Data centers of all sizes are investigating Web-Scale architectures to see if they will meet their needs. In our recent webinar, “Web-scale vs. Enterprise IT“, we asked enterprise IT planners “What Appeals Most About a Web-Scale Architecture?”. The top vote getter by far was “The ability to scale”. But the ability to scale means more than just the ability to grow big.
What is Web-scale?
Built from a cluster of commodity servers, Web-scale architectures provide cost effective, linear scale. Each node (server) adds compute, storage, and network to the aggregate cluster. Organizations like Google, Twitter, Facebook, and Amazon use Web-scale to meet price demand and remain competitive. Scaling, as our polling question indicates, is a key aspect of Web-scale that has caught the attention of traditional IT organizations. But these organizations need to find a Web-scale architecture that can do more than just scale big. It has to start small, scale in granular increments and scale without interruption to applications.
One of the most important scalability requirements is the ability to start small. Just like cloud providers, data centers come in all sizes. The Web-scale architecture needs to start small enough that the organization does not have to over-buy just to get started. One of the benefits of Web-scale is that it enables businesses to attain high levels of resource utilization, meaning maximum use of every resource of a node before another node needs to be added to the cluster. If the organization has to buy a “starter” system that is larger than they need, then the architecture is inefficient.
The second requirement should be the ability to scale in as granular a fashion as possible. In Web-scale, compute, storage, and networking are just-in-time inventory items. Each resource’s (when added) rate of utilization should ideally be high from the moment it is installed. If each upgrade results in the over provisioning of resources by a wide margin, then the architecture is inherently inefficient.
In short, the inability to start small or to scale in small increments leads to buying too much hardware. Since hardware typically decreases in price over time, it makes sense to only buy hardware when you actually need it.
Scale Without Interruption
Finally, each addition to the Web-scale architecture should occur without business interruption. This means integrating another node to deliver additional compute, storage, or networking resources to the cluster should not require any downtime. It also means that failed or old nodes can be systematically replaced when the IT administrators have the time. A failure of a component should not create a “drop everything and scramble” event for IT. Web-scale architectures should be able to be stay in a protected state even after a component failure. This allows IT to address the failure during a scheduled service window.
Storage Swiss Take
There are other aspects of Web-scale to consider, like ease of use and compatibility with existing applications which we cover in our webinar, “Web-scale vs. Enterprise IT”, now available on-demand. When it comes to scale, IT professionals need to understand that scalability means a lot more than just scaling big. Scale also means the ability to start small, scale granularly, and be able to perform scaling without interruption.