UCONN

UCONN
UCONN

Compute Engine

 

Compute Engine 

Google Cloud Compute Engine, a service that lets you run virtual machines (VMs) on Google’s global infrastructure.

Here is a summary of the key takeaways:

Core Benefits

  • Flexibility & Scale: You can easily scale resources up or down based on your needs.

  • Cost-Efficient: It follows a pay-as-you-go model, meaning you only pay for the resources you actually use.

  • Customization: It supports any preferred operating system and leverages Google's high-end security and networking hardware.

Common Use Cases

  • Development: Hosting websites and creating testing/dev environments.

  • Heavy Workloads: Processing massive datasets, running scientific simulations, and training machine learning models.

  • Reliability: Serving as a backup for disaster recovery by replicating on-premises workloads to the cloud.


Four machine families offered by Google Cloud’s Compute Engine, which are categorized based on their balance of CPU and memory.

Core Concept: Machine Families vs. Types

  • Machine Families: Broad categories tailored for specific workloads.

  • Machine Types: Specific configurations within those families that define the amount of vCPU, memory, and storage.


The Four Google Cloud Machine Families

Family

Best For...

Key Features

General-purpose

Web servers, databases, and dev environments.

Best balance of price and performance.

Compute-optimized

Gaming servers, video transcoding, and scientific modeling.

High CPU performance and large core counts.

Memory-optimized

Large-scale analytics and memory-intensive databases.

High memory-to-core ratios; handles data sets in-memory.

Accelerator-optimized

AI, machine learning, and deep learning.

Uses specialized hardware like GPUs and TPUs.


Key Takeaway

Choosing a machine family depends entirely on your workload requirements—whether you need a cost-effective all-rounder (General-purpose) or specialized power for AI and data processing (Accelerator or Memory-optimized).


Virtualization & Computing

Virtual machines are "chopped-up" segments of physical hardware that allow multiple users to share one system.

  • Flexibility: VMs are decoupled from physical hardware, allowing them to move or scale without downtime.

  • Access: Users can SSH into simple VMs for basic servers or customize complex clusters.

  • Performance: Being "close to bare metal" means minimal abstraction layers, giving your code direct access to the physical CPU's power.

Storage & Disks

Storage is the foundation of a VM instance, acting like a highly available, replicated physical disk.

  • Types: * Standard (HDD): Good for large sequential operations and cost-effective bulk storage.

    • SSD: High performance for random operations with no moving parts.

    • Local SSD: Offers the highest possible performance by being physically attached to the host.

  • Performance: Disk speed is directly tied to IOPS (Input/Output operations per second). Importantly, disk size and performance are correlated; larger disks are faster.

  • Constraints: You can increase disk size at any time, but you cannot decrease it. Disks must be in the same "zone" as the VM instance to attach.

Data Management: Snapshots vs. Images

While both act as templates for new disks, they serve different purposes:

  • Snapshots: Used for backups. They use "differential storage" (only saving changes since the last backup) to allow you to restore a disk to a specific point in time.

  • Images: Used as starting templates for deploying new disks across an organization.

Scalability & Instance Groups

Instance groups allow you to manage a collection of VMs as a single entity.

  • Autoscaling: Automatically adds VMs when traffic or CPU usage spikes and removes them when demand drops to save costs.

  • Self-Healing: If a single VM becomes unresponsive, the system marks it as "dead" and replaces it automatically.

  • Rolling Updates: Allows you to deploy new software to a fraction of your fleet at a time, ensuring the system remains stable if the new update has bugs.

Create a Compute engine instance -  search for Compute Engine


Click into link


Hit Create Instance


Select Machine Size


Price varies depending on Machine configurations





No comments:

Post a Comment

Assignment #12 due 12/12/25

  Build 4 graphs using machine learning - linear regression I want two separate publicly traded companies e.g. AAPL & AMZN Linear regres...