A review of industry studies of the service mesh indicates that Istio has an early lead as a preferred control plane. However, service meshes usually also have a data plane component as well as other value-added functionality. The mix-and-matching of these often open-source components into bundled solutions is an ongoing process and may impact the eventual winners and losers in this space.
A recent report by StackRox shows that adoption of Istio in production doubled over the last year, but usage is still minuscule; Only 4% of the respondents say their organizations have it in production environments. Although the use of Linkerd in production stagnated, it is benefiting as companies are paying more attention to service meshes. With last year’s introduction of Linkerd 2.0 may also result in increased uptake.
It is noteworthy that 86% of respondents in both the late 2018 and the December 2019 surveys said their organizations are using Kubernetes as one way to orchestrate containers and microservices. It is probably necessary to look at other metrics, such as a number of production workloads using microservices, to better understand when you actually need a service mesh.
As of today, The New Stack defines service mesh technology as a dedicated infrastructure layer that adds functionality for service-to-service communication. As the market matures, we expect a continuing debate about interoperability between service meshes as well as between related technologies like API gateways and ingress controllers. In the meantime, we are noticing a willingness to take a second look at alternative open source projects, building homegrown systems and bringing in vendor-centric solutions.
The New Stack will continue to interview leading-edge users to determine how different use cases and challenges are affecting the evolution of service mesh technology. Also, check out our recently completed survey.
The table below presents service mesh-related results from five surveys conducted in the last year. The New Stack’s survey was about service meshes, so it obviously over-sampled people that are using or actively considering the technology. Other differences in reported adoption levels may be due to question-wording and a lack of understanding among respondents about what a service mesh actually is.