The edge market continues to grow as emerging enterprise low-latency applications demand workload processing outside of the central data center or cloud. But still, many enterprises remain handcuffed by edge server utilization rates and growing costs. This has so far capped edge compute adoption. Could serverless be the answer for a more pragmatic approach to edge deployment?
Registered users can unlock up to five pieces of premium content each month.
Log in or register to unlock this Insight.
What is Serverless Edge?
Edge computing is a model in which computing and storage are moved outside of the cloud/data center and closer to the end user. This means that data can be processed with much lower latency, which has enormous value for verticals as it provides the capabilities to support performance sensitive applications. Despite this value, adoption has not been universal with many enterprises continuing to run workloads in the cloud, in private data centers, or even on legacy mainframes, hindering performance and deployment of use cases that could have substantial enterprise value. This aversion to edge deployment comes down to several problems that enterprises need to contend with:
- Utilization: Edge compute resources are only necessary for certain applications. This means that when these devices/applications are not running, edge compute resources will not be utilized. This leads some enterprises to reject owning edge compute servers.
- Build and Maintenance: Edge compute sounds good on paper, but often enterprises do not have the space, skills, time, or money to build and maintain the servers. Managing on-premises edge servers is often challenging and can often disincentivize adoption.
- Processing Scalability: Public cloud servers can scale automatically, but edge servers are power constrained meaning that they only have access to a certain number of processors. This could create processing bottlenecks, especially in verticals with significant fluctuations in data creation.
With the slow adoption of edge computing, serverless edge could provide enterprises a reason to shift. Serverless edge is founded on similar principles to cloud serverless. By abstracting the hardware from the software, serverless edge is able to provision enterprises with compute resources close to the end user when they are needed. This model means that you only pay when your code is working on the servers. An example of how serverless edge works: an event is triggered that requires edge compute processing, the edge platform resolves the request to the nearest Point of Presence (PoP) with the lowest possible latency, the servers perform the workload, and then the response is delivered to the end user. These resources are best used for mildly latency-sensitive microservices with significant demand fluctuations.
What Value Could Serverless Create for Edge Deployment?
- Cost Efficiency: Deploying and managing edge servers is costly. Serverless edge outsources these to a third-party and allows enterprises to transition towards a ‘pay-as-you-use’ model. This means that resources will not be underutilized.
- Real-Time Scalability and Elasticity: Owned on-premises servers are inherently constrained by their deployed resources (core, accelerators, memory/storage). Serverless edge can scale up/down to meet any workload demands. This means that during peaks/troughs workloads will not face resource bottlenecks.
- Quick Time-to-Value and On-Going Software Updates: Serverless edge removes any hardware deployment or management lead-times from the enterprise, allowing them to spin up compute resources if and when they are needed for their workloads.
- Reduce Energy Usage: The on-demand nature of serverless edge means that the resources are only ‘on’ when computing workloads. Especially now, this is a key value driver with geopolitical and sustainability challenges placing energy consumption at the forefront of mind for Chief Information Officers (CIOs).
- Poor Performance: The principle challenge for enterprises looking to deploy serverless edge is the latency impact of a ‘cold startup’. A ‘cold startup’ refers to the delay that occurs when a workload is executed on an inactive server. Added ‘startup’ latency means that sensitive applications must still be run on owned edge servers.
- Sovereignty: Serverless edge will provision workloads to a PoP that provides the best performance (which is dependent on latency and availability). However, enterprises (especially in Europe) are facing emerging data sovereignty regulation that requires data to stay within the country or even region it is created. Not being able to control where edge data is processed will slow serverless edge adoption.
- Inflexibility: Currently, commercial serverless platforms requires users to write in code using the infrastructure provided. This means that microservices are purpose built for one location contributing to lock-in. This could potentially have long term cost and flexibility implications, as well as restricting potential multi-cloud/edge strategies.
Serverless Edge is Certainly Not Perfect, But it Could Unlock Edge Growth
At present, serverless edge brings a substantial tradeoff for enterprises—it may lower costs across the board for edge deployment, but it will not deliver the ultra-low compute latency that mission critical applications demand. Overcoming this performance constraint created by a ‘cold startup’ should be the priority for vendors. But this added latency does not mean that serverless edge does not have a pragmatic value proposition for certain enterprises. Startups and Small-and-Medium Enterprises (SMEs) with limited funds, limited IT skills, or even limited on-premises space will see substantial value for low latency, but not ‘mission critical use cases’ like personalized advertising or real-time bidding platforms. To take this value proposition to the next level and potentially extend it beyond resource constrained enterprises, vendors must look to solve some of the solutions’ major features:
- Performance: Deploy more PoPs/edge nodes to decrease the physical distance between the end-user and the compute resources. Potentially, cloud players could look to expand partnerships with telecoms by collocating PoPs within their existing infrastructure. Serverless edge vendors must also look to reduce the impact of ‘cold starts’. Eliminating ‘cold startup’ induced latency will make serverless edge’s more appealing from the perspective of application developers and larger companies looking to optimize their edge strategy.
- Flexibility: Open sourcing platforms to reduce the ever-looming threat of vendor lock-in and support enterprise multi-cloud strategies.
- Observability: Develop and integrate software to support enterprises utilizing multiple Content Delivery Networks (CDNs)/PoPs to run microservices. Applications are comprised of multiple microservices, run across multiple clouds, and leverage more than one CDN, so software should provide enterprises and end-to-end view across this entire environment.
Solving these challenges will encourage adoption across the entire spectrum of enterprise sizes and verticals. But even though this market exhibits a significant opportunity (with some minor tweaks), it has remained largely quiet. Out of the hyper scalers, AWS (Lambda @ Edge) is the only one to make the transition from serverless cloud to the edge. Their main challenger is Cloudflare Workers Unbound, with the remainder of the market made up of smaller cloud players (i.e., Vercel, Fly.io, Deno, Fastly). It is very surprising that both Azure and GCP have yet to make the jump from serverless cloud to edge. Both of these cloud giants have mature CDNs with widespread deployment of edge nodes/PoPs, providing them with the necessary infrastructure to support serverless edge. It is unclear what may be holding them back unless they do not see themselves successfully competing with incumbent AWS. But given the competitive history of these cloud giants, most would expect these players to jump in the ring soon either by acquiring an existing solution or through in-house development.