Cloud migrations can get expensive fast. Between licensing, networking, and training employees on the new system, costs can easily spiral out of control.
But perhaps what costs the most during cloud migration planning is the assumptions that people managing the migration bring to their project.
As professionals with deep expertise in the cloud space, we at Akasia have seen these assumptions cost our clients thousands of dollars over the years.
But it doesn’t have to be this way. With thoughtful planning and help from experts, you can plan a cost-effective cloud migration and save up to 80% TCO in the process.
Here are our top 5 cloud myths to avoid when planning your lift and shift migration.
Jump to a specific cloud myth:
- Cloud Myth 1: Using BYOL is always cheaper.
- Cloud Myth 2: Cloud native VMs are cheaper.
- Cloud Myth 3: One-to-one mapping is simpler and cheaper.
- Cloud Myth 4: We should manage uptime on cloud like we do on-premises.
- Cloud Myth 5: Networking costs are not important until later.
Cloud Myth 1: Using BYOL is Always Cheaper.
BYOL, or Bring Your Own License, is a cloud migration strategy in which you migrate your existing on-premises license to cloud, rather than converting to cloud native or VMware-on-Cloud licenses.
You may think it is simpler and easier to BYOL to your new cloud environment. And while this can be true in some cases, often BYOL makes your migration more difficult and expensive in the long run.
Let’s take a look at a few specific cases. Using BYOL Windows OS or SQL on Microsoft Azure is generally cheaper than other licenses, and you receive unlimited virtualization on Azure dedicated hosts. Similarly, there are a number of financial and logistical incentives to bring an Oracle license to Oracle Cloud.
AWS, Google Cloud, and IBM Cloud allow moving licenses between on-premises and cloud. However, bringing your own license to cloud can present more difficulties and costs in the long run.
Firstly, BYOL are more often used with dedicated hosts. Dedicated host scenarios require you to pay for full 24/7 use of that host, whether the VMs you store there are on or off. This limits your availability to take advantage of uptime management cost savings, as you will have to pay for the dedicated hosts regardless of VM uptime.
Secondly, when you BYOL to Azure, AWS, Google, or IBM Cloud, you are responsible for right-sizing. Thorough right-sizing is necessary to maximize savings, but difficult to do on your own without assistance from an AI tool, like Akasia Infrastructure Modeler. Often, this means that you will end up paying more than you would in a right-sized environment when you BYOL.
Lastly, BYOL can cause problems when decommissioning VMs. With BYOL, it is almost always your exclusive responsibility to keep track of your license, and ensure your compliance with all its requirements. If you migrate this license to a new location or vendor, you must ensure compliance with that environment’s rules and regulations as well.
Microsoft licenses on Azure do offer an eased burden, and AWS has launched a license manager to help with the manual compliance process. But with all other clouds, you must endure strict audits to ensure compliance, which over time can be quite arduous and costly.
In summary, each workload’s situation will be different, and your specific costs depend on VM usage. If you have a great software enterprise deal with an existing license, it is likely worth looking into bringing that license to the cloud. Otherwise, the overhead and managerial costs of managing your license and ensuring compliance on cloud can overshadow any benefits of BYOL.
Cloud Myth 2: Cloud native VMs are cheaper because we don’t have to pay VMware license costs.
This is a common and pervasive myth among cloud architects. The logic is: why add VMware licensing costs on top of existing cloud native costs? How could that possibly be cheaper?
In our experience, it is probably cheaper than you think.
First, let’s talk about the migration process. It is a lot of work to convert on-premises VMware to cloud native VMs. The process is complex and requires format conversion, significant testing, and noticeable downtime (which can decrease revenue). Third party providers that can help with the cloud native conversion process, but with these services comes added costs.
Transforming on premises VMware to VMware-on-cloud solutions is much simpler in comparison. There is no need for format conversion, with less testing and downtime.
Right-sizing, while still important to VMware-on-cloud migrations, is much more critical for cloud native VMs. That is because in VMware-on-cloud environments, VMotion and distributed resource scheduling in the background take care of load balancing for you, saving you money automatically.
It’s absolutely critical to right-size each VM on cloud native to maximize savings—with VMs packed too loosely you waste money, and with VMs packed too tightly you sacrifice performance.
Similarly, uptime management is important for cloud native VMs. Failing to do so is the principal cause of runaway cloud costs. With VMware-on-cloud, you manage ESX services and VMware handles resource allocation for VMs. Therefore, most of your uptime management is already taken care of.
When it comes to operations and management costs, cloud native VMs are new models with new tools. This will require re-training for system administrators and IT staff, which will cost money.
With VMware-on-Cloud, administrators will have a familiar paradigm and familiar tools. This will make the transition for the team easier and less costly. You will likely only need minimal incremental training for staff.
And finally, cloud native requires you to manage each VM independently, which can be extremely tedious for large workloads. With VMware-on-cloud, you manage your environment at the vCenter level. This is much easier and simpler for your system administrators.
