robust disaster recovery and business continuity strategies. Moving stateful workloads to multi-cloud enviroments is complicated and organizations need to modernize their data architectures to work in this new cloud infrastructure.
A multi-cloud data strategy involves solutions to questions like where should the data replication be done, in the database layer or messaging layer?
Tripathi and Wang took the Theory of Constraints approach on how to run data in multi-cloud. Their process includes the following tasks:
- Identify business outcomes
- Identify limitations
- Prioritize tradeoffs
- Brainstorm solutions
Identifying business outcomes includes developing the rationale behind why an organization needs a multi-cloud architecture. The rationale can be that the apps can recover quickly when a hurricane hits (disaster recovery), or the apps are always available and serving customers (high availability).
The next step is to identify limitations and factors the organization is bound by. These factors can be business or organizational (for example, budget, influence factors, skill & expertise, and compliance), or they can be technical (such as networking or hardware limitations, heterogeneous application requirements, and managing legacy data platforms).
In the prioritize tradeoffs step, teams need to decide what they want to optimize for, between different design considerations. For example, is quick recovery more important than the system being always available? Tradeoffs can be maintenance and management related, or they can be vendor related. Vendor tradeoffs include services marketplace selection in public clouds, data redundany and vendor lock-in.
In the last step, teams brainstorm for solutions to identify what tools and services can help them achieve their goals. This includes options like data replication (active-active vs active-passive or sync vs async), cloud footprint (all on-prem vs all public cloud vs public-private hybrid), and what data technologies are used for caching & replication, messaging, event logs etc.
Business applications with tight service level agreements (SLA) must have a recovery strategy at the data center level. The teams also need to consider the data soverignity regulations like GDPR where they may not be able to spin up the infrastructure in a public cloud to manage their organizational data.
Wang explained the multi-cloud data strategies with couple of examples on traditional disaster recovery and point of sale, and solutions for each example.
For more information on SpringOne 2020 conference, checkout the main event website, schedule, and Day 1 and Day 2 highlights posted by VMware team.