'Queue the Questions' AMA Event: VM Families - Join Here!

[Michelle]

Welcome everyone to our 'Queue the Questions' AMA event on VM Families!

This week, we are going to answer your questions about the Google Compute Engine Virtual Machine Families.

You can find more information about the event here.

We are thrilled today to be joined by our special guests @ Moazzam, @ chelsie, and @SubraC, @jamiekinney to answer your questions about the multiple general purpose VMs to workload optimized VM families, including the new addition to our general purpose family, Tau.

Join us by asking your questions below! ⬇️

[Moazzam]

Welcome everyone. Looking forward to addressing your questions here.

[Luis88]

What are Tau VMs?

[SubraC]

Hi Luis88, 

Tau VMs are our newest addition to our Compute Engine VM family. Tau VMs are optimized for cost-effective performance primarily targeting scale-out workloads. For more information, please look at our blog : https://cloud.google.com/blog/products/compute/google-cloud-introduces-tau-vms

[Luis88]

Thank you Subra! Very helpful information 👍

[Francisco04]

When and how can I get access to Tau VMs?

[jamiekinney]

Hi Francisco04,

We are currently accepting requests to access the Preview of Tau VMs. GA is planned for later in the year. Please contact your Google Cloud account team if you would like early access to Tau T2D VMs.

[MickyLynn]

When should I use Tau T2D over other general purpose VMs?

[jamiekinney]

Thanks for the question MickyLynn!

Tau VMs have been designed to support scale-out workloads which benefit from higher per-vCPU performance. For example, Tau VMs offer excellent price-performance for horizontally scaled workloads like web and application serving, image processing, and media transcoding. Databases and other workloads that require our highest performance Local SSD and Persistent Disk storage may be better suited for our N2 and N2D general purpose VM families.

[eddyizm]

I have a VM that I did not update to the latest LTS (UBUNTU 20.04)  How can I update the image, to first 18.04 and then 20.04 via the console? I have a web server (apache) running a python/django? What's the safest way to do this without losing my config or site? I am ok with a little downtime. 

Welcome to Ubuntu 16.04.7 LTS (GNU/Linux 4.15.0-1098-gcp x86_64)

Do I follow the normal process of 

New release '18.04.5 LTS' available.
Run 'do-release-upgrade' to upgrade to it.

Any help appreciated. 

[jamiekinney]

Hi Eddyizm,

I recommend that you first take a backup of your VM so that you can easily revert back to your earlier configuration.  For persistent disk, you can use the PD snapshot feature to backup your block storage.  If you are storing any configuration or site data in GCS, you may also want to make a copy of that data.

Regarding the actual upgrade from Ubuntu 16.04.7 to 18.04.5, we have detailed documentation available here.

[eddyizm]

Thank you!

[eddyizm]

When upgrading, I get the error : 

Checking for a new Ubuntu release
Please install all available updates for your release before upgrading.

When I have already run the sudo apt update/upgrade commands.  I'm not sure if it is asking because I have not installed the security updates but I dont have that ESM SERVICE enabled. 

*The following packages could receive security updates with UA Infra: ESM service
enabled:
libpam0g linux-libc-dev libpam-modules libsystemd0 binutils openssh-sftp-server
udev libpam-runtime isc-dhcp-common apache2-data libx11-6 libudev1 linux-headers
-gcp apport linux-gcp python3-apport systemd-sysv liblz4-1 libpam-systemd systemd
libpam-modules-bin openssh-server libx11-data openssh-client linux-image-gcp lib
xml2 apache2-bin apache2 apache2-utils isc-dhcp-client python3-problem-report
Learn more about UA Infra: ESM service for Ubuntu 16.04 at https://ubuntu.com/16-
04

I believe I got stuck here previously. How do I get around this?

[jamiekinney]

Hi Eddyzim,

I recommend taking a look at this conversation for some suggestions.  If this doesn't resolve your question, I suggest that you reach out to our support team for assistance. We provide a range of support options, include for customers who are taking advantage of the Google Cloud free trial.

[daria_anderson1]

Hello! When should I use Local SSD vs. Persistent Disk?

