Total (consumer) spending for the (Thanksgiving 2012) weekend reached an estimated $59.1 billion, a 13% increase from a year ago, according to the National Retail Federation. Online spending on Friday alone topped $1 billion for the first time, according to the data-analysis firm comScore Inc.
We can analyze the seasonal patterns of traffic by studying the Alexa reports such as the one below for different retailers. In the case of walmart.com the holiday traffic is 400% the average during the year. It remains to be seen how much more the traffic would spike today – Cyber Monday 2012. It is clear that retailers like Walmart need the ability to scale up the e-commerce infrastructure during the holiday season.
A retailer’s e-commerce infrastructure would include software applications, middleware, server, network and storage resources. Virtualization technology has made scaling of e-commerce applications easier. However, virtual machines can only start on physical servers after they are racked, stacked and necessary network and storage access components have been properly configured. The preparation of the physical infrastructure is still mostly a manual task, unless they are using something like the Cisco UCS. The Cisco UCS Manager embedded in the system, handles the simplified configuration. With large environments (thousands of servers) infrastructure management software such as Cisco UCS Central becomes increasingly important. Cisco UCS Central, which simplifies compute infrastructure scaling, became generally available last week. With it, physical compute infrastructure management can now be policy based and automated. This in turn can increase the velocity of changes and reduce the possibility of errors while scaling rapidly. This blog by Steve Kaplan and the short video below give a nice overview of Cisco UCS Central software.
Generally we take into account scaling up, but what about scaling down? For instance if walmart.com added 4 times the number of servers for the holiday season, they need to manually change all configurations to repurpose the infrastructure for other projects in January. Without this capability the infrastructure capacity will be underutilized. Once again converged systems like Cisco UCS can come to their rescue with the service profiles allowing rapid repurposing of servers with the necessary network and storage access configuration.
Additionally, retailers with globally distributed datacenters can now standardize on setups and use centrally proven and tested configurations. A central policy setup capability will allow companies to leverage skills in a central location to manage distributed datacenters. The central policy management capabilities should also come in handy when disaster recovery sites have to be setup quickly. Check these solution overviews for more on Cisco UCS Central and partner solutions which integrate with it.