What is High Availability?
High availability (HA) refers to a system or component that provides a measurably high level of uptime and redundancy. It is a core component in the design and delivery of a business continuity plan, and provides consistent access to your network, applications, data, and services during the most critical periods of business activity.
The advent of virtualization has had a dramatic impact on HA. Prior to virtualization, a server was purchased, the OS installed and then the applications providing key business services added. Problems arose when a server failed, and key business functions went down along with the hardware. Over time, IT departments began to segment the applications and services in a one-to-one ratio for the physical server and OS. Therefore, a loss of hardware resulted in only losing a single application or component in a system. Over time, this led to physical server sprawl, increasing capital expenditures, operational costs and servicing.
As hardware became more powerful, the utilization of the hardware purchased decreased dramatically. At the beginning of the virtualization movement, a physical server typically averaged around 11% CPU utilization. Consolidation was the primary objective of the initial waves of physical-to-virtual migrations. OS-to-application ratios remained fairly unchanged; therefore the end state of consolidation was a many OS-to-single physical server implementation. This led to a dramatic increase in risk per physical host.
However, virtualization technologies (combined with shared storage) permit the migration of live running virtual machines (VMs) to a new host. This enables administrators the ability to perform maintenance by shifting workloads from one physical host to the next. If a physical host should fail, the existing VMs (which physically reside on the SAN) can be powered up on a separate physical host within a few minutes (sometimes seconds). This HA scenario is, and should be, inherent in any cloud service deployment.
Fault tolerance (FT), where two virtual machines are clustered and treated as one machine but reside on different hosts, can also be achieved through virtualization. This HA scenario provides no interruption in the event of a physical host failure. There are some limitations, and this configuration cannot be deployed in all business cases.
HA began with hardware redundancy, grew with virtualization, and is fully maturing in the cloud at the application and services layer.
The Role of Geo-diversity
Geo-diversity plays a role in HA because it increases the availability of IT services by diversifying the physical location, providing protection against larger regional outages. This is typically deployed in disaster recovery (DR) solutions. For example, when a catastrophic event renders a data center inoperable, a secondary site can be brought online in a geo-diverse infrastructure. To take this further, we can provide geo-diversity to applications and services. We see this deployment in large public cloud providers with data centers all over the world. This is why you may never know where your data is located physically at any given time. Latency plays an important role in geo-diversity. Some applications simply do not cooperate if latency falls outside of specific design parameters.
The Future of HA
Customer demand is driving technology to accommodate HA solutions. These services and technologies are pushing the exponential growth of data centers. The future holds significant improvements where applications and services are geo-diverse and self-healing. HA at the application and services layer will provide the ability to flex workloads based on demand as well as to self-heal. For example, a web company may require more web services during peak hours. Should any servers fail, new servers can be automatically spun up to take their place. Demand for scalable resources is no longer limited to specific industry as more companies are asking for higher availability and performance at lower costs.
Going forward, businesses will consume applications and services with the same expectation we have today for electricity. Workloads will be “to go,” performing with the same elastic foundational computer network and storage on a plane as they do at the office. HA will enable your capacity to do business, at all hours, with imperceptible changes in availability and performance.
In the future, growth in the volume of data centers is expected to heal geo-diversity issues. Even more exciting is the expectation that automation will create self-healing services and applications. Whether your server experiences a power outage or security threat, your unaffected servers will perceive the interruption and duplicate themselves into a new location, without you even noticing the shift.
HA is foundational here at Peak 10. Rest assured that we’re doing all we can to drive towards higher availability and flexibility in our clients’ workloads.