In this post, we survey the offerings from major players in the cloud auto-scaling and see how they compare to each other, and to our own product, the Elastisys Cloud Platform. We have chosen to include both cloud infrastructure providers, who have native auto-scaling functionality to offer, and cloud auto-scaling-specific companies in our comparison.
In our comparison, we include the three major public cloud infrastructure providers: Amazon Web Services, Google Compute Engine, and Microsoft Azure. We also include cloud auto-scaling-specific companies: RightScale and Scalr.
All auto-scalers in this comparison can do the very basic tasks one would assume, i.e. react to the event that some threshold value for some metric has been reached and scale up or down accordingly. So, for this comparison, we look at the following more advanced set of features:
- Whether one can specify custom metrics or not.
- To what extent the software is open source.
- Whether one can specify scheduled scaling limits or not.
- To what extent the auto-scaler automatically recognizes recurring patterns, and uses those to anticipate future load (e.g. Monday morning always sees load spike when people come to the office).
- If the auto-scaler can predictively scale by an algorithmically determined amount to meet estimated future demand based on current usage patterns.
- To what extent the auto-scaler is able to manage cloud resources from multiple clouds.
Recognizing recurring patterns helps reduce manual intervention from system administrators or DevOps specialists. Predictively scaling based on current (previously unknown) usage patterns helps quickly converge on the correct amount of resources, which a more simplistic model of “add X instances if we reach threshold value Y” rule will not be able to do.
|Auto-scaler||Custom metrics||Open source||Scheduled scaling limits||Recurring pattern recognition||Predictive scaling||Multi-cloud capabilities|
|Amazon Web Services||Yes||No||Yes||No||No||No|
|Google Compute Engine||Yes (beta)||No||No||No||Uncertain3||No|
|Elastisys Cloud Platform||Yes||Partially2||Yes||Yes||Yes||Yes|
- 1 Scalr has an Open Source edition that is not exactly the same as one gets via their SaaS. It of course comes without support.
- 2 Elastisys offers its cloud pools as open source, which are the only components that need to contain your cloud provider credentials, so you can host these yourself to avoid disclosing secrets to us. The auto-scaling engine doing the calculations is proprietary.
- 3 The inner workings of the autoscaler in Google Compute Engine is not documented, but reading between the lines, one can guess that the amounts to scale are determined according to some formula other than merely “add two instances”.
Cloud Auto-scaling Comparison 2015 Summary
Microsoft Azure has opted for a very clean and simple look for their auto-scaler, but feature-wise, it also comes up lacking. Especially jarring is the inability to scale based on custom metrics and that there are no schedules (for virtual machines). Amazon Web Services provides schedules and custom metrics, but no more advanced features. Google Compute Engine rises above both Azure and AWS by using some undisclosed method of figuring out the amount of instances to scale up and down. This is a step up in administrator-friendliness, at least.
Scalr and RightScale both include monitoring systems and can scale on custom metrics. They can both be used for managing multiple cloud infrastructures. They do, however, not have any advanced features for recognizing patterns or any provisions for dealing correctly with sudden spikes. Only Elastisys Cloud Platform does.
Since Elastisys Cloud Platform also includes a monitoring system, and can scale on multiple metrics, it offers the fullest and most feature-complete auto-scaler on the market. It helps you manage multiple cloud deployments, across multiple clouds, with ease. And because it does not attempt to be an all-in-one solution, administrators can use their own favorite tools for all other tasks, such as configuring instances.
This comparison has not included every feature of the Elastisys Cloud Platform, only the ones related to auto-scaling. Please check out our features page for more information on what else our software offers you!
We look forward to doing this comparison again next year and see how the field has evolved until then! Did we miss anything we should have included? Let us know in the comments below!