Skip to content

Latest commit

 

History

History
188 lines (171 loc) · 8.91 KB

lec19.md

File metadata and controls

188 lines (171 loc) · 8.91 KB
Raw

Computer Systems - Lecuture 19 - Future of Computer Systems

Computer Systems

  • People are beginning to use tablets more than laptops and desktops
  • Online services
    • Desktop as a service
    • Citrix XenDesktop
      • Everyone connects to one server
    • Amazon Workspace
      • Top level machine for 22$ a month
  • u disrespecting my rig????
  • This has a big impact on Development
    • Web-first approach
      • People expect shit to work on whichever device and integrate with services
    • Collaborative tools
    • Multiple platforms
      • iOS, Android, Windows, Mac, Linux etc.
      • vs Making 1 purely online service

Computer Systems Servers

  • Multi-user (server)
    • Mainframe 1960s
    • Microcomputer servers
    • Large scale data centers
    • Virtualisation
    • Software as a Service
      • Video games
      • Email
      • Spotify
      • Office 365
    • Inrastructure as a Service
      • AWS
      • Virtual Machines
      • Servers
      • Storage
    • Platform as a Service
      • Google App Engine
      • Heroku
      • Execution runtime
      • Database
      • Web server
      • Dev tools
      • Targeted for app devs

Software as a Service

  • Where cloud services began
  • Enterprise applications delivered through the browser
  • Streamlines enterprise IT management
    • Vendor does all of the maintenence shit
  • Google Apps, Office 365, GitHub, GitLab
  • Entirely dependent on the vendor not being bad
  • Depends on internet connectivity
    • No internet = no job
  • Trust in a single provider
    • Failure means you lost all your shit
      • Also means provider may not be trustworthy
  • Privacy and Security Challenges
    • Where is data stored?
    • What protections does it have?

Infrastructure as a Service

  • Usually virtualised access to hardware instances
  • AWS EC2 is one of the biggest examples
    • Various hardware specifications available
    • Both shared and dedicated instances
      • Dedicated instances provide a greater degree of security and predictability of performance
  • Able to reduce hardware purchase and support costs
    • Allows rapid rescaling of resources - able to handle peak demand more efficiently and at a lower cost cause you can just turn on/off things instead of having to set up your own machines that just idle there when it's not peak hours
  • Software is still the responsibility of the enterprise
    • Amazon will not touch your software
    • Have to install patches, keep OS to date, manage software/applications
  • NeCTAR provides Infrastructure as a Service
  • Requires skilled System Administrators to manage IaaS
  • Can be expensive
    • Forgot to shut down things during prototyping = still paying for idle things
    • AWS has no billing warnings, just spends till you run out of money so be careful
  • Replication makes scaling easy but can also lead to replication of vulnerabilities
    • Old images not being updated
    • Poorly configured server instance replicated 100's of times
    • Incorrect security settings lead to massive data lost
  • Amazon Simple Cloud Storage Service (S3)
    • Really easy to use and get terabytes of data
    • Nothing inherently insecure about it, just really easy to setup incorrectly
  • In 2013 Rapid7 survey 12000 s3 buckets and found that 1 in 6 were left open to the public
    • Accessible via a guessable URL: s3.amazonaws.com/[bucket_name]/
    • This is getting worse lol

Some idiots getting their data breach'd

  • WWE - 3+ million member details
  • Verizon - 14 million records of subsciber calls, including name and account PIN
  • Republican National Committee - 198 million voter details (o shit that's like everyone in the US)
    • Had modelled religion/ethnicity lol
  • Dow Jones & Company - 2.2+ million customer records
  • Election Systems & Software - 1.8 million voter records
  • Power Quality Engineering (PQE) - lost sensitive electrical infrastructure data
    • Found top secret government shit
  • Here's some more
  • Holy shit guys actually stop please

