ovh
Support for ARM-based computing instances
Yeah definitely, ampere altra based instances would be very cool. We have switched to hetzner because these vms and servers offer a significant performance and price benefit. With the coming arm 9.2 platform it will be even more interesting as confidential compute workloads will be supported. Would be great if ovh could support this architecture.
Hi, We are investigating ARM-based instances and we will try to validate some Hardware profiles in the next quarters with our engineering teams. As a product manager, I will be really interested to better know :
- your typical machine size in number of vCPU and GB of RAM ?
- your system disk needs ?
- your private and/or public bw needs ?
- which specific OS do you use on your ARM machines ?
- intended use in pure IaaS or via Managed Kubernetes ? Thanks for your inputs !
Hi, We are investigating ARM-based instances and we will try to validate some Hardware profiles in the next quarters with our engineering teams. As a product manager, I will be really interested to better know :
- your typical machine size in number of vCPU and GB of RAM ?
- your system disk needs ?
- your private and/or public bw needs ?
- which specific OS do you use on your ARM machines ?
- intended use in pure IaaS or via Managed Kubernetes ? Thanks for your inputs !
8 Cores / 16 GB RAM 120 GB 1 Gbit Debian (Want to self install from rescue system via debootstrap) Pure IaaS
Hi, I'm the dev behind <PirateWeather.net>, and was looking into moving some/ all of my infrastructure off AWS (the $15k/year bill is getting a little unreasonable, and it's not as Canada friendly as it used to be). This is exactly the sort of thing that I'd need to spin it up on OVH, and in particular,
- 2-4 cores and 8-32 GB RAM
- 100 GB NVME (this is key)
- ~ 1 Gigabit bandwidth
- Any Linux OS (it's all docker based)
- Pure IaaS!
Happy to answer any questions or try out some new infrastructure!
Hi, We are investigating ARM-based instances and we will try to validate some Hardware profiles in the next quarters with our engineering teams. As a product manager, I will be really interested to better know :
* your typical machine size in number of vCPU and GB of RAM ? * your system disk needs ? * your private and/or public bw needs ? * which specific OS do you use on your ARM machines ? * intended use in pure IaaS or via Managed Kubernetes ? Thanks for your inputs !
We have two different uses cases, for which we already saw a cost strike by passing them on ARM, so I describe them separately.
- Quite light loads who needs high resiliency, essentially pass-thru APIs and monitoring, positioned with placement groups.
- Heavy workloads for which basic SLA is ok.
So, for the 1:
- 4 cores / 16GB
- 20 GB space
- ±1 Gb bandwidth
- Any Linux OS (as long as Kubernetes deal with it)
- Full Managed Kubernetes
And the 2:
- 16/32 cores / 64/128GB
- 1TB to 5TB space
- ±1 Gb bandwidth
- Any Linux OS (as long as Kubernetes deal with it)
- Full Managed Kubernetes
For our SAAS product we are planning to utilize OVHCloud for european customers.
Below you will find an overview of instance types we currently use in AWS that work well for us.
All our processing heavy components run on m7g Graviton instances, https://instances.vantage.sh/aws/ec2/m7g.4xlarge
CPU GB
m7g.4xlarge 16 64
m7g.8xlarge 32 128
m7g.12xlarge 48 192
m7g.16xlarge 64 256
For our analytics datastores we use m7gd Graviton instances since they come with local ephemeral NVMe SSD that we utilize for caching, https://instances.vantage.sh/aws/ec2/m7gd.12xlarge
CPU GB
m7gd.8xlarge 32 128
m7gd.12xlarge 48 192
m7gd.16xlarge 64 256
We migrated everything last year from regular m5/m6i instances to Graviton. We gained 20-40% performance improvements across all components (mostly GoLang and Java) as well as 20% cost reduction.
Answers to your questions
-
your typical machine size in number of vCPU and GB of RAM ? see above
-
your system disk needs ? SSDs up to 10TB
-
your private and/or public bw needs ? 15,000 Mbps (15 Gbps)
-
which specific OS do you use on your ARM machines ? Linux (currently using Amazon Linux)
-
intended use in pure IaaS or via Managed Kubernetes ? Only Managed Kubernetes
Arm Instance:
- vCPU: 4
- Memory: 16gb
- Disk: 60gb
- Network: 1+ Gigabit bandwidth
- usage: Managed Kubernertes
Hello, I hope I'm not too late on this thread but it would be nice to have small instances for testing purposes or very small edge ARM servers, like 2 vCPU + 8GB RAM (or even 4GB or RAM, Hetzner cloud provides some 2 vCPU + 4GB RAM instances for a really small price). This kind of usage would be mostly as pure IaaS for me but 2 vCPU + 8GB RAM can be a nice fit for a small managed Kubernetes cluster.