The bottom line is that given all these factors, transforming on-premises VMware to cloud VMs can be 15-30% lower than current on-premises costs in our experience. This assumes on-premises commitment ratio is low and you manage uptime and right-size.
Akasia Infrastructure Modeler provides cost planning for lift-and-shift migrations from on-premises to VMware-on-Cloud. It discovers on-premise resources and provides equivalent and right-sized cloud templates in minutes. Akasia’s automatically generated reports provide a complete cloud bill of materials and costs that form a starting point for lift and shift migrations.
Transforming to Vmware-on-Cloud can be significantly less expensive. You can save 40% or more than current on-premises costs with optimal bin-packing. Our clients have found 40-60% savings with VMware on Cloud– it’s more economical than you might think.
Cloud myth 3: One-to-one mapping is simpler and cheaper.
One-to-one mapping is a migration planning process in which companies submit server information to a cloud architect, who will then find a comparable server on cloud for each VM.
The problem is that there are so many choices when it comes to cloud configurations. On each platform, there are different virtual machine types, storage types, number of regions, instances, tenancy variations, etc. This multiplies quickly to a dizzying 2 million options for VM migration, before including right-sizing.
Moreover, right-sizing is key because most on-premises VMs are over provisioned. One-to-one mapping without right-sizing means you will likely be over provisioning in the cloud and paying more than you need to.
And right-sizing before you migrate is essential—changing and resizing once an instance has been deployed is extremely difficult and disruptive.
Right sizing requires analysis of CPU usage, CPU type/class, memory usage, storage usage, and network I/O. Once you select a host, you must find the best fit placement for each VM.
Trying to do this while one-to-one mapping is difficult and time consuming, and is arguably impossible without automation.
One-to-one mapping may seem cheaper in the short term, but you will pay for it in the long term with over provisioned VMs and a poorly sized cloud environment.
Cloud myth 4: We should manage uptime on cloud like we do with on-premises servers.
Uptime management is not typically a concern with on-premises VMware. On-premises servers are always on, whether they are being utilized or not.
With the cloud, there are more opportunities to manage uptime and save. This necessitates a different approach.
In brief, there are 3 main types of instances on the cloud: spot, reserved, and on-demand instances.
Spot instances can be preempted anytime with little notice, because your workload will be removed if someone pays more or has a more urgent need than you. While these tend to be an economical option, their lack of reliability make them appropriate only for stateless applications.
Reserved instances are VMs that you commit to use for a period of time. You can receive deep discounts for committing to reserved VMs for a given period of time. Reserved instances can achieve up to 30% cost savings relative to on-premises, and these instances usually break even around 60% uptime.
Generally, on demand instances are the most expensive option, but can be a cost-effective option for systems with less than 60% uptime. It’s especially important to manage uptime here to avoid runaway cloud costs.
When done correctly, uptime management in the cloud can save you about 15-30% in overall cloud costs.
Cloud myth 5: Networking costs are not important until later.
Networking is usually the last thing considered by our clients, since networking costs are a given in cloud migrations.
However, networking can end up being very expensive. It’s important to factor these costs into your overall cloud migration budget.
Since ingress costs are free for all clouds, most people end up spending a significant amount on egress costs.
Egress traffic can be destined for on premises, another cloud, or another region in the same cloud. Intra-region traffic typically has a free tier, and then a charge by usage that varies from cloud to cloud.
Unfortunately, there is not much you can do to reduce bandwidth consumption. But, you can be aware of what you are going to need, model proactively, and negotiate costs with your cloud platform. Both of Akasia’s modeling solutions automatically include all hidden costs, including network and I/O costs, in our cost comparisons.
Some clouds, including IBM Cloud, offer intra-data center ingress and egress for free. In an environment where you’re replicating a lot of data (i.e. for HA), this can be a massive cost savings on a network-intensive workload.
In summary, it is critical to model out your network bandwidth and determine what network costs are before you choose your cloud. These costs may end up being significant, and can affect which cloud provider is most cost-effective for your workload.
As you can see, there are a lot of factors to consider when planning and executing a cloud migration.
We’ve shared our general advice, but every workload and every migration is different. Modeling out and comparing costs for BYOL, cloud native, and VMware-on-Cloud, on top of right-sizing and managing uptime, is extremely tedious to perform by hand or Excel spreadsheet.
Akasia makes cloud cost optimization and migration decision-making easy by enabling cost comparisons between:
- All major cloud platforms: AWS, Azure, GCP, IBM Cloud and Oracle Cloud
- VMware-on-Cloud vs Cloud Native virtualization
- On-Premises vs Cloud
- As-is vs Right-sized infrastructure
- Always-On vs Pause-and-Resume infrastructure
- BYOL vs License Included
Akasia Cloud is unique because of our easy to use models for on-premises to cloud or cloud to cloud migration planning. We also offer highly sophisticated VM and resource mapping algorithms, and unbiased cost comparison of the full stack of cloud services beyond just VMs. Akasia’s Infrastructure Modeler and Cloud Modeler are road tested and proven to save our clients time and money, with over 1.5 million customer VMs modeled.