IT survival guide: Preparing for a major web event
Guest post by Stuart Franklin
There are many important aspects and considerations a company must consider before a major web event or launch. That old adage, "People don't plan to fail, they fail to plan!" applies to the most recognizable problem in any project launch. The way to avoid this issue is to do just that--plan, plan and plan some more! It starts with picking the right team and identifying their specific roles. A good project manager will seamlessly take the project from planning to execution, just like the right coach in baseball. In baseball, if no one knew who was playing outfield, first base, catcher, etc., team members would go anywhere and only do what felt right to them on the field, turning the game into a picnic without the meal. The best and most successful project plans include staffing with detailed role identity, strategy, requirements and communication.
Aside from the team, plans are easy: a beginning, middle and end. Simple right? Not so if you lose sight of the goals. Don't forget that setting milestones or checkpoints is a great way to ensure you meet your goals.
Glitches in the system
During the entirety of this process, the project team needs to anticipate any and all glitches. Many times these glitches are the hidden details that either were not captured in requirements gathering, or had been assumed to be obvious. The devil is in the details! Capturing requirements is so obvious on paper, but there are many pitfalls that easily work their way into the equation. For example:
- A web service or database connection passes all tests in the process of execution, however, when it goes live for testing you can't even access the application.
- The application works, it just doesn't capture the user's information because no one considered the firewall settings or a critical DNS entry.
- A Secure Sockets Layer (known as SSL) requirement for web forms or logins may not work because the certificate wasn't installed correctly or the form is pointing to another site.
It's usually the small things that are the biggest problem because most people don't sweat the small stuff well, at least not until it becomes big. Simple name changes when upgrading a site can cause a myriad of issues. Making sure that the old name (e.g. ...MySite\* to ...\OurSite\*) is one of the more common oversights. Redirects on a server level should be reviewed not only for an upgrade or fundamental site enhancement but even when starting fresh with a new site.
Worried that your network will be overwhelmed by an influx of visitors? Great! Performance testing is essential to measure the new system/site with simultaneous simulations of many visitors. "Expect the unexpected!" is the motto for performance testing. In my experience while performance testing, the most common problems discovered relate to individual server limitations. For example: The servers are put into a Virtual Local Area Network (or VLAN) to ensure their own guaranteed bandwidth. In eliminating this other network "noise" (e.g., service broadcasts) we've stumbled on settings that limit how much bandwidth our new servers get. While this is common, the more common issues are buried in the systems, such as limited database connection pools, network card and operating system configurations that aren't set for anything more than the default. This is equal to a simple stroll in the park. Your new site is supposed to run marathons, not check out the scenery or feed the pigeons. These missed essential configurations can lead to disastrous performance hits.
Sharing the load is an effective way to overcome network limitations and save money. If your site is going to be global and your customers will be coming from almost anywhere, you could consider a Content Delivery Network. Using a CDN to cache a majority of your website's static content takes the load off of your individual servers. If the CDN is configured properly, it's a safe way to deliver content to your customer from a location that's closest to their computer. A CDN saves time and increases efficiency from almost every aspect of a website. Here's how: when you connect to your account on a CDN you are connecting to the Point of Presence that's closest to your Internet connection. Files are uploaded quicker and this is the same way it works for your customers. The website information comes from the POP closest to their Internet connection. Whether you're in America, Asia or Europe, the CDN knows where you are and, with mirrored copies of your site information, delivers it from the nearest POP. This provides a better user experience and quicker delivery to your visitors. It's not bad for your new site's SEO scores either. Search engines have more recently started basing search result scores on delivery times.
There should always be a Plan B. In the initial planning stage you should have addressed "continuous operation" through a mirrored hot stand-by, paired/load-balanced servers, and high availability options, etc. If planned well, you shouldn't have many worries regarding recovery in the event of failure. Every website critical to a business should have a Disaster Recovery Plan. This, for all intents and purposes, is your ultimate Plan B. Testing your DRP and capabilities is a part of the launch as much as performance, configuration and load testing. Thorough testing of the website's functionality, servers and network generally uncovers a number of issues needed for the final DRP.
Failures are disheartening. Sometimes even planning for failure can be depressing, but stick to the plan. Realize that every Plan B is just that: a plan that requires as much attention (and sometimes more) than the operational plan for the site. Personally, I find this to be a favorite and one of the most exciting parts of a project. My best experiences have stemmed from finding a better way to do operational functions while in the process of establishing the fallback/recovery position.
How can you make sure customers will stay or return if they can't access the site on the first try? One part technology, six parts design, add marketing and stir. The way to keep customers coming back is more complex from the visual and interactive components to a website (IT has the easiest part of the entire recipe if we did our job right). The planning, testing, sweat and tears should have brought us to a point where the system is always available. If we have to use our DRP, so be it, but we're up and running. With high availability, something interesting to present and quick webpage response times, consumers will always find their way "Home" no matter what extension they use--aspx, php, etc.
Top tips for running a seamless web launch
Safety in numbers: Create a "war room" on launch day to ensure a common location, quick assessments of issues or challenges and real-time communication. This could be the project manager's office, a conference room or an open conference call bridge.
Communicate often and openly: Every team member is involved for a reason--they add value. Open and frequent communication provides many tacit benefits. As an example, "We're still looking good!" instills confidence in the team members as the launch progresses. Other communications such as "Load is increasing approaching upper memory limits" get everyone on their toes and proactively looking for problem areas.
Monitoring can save your site: Some monitoring tools have come so far they are almost psychic. Most operating systems come with a very good basic set of monitoring tools. These can be significantly augmented through the use of more advanced tools such as PRTG, Splunk, Pingdom and others, each with their own set of advantages. Be sure to review these and other tools to meet not only launch, but long-term monitoring goals. The inherent analysis capability in the better monitoring tools can be invaluable at stopping a problem before it gets out of control.
Nerves of silicon!: How do you overcome the jitters when launching a website? You don't! Even the best project managers demonstrate pride of ownership though their concern. Nerves can be a good thing, just don't get overwhelmed. Great leadership dictates, "Don't lose your wits while others around are losing their web heads!" Calm and steady will always get you through the storm.
Stuart Franklin is the IT director for Siteworx.