Cloud Computing Final - 2nd Half

Supervisory Control and Data Acquisition (SCADA)

-data-gathering oriented -remote monitoring, remote telemetry, report back to control systems

Programable logic controller (PLC) (aggregators)

-handle specialized input and output -ensure uninterrupted processing Computer that performs some specific function you coded, burns code into firmware + continuously repeats it

ec2 instance savings plan v.s. compute savings plan for ec2

1. commiting to hourly spend of a specfic instance 2. flexible and can move to different instance types (more expensive per hour tho)

triggering a function

1. schedule driven (time or frequency) 2. event driven trigger (new file in s3 bucket)

5g

20 gigabits

Internet of Things

A global infrastructure for the information society, enabling advanced services by interconnecting (physical and virtual) things based on existing and evolving interoperable information and communication technologies

IOT control software (4)

AWS IoT core - connects + lets devices interact w/ cloud AWS iot device defender - security auditor AWS IoT device management - remote management at scale Aws iot things graph - connect devices + cloud services for iot applications

API gateway

added a web-interface to lambda in order to pull qualtrics results qualtrics triggers the API gateway, which triggers lambda

Kinesis

allows real time processing of large amount of data assists in data dumping + structuring

AWS Athena

allows you to access the files stored in a data lake using SQL without needing a database

IOT device software (2)

amazon freeRTOS - OS for managing microcontrollers Aws IOT Greengrass - software for local compute, messaging, data caching, sync

IOT data services (3)

aws iot analytics - sophisticated analytics w/ massive volume AWS iot events - detect and respond to events from sensors aws iot sitewise - collect, structure, and search IOT data

AWS Sagemaker

builds and deploys fully functional machine learning models in the cloud integrated environment for running jupyter notebooks in aws

overprovisioned services

capacity exceeds demand- services should be downsized

Cloud computing costs

considered operating expenses

cost optimization

cost engineering into design decisions

Data Warehouse

data stored structurally, optimized for online analytical processing (OLAP)

underprovisioned services

demand exceeds capacity- services should be upsized

Distributed Control System (DCS)

focus on controlling processes use sensors and feedback systems

Qualtrics survey

has interview logic, branching, data validation

Classification techniques

help assign instances to pre-defined categories

regression technqiues

help predict numeric values

Clustering techniques

help us identify groupings we may not have recognized

artificial intelligence techniques

help us recognize patterns in data using neural networks

Association discovery techniques

help us to identify common pairings

Cost awareness

individuals in org have a sense of what things cost

drawbacks of serverless

limit on resource consumption (3gb of mem) (max 15 mins) stateless nature - runs independently, no memory (state) between the operations of functions dev / debugging can be more difficult (complex)

when to use reserved instances

long duration + unchanging ex) running a ERP database

Data Lake

massive amounts of data stored in original format (csv), object store, unstructured so more like a swamp

cost management

monitor and manage cloud costs

Lambda

native support - write code within it triggers from most other aws services, also can trigger something to happen in aws (reboot server) transparent infrastructure (autoscaling) 1 million free requests (400,000 gb/s compute time) infinite scalability

prescriptive analytics

provide advice about actions we should or should not take "guiding our actions"

Serverless Computing

runs code on infrastructure is a PaaS (Function as a Service)

Why cloud analytics? (4)

scalability becomes easy complexity of data environment simplicity of cloud analytics tools manageable cost structure

IOT Primitives (5)

sensor aggregators (collecting and summarizing data) communication channel (allowing coms between them) external entity - the cloud decision trigger (business logic)

Reserved instances (convertible)

set aside capacity at a discount if convertible - can flex around the instance family

savings plan

set aside capacity at a discount - commiting to an hourly spend rate

SNS

simple notification service

benefits of serverless

simplicity of implementation (no servers) tremendous and immediate scalability (infinite) wasteless approach to CC, use only what you need manageable costs

when to use spot instances

something that doesn't generate new master data risk-tolerance dependent running reports/functions that arent entirely essential

When to use on-demand instances

something you've never done before for a known duration autoscaling (reserve for min, on demand for peaks)

Aggregators

summarize sensor output without 5g some of it done on device (edge)

Descriptive analytics

summarizing data we have about the past "understanding the past"

tagging

tagging resources allows the allocation of costs

Analytics merges three worlds

technology, business, statistics

Database

transactional data usually kept in relational form OLTP - online transaction processing

communication channel, external entities, decision triggers

transmit data products/ services implemented in hard/software conditional expressions that trigger actions

designing serverless

trigger inputs calculation output

predictive analytics

use past data to predict future outcomes "understanding the future"

Spot instances

utilize excess capacity of servers at extreme discount can be inturrupted within 2 mins notice differences in discount and frequency of interruption dependent on region!

Reserved instance (RI) Utilization

what % of RIs are being used by active workloads

Reserved Instance (RI) coverage

what % of active workloads are using reserved instances