Deliver the right amount of capacity for your apps on Lumen Cloud with
horizontal & vertical Autoscale, automatically
On Lumen Cloud, Autoscale makes it possible for a server or group of servers to self-regulate and deliver only the capacity it needs at any given time. This provides our public cloud users with two key benefits: cost savings, and a reduction in administrative overhead.
Instead of requiring system administrators to closely monitor and scale cloud servers based on changes in utilization, you can create policies that add and remove capacity automatically. So Autoscale ensures that you don't have unnecessary CPUs allocated or additional servers powered on unless you need them.
Ensure the number of servers increases automatically during demand spikes to maintain performance, and decreases automatically during demand lulls to minimize costs.
Define the policy characteristics by providing a friendly name, the minimum number of servers to remain on at all times, and the metrics (CPU and/or memory) and thresholds to determine when to scale out and scale in. The maximum number is determined by the number of provisioned servers with the group that the policy is applied to.
Set the threshold period, which defines how long a server must be at the minimum or maximum usage before an Autoscale action occurs. If a server is likely to have temporary spikes that don't warrant a scaling event, choose a longer period. Then set the cool down period, which tells the Autoscale engine how long to wait after performing one Autoscale event before considering another. This setting helps prevent a rapid fire set of Autoscale events before the server has a chance to recognize the positive effects from the initial scale event.
Horizontal Autoscale policies are applied to server groups and can be used on more than one group. An optional load balancer and port combination maybe applied, making horizontal Autoscale ideal for web server based workloads.
Automatically rightsize the compute capacity of your servers during demand spikes and lulls. Many enterprise workloads can see an increase in performance just by adding capacity to existing servers without the need for additional servers. This is often the case if the workload is purely CPU-constrained and not I/O bound.
Define the vertical Autoscale policy by providing a friendly name, defining a CPU Range which defines the allowable minimum and maximum number of CPUs for the server. Also define the policies scale thresholds and scale up increment, which determines how many CPUs to add (1, 2, 4) during a scaling event.
Set the threshold period, which defines how long a server must be at the minimum or maximum usage before an Autoscale action occurs. If a server is likely to have temporary spikes that don't warrant a scaling event, choose a longer period. Then set the cool down period, which tells the Autoscale engine how long to wait after performing one Autoscale event before considering another. This setting helps prevent a rapid fire set of Autoscale events before the server has a chance to recognize the positive effects from the initial scale event.
Finally, apply the policy to an existing server, or while you create a new server.
Autoscale makes it possible for a server or group of servers to self-regulate and only have the capacity it needs at any given time, based on real-world usage. This saves you money while reducing administrative overhead. Instead of having a system administrator closely monitor and scale servers based on an increase or decrease in utilization, you can create policies that add or remove capacity automatically. This ensures that you don't have CPUs allocated or additional servers powered on (costing you money!) unless you need them.
Many enterprise workloads can see an increase in performance just by adding capacity to existing servers without the need for additional servers. This is often the case if the workload is purely CPU-constrained and not I/O bound. For these types of workloads, vertical Autoscale may prove to be enough. If you find that your workload has very large spikes that require bigger bursts than vertical Autoscale may be able to provide, horizontal Autoscale is definitely a powerful option as well. You can always start out using vertical Autoscale, and later change to using a horizontal Autoscale policy instead, if you determine that it isn't enough for you. Or, if you know your workload causes your servers to become I/O bound, you may want to use horizontal Autoscale from the start. Also, horizontal Autoscale relies on load balancers to scale the workload. If you have a workload that doesn't scale out well this way (i.e. database servers), vertical Autoscale is the better bet. Conversely, the web tier is a great candidate for horizontal Autoscale as it scales out behind a load balancer well.
Vertical Autoscale can be applied to any server type that supports a "hot add" of CPU resources without a reboot. Those OSs include: Windows Server 2012 Datacenter Edition, Red Hat Enterprise Linux 5 or 6, and Ubuntu 10 or 12 x64. Horizontal Autoscale is applied at the group level and will work with any server types within that group.
Vertical autoscaling (scaling up/down) adds (or removes) CPUs to (or from) an existing server based on the policy thresholds set. This policy gets applied on a server by server basis. See details below on when servers are scaled up or down for more information on what triggers a vertical Autoscale event. Horizontal autoscaling (scaling out/in) powers on (or off) servers within a group and adds (or removes) them from the load balancer configuration based on the thresholds set in the policy. This policy is applied at the group level and will affect every server within the group. See details below on when servers are scaled out or in for more information on what triggers a horizontal Autoscale event.
Though it is not explicitly disabled by the platform, it is not recommended that you use both options in conjunction with each other (i.e. servers with vertical Autoscale policies inside a server group with a horizontal Autoscale policy). While this case should not cause anything to break, it can result in unpredictable behavior. (In one case, for example, if a horizontal Autoscale policy scales down, it may turn off a server that had a vertical Autoscale policy configured on it.)
A policy takes effect instantly.
While the minimum threshold period for an Autoscale policy is 5 minutes, the Lumen Cloud platform collects data points far more frequently for each server. In fact, each 5 minute threshold period contains hundreds of data points. This ensures that your server scales because of an overall utilization trend, not just an isolated spike in utilization over a few seconds.
Lumen Cloud is an elastic set of cloud services. That means you can tailor your environment to deliver the right amount of capacity during peak traffic levels, as well as during less busy intervals. Autoscale takes elasticity and scalability to the next level, by automatically adjusting resources, based on conditions you set, to deliver the optimum level of capacity. This delivers on one of the most powerful – and empowering – promises of Public Cloud: Pay only for what you need, when you need it.
Horizontal Autoscale policies work in tandem with Lumen Cloud’s load balancers to “scale out” and “scale in” virtual machines. This function is especially useful for sizing Groups of web servers that manage and process user requests to and from logical and database tiers of common applications. Horizontal Autoscale also can improve performance for purely I/O bound architectures.
Horizontal Autoscale just isn’t feasible for some tiers of application architectures – including database servers and other I/O bound applications. In this scenario, vertical Autoscale is a better choice. This feature adds CPU capacity to increase performance under duress, and then reduces compute resources as usage falls. Other common examples where vertical Autoscale should be considered: enterprise collaboration and messaging applications (like Exchange and SharePoint), batch processing services (e.g. visual effects rendering, simulations), and high traffic web applications.
Enterprise-grade public cloud servers available on-demand for legacy and greenfield business applications.
Persistent, high-performance storage for every enterprise scenario. Easily add backup & DR capabilities as needed.
Ensure maximum uptime and reliability, the Lumen Cloud offers a portfolio of cloud load balancers to meet the unique requirements of your apps.