AWS EC2

  • By default AWS instances will have some degree of insecurity
    • Gobally accessible
    • Password based access (as in your AWS account not the actual instance)
      • Holy shit your credit card stuff
    • Publicly accessible metadata available at /latest/meta-data/
  • Threat stack analysed their own customers and found 73% of companies had critical AWS misconfigurations
    • This is really hard to fix since most people offshored all their sys admins and they have 100s of instances

AWS Credentials

  • AWS credential compromise is a growing problem
  • Never hardcode your AWS credentials or store them in a git repository
  • DXC lost their credentials via a developer publishing them to a public repo by mistake
    • Started 244 AWS instances and had a 64k USD
  • Uber had a private repo compromised that contained AWS credentials giving the attackers access to customer data held in AWS (50m customers and 7m drivers)
    • Uber secretly paid 100k usd to the attackers to delete the data

Platform as a Service

  • People thought this would be better since it's not as low level
  • Provides a framework for rapid development/deployment of applications
    • Potentially across multiple platforms
  • Select frameworks to add to application, then develop the application
    • No managing the installation or updating or the frameworks
    • Often uses Git to deploy applications
    • Generally provides auto-scaling
  • Salesforce Heroku, AWS Elastic Beanstalk, Microsoft Azure, RedHat OpenShift, Google App Engine
  • Not grown as quickly as expected
  • In many ways provided a middle ground between IaaS and SaaS but somewhat blighted by history
    • Vendor tie-in
      • Moving between different Platforms can be difficult and expensive (like Google App Engine requiring certain things)
    • Scaling costs grow rapidly
      • Good for proof of concept but wouldn't want to run a large business on PaaS
      • Infrastructure cost is like the half
    • Runtime limitations - limited support and often restricted features
    • Service providers have shutdown, abandoned platforms, forcing customers to migrate/update their applications using a different infrastructure

IaaS vs PaaS vs SaaS

  • Probably still need on-premises for academic/really high powered shit

Severless - Function as a Service

  • Next step from PaaS - developer takes no interest in infrastructure or software stack, just the functionality of the application
    • It isn't actually serverless - there are still servers, it is just they are invisible to the developer
    • The scaling works really well
    • Costs are tiny
  • Hook.io and AWS Lambda introduced in 2014
    • Google Cloud Function (Firebase as well), Azure Functions
    • Aligns with microservice development technique
      • Build services to do tiny things
      • Call these services for your app
      • Some people don't like this cause it splits stuff up
  • Development is simplified, no low level handling of requests, functions are small pieces of code expected to run in milliseconds
  • Static content served via a CDN (Content Delivery Network)
    • CDN is good at delivering static stuff
    • Moves all your dynamic stuff to the cloud network

Severless - Example - Alexa

  • Allows easy access to advanced APIs and functionality
  • Including Amazon Alexa
  • Can handle really simple voice based queries

Severless - The Future

  • Severless adoption grew by 667% in the fourth quarter of 2017
  • Allows companies to focus on what differentiates them, which is their product/application and not on infrastructure which is largely commoditised anyway

What does this mean for SysAdmin?

  • The role isn't going to disappear

  • Skillset is changing

    • Configuration vs Installation

  • Severless may well dominate in the application dev space but it it's suited to long running/high performance tasks

  • Could see a shift away from PaaS to Serverless but IaaS will always be around

  • Awareness of best practise, network config, security and privacy will be paramount

  • Role is gradually moving upwards, away from low-level OS and network management to strategic planning and anomaly management

    • Understanding bottlenecks in the network
    • Understanding security and privacy
      • Avoid losing all your shit on AWS S3
    • We've been convinced cloud is secure due to marketing
    • 198 million records on a webserver is the same as 198 million records on "the cloud"

  • A lot of what we have covered has been about understanding the infrastructure in order to protect it and yourselves from attack

  • Gotta change how systems are being compromised and where data is being stolen

  • We're getting really shit at this holy crap

  • Dystopian cyber future where hackers rule the world let's go

  • The scale is almost becoming unmanageable

    • This is hitting your server all the freakin' time

  • Holy crap we're all idiots