History
We started our first Magento Cloud project in 2019 and both Magento and us received several complaints about the bad performance (slow response-times of non-cached requests) from that customer back then.
We did several other Magento/Adobe Commerce projects since 2019, some hosted on Magento cloud, others hosted on different infrastructure solutions, which included custom multi-node AWS-setups and single-node setups.
Over all those years we started several attempts to optimize both the code-base of the projects and the configuration of the Magento Cloud infrastructure, which brought some improvements, but the resulting performance never came close to the one of the single-node setups that were used by most of projects.
By numbers we are calculating an average of 2.5-times the response-time for Magento-Cloud projects compared to projects hosted e.g. on a single https://www.hetzner.com/dedicated-rootserver/ax102/
Current situation
We received another complaint from a customer, again optimized the code-base and Magento-Cloud configuration, but also came up with our own considerations regarding the Magento Cloud infrastructure in general.
We initially raised those requests at our onboarding team, the customer's suggest manager, were forwarded to the support, which forwarded us to this section of the forum at the end.
Considerations
Overall architecture
To our knowledge the usage of a Magento Cloud infrastructure always implies the use of at least 3 different nodes on 3 different availability-zones. (Visual representation: https://experienceleague.adobe.com/docs/commerce-cloud-service/assets/az-stack-diagram.png?lang=en)
From several cloud-projects (AWS and Magento Cloud) we also learned that the usage of several nodes/servers for the different services increases the over-all response time, as network-latencies are introduced into the communication between the application and its services and the replication of the data between the same service on the different service-nodes. Those network-latencies are even worse, if the the different nodes are distributed over several availability-zones.
Is there a possibility to drop either the 3-node architecture, or at the least the use of different availability zones?
We are aware that this would reduce the redundancy of such a setup, but from the feedback of our customers a better performance would be more important to them then the redundancy currently offered.
An a number of those customers do not have the high number of traffic-volume that would require a multi-node setup.
Type of CPU
For our current customer executing a cat /proc/cpuinfo on one of the cloud-environments reveals the following data:
model name : Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
cpu MHz : 2095.081
So Magento Cloud currently provides a CPU that was released back in 2017 and has only a CPU clock speed of 2.1 GHz.
Whereas a hosting-option from another hoster (https://www.hetzner.com/dedicated-rootserver/ax102/) would give us a AMD Ryzen™ 9 7950X3D (from 2023) with 4.2 GHz.
And yes, we are aware that the most recent hardware is not available at every location of AWS or Azure, but at least more recent hardware with higher CPU clock speed would be available at both of those provided, some of it even for cheaper prices than the hardware currently provided.
https://learn.microsoft.com/de-de/azure/virtual-machines/linux/compute-benchmark-scores#hbv3
Is there a chance to get more recent hardware for our Magento Cloud customers?
