Now that your marketing team has started planning heavy promotions for the upcoming Black Friday/Cyber Monday sales, your IT team should be in full force getting the hosting infrastructure ready to handle the peak traffic. With holiday season sales projected to account for 30% of overall annual revenue for some retailers , the pressure on the IT team to ensure systems run smoothly is high. This is important – as retailers compete for consumer spend this holiday season you don’t want to push your customers to your competitor because you are 1) down.
Many believe that cloud is a fool-proof way to handle peaks and valleys in traffic. While cloud does allow you to scale up and down at a rate unimaginable on bare metal infrastructure, it is not entirely automatic. Having helped our customers prepare for and operate through the holiday season for nearly 15 years, we’ve learned a thing or two. Here are the 5 main tips to consider when preparing your cloud infrastructure for the holiday peak traffic.
1) Forecast demand for the upcoming sale
We had two customers last year who experienced a short service interruption simply because they didn’t accurately predict how much traffic they could potentially generate during their Black Friday sale. Some organizations forecast the demand using their gut-feel or a rule of thumb. Some are a little more advanced and look at industry reports and/or historical traffic for past 5 years. Others are way ahead of the game and are leveraging machine learning algorithms.
Artificial Intelligence (AI) tools can get way more advanced with predicting the actual demand for the upcoming sale. They could take into account things like user profiles and behavior, relevant product category performance on competitive websites, and even run a query on keyword search trends.
We do realize that not every company has access to a fancy machine learning platform (although those are becoming a lot more popular), and therefore always recommend to triple the expected increase in traffic to be safe (i.e. if you predict a 25% increase, add 75% to your capacity). It is always better and cheaper to overestimate than underestimate!
2) Decide on your scaling strategy: horizontal vs vertical
The first step in planning your cloud capacity management strategy is to use your projections to determine what kind of scaling you will use this holiday season. You have a few options based on whether you want to scale horizontally (add more instances) or vertically (increase the size of your existing instances).
To help decide, you should know the impact of how scaling or adding instances affects your licensing. If you are not able to add more instances because of licensing restrictions, you can increase the size of your instance. This does require a service interruption and will obviously increase your bill. Make sure you plan for this event.
If you are scaling horizontally, determine if you are going to scale automatically based on load or pre-scale your environment. A good rule of thumb: if your environment has to scale more than 3 times its starting size within one hour, auto-scaling is not for you. Auto-scaling is also not suitable for all applications and needs to be considered on a per-case basis.
3) Prepare for auto-scaling
If you do decide to take advantage of the auto-scaling option, make sure the threshold set to trigger automation is right for your application.If you set it too high, the site will suffer before new instances are spun up. If it is too low, you are incurring additional costs unnecessarily.
Auto-scaling needs to be configured to not only scale up properly but also at the right rate. Test and review your auto-scaling configuration including setting the minimum and maximum number of live servers. If you don’t set your maximum to the appropriate level your bill might surprise you.
Ensure your virtual machine image contains the latest code release and OS level.
Ensure your latest code release is in a repository that can be accessed by new instances (servers).
Load test your application when the environment is scaled at different levels. For example, you can test with a single instance to understand how many users that instance can handle and at which point performance begins to degrade.
4) Plan for capacity beyond your instances
Instances are just one part of your cloud capacity planning process. Understanding when to scale your RDS (Relational Database Services), your admin instances and other components should also be assessed.
Two areas that are often forgotten are the elastic load balancers and SES (email services). Make sure to inform the cloud provider or your managed service provider when traffic is expected to increase. This will allow their teams to pre-warm these services so they can process the extra traffic.
5) Work with your managed services provider to help you prepare
Cloud hosting providers offer a wide range of services to support you in your capacity planning strategy but they could also be a tad overwhelming with a number of services on the list. Talk to your dedicated Managed Services Provider who can provide in-depth expertise on a given cloud provider and suggest the best course of action for your organization.
We have put together a list of services specifically developed for a Holiday Readiness program to help our clients sail through the holidays without a hitch. You can learn more about these offerings here.
Latest posts by Aisling McCaffrey (see all)
- Five Tips to Prepare Your Cloud Infrastructure for the Holiday Season Traffic - September 24, 2018
- 3 Steps to Get Your Digital Innovation Journey Started - May 30, 2018
- What not to miss at IRCE 2018! - May 23, 2018