[jamiekinney]

Hi Daria_anderson1,

Google Cloud offers a number of storage options.

Local SSD (docs here) is our highest-performance block storage. Local SSD is typically used for high-performance filesystems, caching systems, and databases which are able to run on ephemeral storage. 

Persistent disks are durable network storage devices that your instances can access like physical disks in a desktop or a server. The data on each persistent disk is distributed across several physical disks. Compute Engine manages the physical disks and the data distribution for you to ensure redundancy and optimal performance.

Persistent disks are located independently from your virtual machine (VM) instances, so you can detach or move persistent disks to keep your data even after you delete your instances. Persistent disk performance scales automatically with size, so you can resize your existing persistent disks or add more persistent disks to an instance to meet your performance and storage space requirements.

Google Cloud customers typically use Persistent Disks for the OS filesystem on their virtual machines and to provide storage for any workload that required highly-available, durable storage.

[daria_anderson1]

thank you!

[Oliver00]

I've heard that GCE supports confidential computing. How are customers using this feature?

[chelsie]

Hi Oliver00, Thank you for your question. 
Yes, Google Cloud supports Confidential Computing and ecrypts data in-use with Confidential VMs and Confidential GKE nodes. Our Goal has been to make Confidential Computing easy. The transition to Confidential VMs is seamless- all workloads you run today can also run as Confidential VMs. No extra code changes to applications, just one checkbox to protect against rootkit and bootkits. Read more here:  https://cloud.google.com/blog/products/identity-security/introducing-google-cloud-confidential-compu...
More on Confidential GKE Nodes here: https://cloud.google.com/kubernetes-engine/docs/how-to/confidential-gke-nodes

[diannegibson401]

Does Google Cloud support GPU accelerated VMs?

[chelsie]

Hi Diannegibson401, Thanks for the question! 
Within our Accelerator Optimized VM family, A2, we support Nvidia's A100 GPU. Google Cloud is the only platform that provides up to 16 A100 GPUs attached to a single A2 vm for mega performance. Read more about what you can do with the Accelerator Optimized Family here: https://cloud.google.com/blog/products/compute/a2-vms-with-nvidia-a100-gpus-are-ga.
We also have T4 and V100 GPUs avaialble via our original N1 VM family. To know which regions and zones each GPU is available in check here: https://cloud.google.com/compute/docs/gpus/gpu-regions-zones

[vinnie17]

Hi there! What are Google Cloud Compute Engine differentiating features?

[Moazzam]

Thank you for your question vinnie17,

Google Cloud Compute Engine is an Enterprise grade, innovative platform that easily allows you to manage your VMs through their lifecycle in a cost effective, sustainable way. Compute Engine provides Custom Machine Types which offer unique flexibility to configure CPU and RAM independently to only provision and pay for what you need to effectively mange cost. Live Migrations are able to keep your VM running even when a host system even occurs keeping your application more secure and reliable. Google Cloud provides sustainable compute power as we publish per-data-center quarterly power usage efficiency metrics and an annual environmental report. Accelerator Optimized VMs are also the only platform that provides up to 16 A100 GPUs attached to a single A2 vm for mega performance. You can read more here: https://cloud.google.com/blog/products/compute/a2-vms-with-nvidia-a100-gpus-are-ga . Memory optimized VMs are SAP Certified with up to 12 TB VMs. Just to list a few.

[vinnie17]

Thank you, Moazzam! 

[daria_anderson1]

Also, how should I determine which Google Cloud region to use? Thanks a lot!

[jamiekinney]

Hi again 🙂

Google Cloud currently has 27 regions, 82 zones and 146 network edge locations with new expansion regions coming soon. There are plenty of options to optimize for latency, redundancy, cost and regulation. https://cloud.google.com/about/locations

In choosing which Google Cloud region to host your application there are multiple considerations.

• Latency to your end users can be different from one region to the next:
• The price of services also differs from region to region.
• The electricity used to power your application may have a different carbon intensity.

This document explains how to include carbon emissions characteristics into the location choice for your Google Cloud services.

[MickyLynn]

What are ways to optimize cost on Google Cloud Engine?  Thank you.

