Turbonomic Blog

3 Ways to Avoid Cloud Lock-In

Posted by Ben Yemini on Oct 24, 2019, 10:40:00 AM
Ben Yemini
Find me on:

If you’re a regular reader of this blog or have ever had the pleasure of speaking with Dr. Shmuel Kliger about how IT is managed today, you know that thresholds hold a special place in “IT hell” for us.

Rather than discussing the issues with setting thresholds and monitoring them to observe abnormal behavior and then trying to get back to normal, I wanted to discuss a different type of threshold.

What is your monthly public cloud spend?

What should you do if you’ve passed $100k per month with a single provider?

Over the past decade thousands of startups have leveraged the ease of use of AWS to quickly set up and deliver services to end users without having to master the complexities of the underlying infrastructure. At the same time, we are seeing more and more of our enterprise customers place a significant amount of workloads in AWS or Azure and several are operating under a “move everything to the cloud” mandate.

Ending Up in Cloud Jail

CloudJail_In_b

As enterprises migrate and startups scale with one cloud provider they run the risk of ending up in cloud jail. Cloud jail is simply waking up to discover you’re spending way too much money on infrastructure and are completely beholden to one provider. Once this happens it becomes very hard to switch. Inmates have often architected for one cloud. It becomes very difficult to migrate to another (the network tunnels cost money) and re-architecting to operate across multiple providers is a significant undertaking.

Even for those that sleep well at night knowing they’ve chosen only one provider, they often know they are spending too much and have an opportunity to reduce waste without moving. In fact, a recent analysis from RightScale reveals that 35% of cloud spend is wasted via instances that are oversized and not optimized.

How to Avoid Cloud Jail

CloudJail_GetOutOf_b

So what should you do?

How do you avoid getting into cloud jail in the first place?

Here are some guidelines to consider:

Identify the signs that you may be heading into cloud jail and act before it’s too late. This should depend on primary benefits you are trying to achieve.

1. Cost Optimization

If your cloud journey is all about cost optimization, make sure you are tracking all of the various services you are consuming from a provider (e.g. instances, load balancing, SQL and NoSQL services) and have a good sense of the costs. If your usage patterns are predictable (i.e. application demand rarely changes) you can probably manually look at opportunities to save. As you see cost go up per month and approach that $100k threshold realize that having multiple providers will give you options and the cost for one service on AWS vs. Azure vs. Softlayer vs. Google will always vary. Having options will keep you in the driver seat.

2. Great App Experience

If your cloud journey is all about that next great app and user experience is paramount to delivering the best app you can. Make sure you are set up to tracking the end-user response time and performance of your app from day one and be particularly leery of network performance issues your provider can’t or won’t fix (e.g. packet loss for certain geographies or internet providers). If you can’t solve issues by leveraging CDNs or SD-WAN acceleration you may want to think twice about your chosen provider. Remember that cloud providers will not take care of performance for you. The good ones will arm you with best practices for how to architect for auto-scaling and knowing that an instance can randomly disappear but getting the best end-user performance for your app is up to you.

3. Multi-Cloud From Day One

Enterprises and startups that are more advanced in their cloud journey start very early with the perspective that it is a multi-cloud work. As you start your journey, set up a test environment that has cross-cloud connectivity. This could be as simple as vCenter and AWS or more complex with multiple private and public connections. Even if you’re going all public in a small on-prem data center (e.g. a few RUs at a colo) that your engineering team can work in and experiment with will help in the long run. Having the skill sets on staff and thinking through what performance and cost you can achieve for which workloads as you scale will become easier when you have options.

As a next step see how Turbonomic can help with your multi-cloud architecture or if you’re looking at a sub-set of workloads and want to find the best placement in AWS or Azure check out our Cloud Cost Calculator.

Most importantly don’t end up in cloud jail.

Subscribe Here!

Recent Posts