SpotInst, the Israel based startup focussed on streamlining the use of spot instances in the cloud, today announced their Series B funding of $35 Million. This round is led by Highland. Existing investors, Leaders Fund, Vertex Ventures, and Intel Capital are part of this round. SpotInst is a company that has been under my radar since Google I/O 2017 and Rishidot Research has spoken with Kevin McGrath of SpotInst about how they fit into the Cloud Native ecosystem. Even though SpotInst started off with helping customers use Spot Instances on top of AWS, they have expanded further into other cloud providers and going beyond virtual machines on spot instances to containers and serverless.
SpotInst offers a new way to consume cloud
SpotInst’s main product is ElasticGroup which is a multi-cloud management and application scaling platform which helps customers leverage spot instances more effectively. Since they have been hosting customers for few years and have lots of data on the trends associated with how cloud providers release and take back instances from the spot instance pool, they could tap into machine learning to predict 15 minutes in advance the end of life of a spot instance. This capability in their management platform makes the user experience with spot instances seamless and customers need not sweat about any potential disruption due to an unexpected removal of spot instance. More importantly, this gives an even more seamless user experience in a multi-cloud scenario.
The biggest problem faced by users in the multi-cloud is the barrier to seamless consumption and seamless management of workloads across clouds. SpotInst solves this problem and allows customers to effectively tap into both spot instances and reserved instances to ensure that their applications are always available. More importantly, their platform helps customers realize dramatic savings and in some cases up to 80% savings. They do this for virtual machines, containers, serverless and GPU (which was also announced today). Since SpotInst uses on-demand instances as a fallback option, the users are never left in trouble with unexpected downtimes.
A new trend in cloud consumption going forward?
Even though SpotInst service appears to be a good model for the consumption of spot instances and serverless functions (essentially jobs running for shorter periods of time), I believe that this will change the very face of cloud consumption. I have argued for long that the underlying infrastructure doesn’t matter because of the higher-order platform abstractions like container platforms and serverless. But, for long, these multi-cloud abstractions served a limited purpose because of the costs involved in moving the workloads across cloud providers and other operational headaches. SpotInst and other platforms similar to theirs could help change the consumption model.
One of the advantages of multi-cloud is the potential cost savings. As cloud providers compete with each other, both through cuts in cloud instance pricing and building excess compute capacity at scale, cloud users can benefit by using the cost-effective resources from different cloud providers for their workloads at any particular moment and tactical use of spot instances. A platform like SpotInst makes this experience seamless for developers and has the potential to change the nature of cloud consumption.
Possible use cases
Some of the use cases where this kind of consumption model may prove effective are:
- A microservices architecture where different components of an application are independent of each other is a perfect candidate for this consumption model. Microservices being relatively lightweight, a platform like SpotInst can easily move the services across cloud providers. Ever since cloud computing became the norm in enterprise IT, we have been advising enterprise applications developers to focus on resiliency than robustness or reliability. With SpotInst consumption model, we argue that application architectures should focus on high levels of resiliency. The traditional cloud thinking is to assume that virtual machine instances will fail and build resiliency on top of it. We are arguing that application architectures should move one step further and assume that the underlying compute is short lived and ensure that the workloads are resilient enough to handle this. If the traditional cloud world ushered in the idea of stateless twelve-factor applications, we expect this new consumption model to use what we call as disposable applications, aided by containers and serverless functions
- A high availability architecture where spot instances from other regions or cloud providers can serve as failover instances
- Disaster Recovery use cases where spot instances and fallback on-demand instances are a great fit for this consumption model
As organizations embrace microservices and serverless functions for the newer applications, it will be easy to shift to this new consumption model of spot instances and multiple cloud providers. Running legacy monoliths on the cloud is a challenge. Similarly, running 12 factors apps using this consumption model will be a challenge. However, newer resilient architectures with disposable (micro) services will be a good candidate for the next generation of cloud consumption. SpotInst is well placed to kickstart this new consumption model. Thoughts?