According to the "RightScale State of the Cloud" report, virtually every company (84%) that uses infrastructure-as-a-service (IaaS) or platform-as-a-service (PaaS) clouds uses more than one provider. Most employ three or more public clouds from leading providers such as Amazon Web Services (AWS), Azure, and Google Cloud Platform (GCP).
There are a number of trends driving this multicloud proliferation, including the need for greater agility, flexibility and scalability, better network performance, improved risk management, avoiding vendor lock-in, and getting more-competitive pricing.
Each of these cloud platforms uses a different, built-in identity system that creates separate silos. Meanwhile, identity-as-a-service (IdaaS) solutions that manage access to SaaS apps introduce yet another separate identity store. Despite this accelerating cloud adoption, many organizations still run the majority of their critical business applications on-premises using a hybrid model, creating even more identity silos.
As a result, the coexistence of multicloud and hybrid clouds introduce an inherently distributed architecture that spans multiple platforms and identity systems. This "new normal" is transformational, not incrementally different from what came before.
Adopting a Distributed Security Model
The challenges associated with implementing security that can span multicloud and hybrid-cloud architectures is being complicated by the fact that identity has become the new security perimeter because many users are accessing apps and data from outside of the firewall.
Initially, many organizations try to set up multicloud infrastructures manually but quickly discover that identity management efforts don't scale due to a lack of staff and expertise, which results in misconfiguration errors, slow process times, and increased costs. In addition, most organizations have significant investments in legacy systems they must maintain while migrating to new cloud infrastructures. This will require IT departments to support a mix of cloud and on-premises apps for the foreseeable future, further complicating the multicloud challenge.
Here are five steps that will lead organizations to the creation and implementation of a successful multicloud identity strategy.
Step 1: Embrace Decentralization
The best way to solve multicloud identity problems is with a distributed architecture. Orchestration software can be used to serve as a connectivity "fabric" between cloud and on-premises identity systems and unify policy management and enforcement.
Step 2: Modernize! Don't Migrate Legacy Infrastructure
Don't move messy identity silos as part of migration projects. Instead, upgrade and modernize identity capabilities with new identity-as-a-service options. This makes it possible to consolidate policies and simplify roles and groups, check for compromised credentials, lock out dormant accounts, and force users to securely reactivate their accounts.
Step 3: Unify Visibility into App and Identity Ecosystems
To effectively implement a new identity strategy, it's important to understand where existing apps and identity systems are located. What are the dependencies between each app and identity systems? What identity workflows exist today and who has access to what? Having the right tools can speed up and streamline this discovery and mapping process.
Step 4: Use Standards, Not APIs, for Integration
Avoid lock-in to proprietary systems or writing to outdated APIs when updating identity systems. Instead, use standards like SAML, OIDC, and SCIM. Rather than performing multiple 1:1 integrations, an abstraction layer can provide 1-to-any flexibility. This approach also eliminates the need to rewrite apps in order for them to interoperate with a new identity system.
Step 5: Develop a Phased Migration Plan
A complex migration is made easier by breaking it into smaller, more manageable phases. For example, grouping migrations and planning an incremental shift will minimize disruption and risk. Grouping apps in terms of their migration complexity also helps anticipate and plan for potential speed bumps during a migration, for example:
- Complex = Cookie-based sessions and IAM SDKs
- Medium = HTTP headers and custom attributes
- Simple = SAML and OIDC standards-based sessions.
Review the capabilities you currently use that can be carried forward and determine what new capabilities will be needed to support the multicloud environment. There will be some level of coexistence support required until all legacy infrastructure can be decommissioned. Finally, calculate the costs that will be saved by retiring legacy infrastructure, which can be earmarked to fund other innovations.