[chelsie]

Hi MickyLynn! 
Google Cloud will automatically apply Sustained Use Discounts (SUDs) to your VMs that are running specific Compute Engine resources a significant portion of the billing month. (https://cloud.google.com/compute/docs/sustained-use-discounts)
Users can also purchase Commited use Discounts (CUDs) for Compute Engine resources such aas vCPUs, memory, local SSDs at a discounted price for 1 year or 3 years. (https://cloud.google.com/docs/cuds?hl=en)
Custom Machine Types (CMTs) are another way users can provision and pay for only exactly what is needed. For some use cases, you could save 50% or more compared to predefined types at other public cloud providers by configuring an optimal machine for the workload. (https://cloud.google.com/custom-machine-types?hl=en)
Google Cloud provides recommendations based on your usage data for right-sizing your VMs as well as reconfiguring your VM family selection. These recommendations can be found on the Compute Engine VM page as well as within the recommender hub.
Also Check out INF201 session at Google Cloud NEXT for more details on best practices of Cost Optimization. https://cloud.withgoogle.com/next/

[vinnie17]

Hi again! How can I choose between N2 and N2D VM?

[SubraC]

Hi Vinnie17.  Thank you for your question. 

Both N2 and N2D are suitable for a wide variety of general purpose workloads such as enterprise apps, most databases, web serving, media transcoding, image processing. N2D offers up to 224 vCPUs, so high throughput workloads that can benefit from higher number of threads can take advantage of N2D machine. Currently, we also offer confidential computing feature in the N2D family. This would be useful for workloads that may require memory encryption. N2 machines will be a good fit for workloads that requires high single thread performance.

In addition, the choice between N2 and N2D could largely be a matter of price-performance for a specific workload type. For this, refer to representative benchmarks published by Google at https://cloud.google.com/compute/docs/benchmarks-linux.  Please note, it would be good to benchmark your specific workloads to get an accurate comparison of price-performance that meets your unique workload's requirements. 

Lastly, customer's preference for Intel vs AMD processors could also be a factor to choose between N2 and N2D. N2 family is built using 2nd generation Intel Xeon Scalable Processor (Cascade Lake). N2D is built using 2nd generation AMD Epyc processor (Rome).

[diannegibson401]

Will GKE support Tau and other upcoming VM families?

[jamiekinney]

Thanks for the question DianneGibson401,

GKE plans to support Tau VMs on day one of the public preview.  GKE will also support other VM families as soon as they become available.

[Luis88]

One more question - Does Google Cloud offer advanced networking for HPC?

[chelsie]

Luis88, great question.
We have the options of 100, 75, and 50 Gbps high-bandwidth network configurations for General Purpose N2 and Compute Optimized C2 Compute Engine VM families. If you're running an application that is senstitive to both latency and throughput, you'll be able to combine compact placement policies on C2 instances with the 100Gbps bandwidth for suprerior performance. Read more in the blog here.

[Luis88]

Thanks so much! Thank you also for sending the blog post!

[Francisco04]

Can you walk through GCE's naming scheme - for example what does n2d-highcpu-64 mean?

[chelsie]

Hi Francisco04,

The first part of the name refers to the machine series. This will tell you the category of machine. N2D denotes AMD while N2 machines run Intel.
Each series is available as different machine types - that’s listed in the middle as "Highcpu". There are either preconfigured machine types or custom machine types. Each type has a different virtual CPU to memory ratio to fit different workload requirements, like “highmem”, “highcpu” or “standard”, etc. 64 listed at the end denotes the amunt of vCPU that is provided per machine. Bringing it all together: n2d-highcpu-64 means that the machine is an AMD machine in the Balanced family preconfigured with high cpu type with 64 vCPUs.

[MickyLynn]

One more question:  When should I consider using Memory Optimized VMs?

[chelsie]

Appreciate all the questions MickyLynn! 
Memory Optimized VMs (MOVMs) are made to ensure optimized performance for memory intensive workloads and workloads requiring more memory than what you can get on General Purpose VMs. The typical use case is large in-memory databases, in cache databases or SAP HANA. When sizing what VM is right for you, we recommend you visit [Memory Optimized VMs]
[SAP on GCP documentation page]