Signs It's Time to Move to Metal

Lately, we all have been hearing a lot about "cloud repatriation", how companies are moving their workloads from virtual environments in hyperscalers back to on-premises or hybrid environments. As we know, Metal offers the best of both worlds: cloud-type management with bare metal servers and direct network access. Of course, before you even can figure out how to move, you need to know when it is time to move.

What are the signals and signs that it's time to move to Metal?

It’s important to note that there is no single reason that you should move workloads and each scenario should be evaluated on its unique circumstances and objectives.

What are the main signs?

Sign #1: Timing

Considering a move when you’re in the middle of a contract, let alone right after you signed a new multi-year commitment, sometimes doesn’t make much economic sense; after all, why would you pay Metal and your current hyperscaler for compute to host the same workloads?

However, even if you are in the middle of that contract, sometimes it does:

• Do you have other workloads that fit better in the cloud? If so, you can use your existing contract for those, and move your critical loads to Metal.
• Is the cost difference so large, that moving to Metal more than outweighs the sunk costs of your hyperscaler contract?

Even if not, inevitably, a window of time will come up where you should reflect on the cost and performance of your application. Even if you are at the very beginning of a contract, you should prepare for the end by calculating how much time and cost it would take you to move those workloads. This, in turn, will give you a window of time before your multi-year contract ends, when you should prepare for a potential move.

Less predictably, when a cloud provider discontinues support for a specific service on which you depend, you have an excellent opportunity to reevaluate, even if you are in the middle of a multi-year contract.

Sign #2: Overrun costs

It’s notoriously easy for cloud costs to become overrun - to the point where it’s not uncommon for organizations to have someone on their payroll to keep track of cloud costs. Indeed, entire businesses have arisen to help companies manage and negotiate their cloud costs.

Between idle and unused resources, over-provisioning resources, and the lack of visibility into cost details it’s easy for cloud costs to become a frighteningly high monthly figure.

The most common complaint that we hear from customers, and that an on-premises solution is a great fit for, is the cost of data transfer and egress costs. Cloud providers typically charge for both a data transfer cost and a data access cost. In addition, there often are charges for data transfer between zones in the same region!

With Equinix Metal, network connectivity is a fixed cost, and there are no charges for traffic within a Metro. Further, it is a great first step, as it is relatively simple to re-architect your application to pull data from a new network source.

Sign #3: Regulatory requirements

SaaS applications with a global presence (like Uber and Paypal) often operate in highly regulated industries and countries that have strict data sovereignty regulations. Applications are putting data in the cloud but need controllership for Personally Identifiable Information (PII) and sensitive data. Further, they are highly sensitive to customer-perceived performance; no one wants their Germany-targeted application to be zippy in the USA but slow and laggy in Germany!

If your regulatory and compliance requirements become tighter, affecting not only where you can process and store customer data, but also what can share the same hardware, Metal has your back. Metal's metros with precise locations ensure better control and adherence to regulations. The broad global distribution of Equinix locations further enables you to deploy your workloads as close to your customers as needed.

Sign #4: The cost of NOT doing it

Let's expand a bit on a point we made in Sign #1: Timing. Specifically, we said that even if you are in the middle of a multi-year contract, sometimes the nature of your consumption can change so much, that it is worth moving even if you leave some of the contract on the table.

For example, let's say you are in the middle of a 3-year contract with a hyperscaler. You have a 3-year plan to grow your business, and you expect to store and access a certain amount of data, and build your calculations on top of that. But (lucky for you) business absolutely takes off. Instead of the expected 50 PB of data that you figured maybe you would hit at the end of the 3 years, you aren't even halfway through, and you already have 150 PB of data! Even more, you are on track to double that in the next 3 years.

You do the math, and your cloud storage, and especially your data access and egress costs, are far far higher than you expected, and leaving you at risk of really "breaking the bank".

In this case, the cost of NOT moving to Metal is far higher than the cost of moving, even if you leave some of your contract on the table.

What to look out for when repatriating your workload

We’ve now decided that it’s the time to migrate our workload over to the cloud. What are some things to do as we start that process:

• Figure out an order of operations
• Get the right team together with the relevant skills, including hardware, software and operating system engineers, and the project manager
• Use an on-demand service with an easy-to-use Web UI or CLI, like Equinix Metal, for experimenting and creating small proofs of concepts
• Figure out your specifications
• Determine your data storage needs
• Get hardware! This can be one of the following:
  • Running it yourself on premise (which, likely as not, means a colocation facility, like Equinix)
  • Using a true bare-metal cloud, like Equinix Metal

Going down the Equinix Metal path, some helpful details in planning out the technical aspects of your migration are available in our Journey to Metal.

Follow-up and evaluation

Once you have moved your workload over, you will want to evaluate how it went and what you learned. This is critically important, both to evaluate your own decisions, as well as to make further migrations easier.

1. Keep track of your lessons learned so you can do it again with other workloads. Do this throughout the process, rather than waiting for the end.
2. Track costs and savings and see if it makes sense to do other workloads. You know your pre-migration costs, you are a whiz with Excel and calculated your expected post-migration costs, now see how accurate your predictions were. Did you underestimate some, or overestimate others? Use those to adjust future calculations.
3. Update your security processes and architectures based on your security posture. Your overall security posture probably changed, as happens with any such change. What changed? How did you adjust?
4. See if you can find ways to manage both your Metal resources and your hyperscaler resources. Obviously, if you moved everything over (lucky you!), there is nothing to worry about, but likely you have resources in multiple places. Some applications can provide a single pane of glass for multiple clouds and on-prem resources.

Further Reading and Inspirations

There are many great readings about migrating to the cloud and back again. David Heinemeier Hansson recently moved hey.com to on-prem; it’s a great read about a real life repatriation:

• Why we're leaving the cloud
• Sovereign clouds
• Our cloud exit has already yielded $1m/year in savings