Cloud computing Stelios Sotiriadis Euripides G M Petrakis

Cloud computing Stelios Sotiriadis, Euripides G. M. Petrakis 11/06/2021 1

Διαίρει και βασίλευε! “Work” w 1 w 2 w 3 “worker” r 1 r 2 r 3 “Result” 11/06/2021 Partition Combine 3

Κατανεμημένα συστήματα Εφαρμογές • GPS • Αυτοματοποιημένα τραπεζικά συστήματα (visa) • Air-traffic • Roaming cellular telephones • Social Networking (Facebook)! • P 2 P (Torrents ; ) • WWW? 11/06/2021 4

Ιστορική εξέλιξη… 2007 2014 ? ? ? 2000 Cloud 1999 Utility 1999 1997 HPC -HTC 1999 Volunteer Computing Grid Parallel Computing 1987 Metacomputing 1960 s Supercomputers 11/06/2021 6

Δρομολόγηση & meta-Δρομολόγητές (Scheduling) FCFS JOB A RESOURCE A QUEUE FCFS JOB B Scheduling problem: Μία εργασία JOB A εκτελείτε σε χ ms) και έχει δρομολογηθεί στον υπολογιστικό πόρο A με ουρά εκτέλεσης FCFS RESOURCE A JOB B JOB C RESOURCE B RESOURCE C QUEUE FCFS QUEUE BACKFILLING QUEUE SJF JOB D RESOURCE D JOB E RESOURCE E QUEUE E RESOURCE … QUEUE … JOB … 11/06/2021 Meta–scheduler 7

Why – When ? • Cloud computing is one of the most prominent trends in IT sector • Emerged from Grid computing • New challenges arose when CERN started experiments in the Large Hadron Collider • Demands for high-performance computing (HPC) • Later, demands for HPC independent of location • And later, profitability for other institutions became obvious as it was not necessary to buy – host – maintain large computing resources 11/06/2021 8

πηγή δεδομένων 11/06/2021 CMS CERN Source: http: //en. wikipedia. org/wiki/Compact_Muon_Solenoid 9

CERN Datacenters 11/06/2021 10

Παράδειγμα datacentre • 136. 8 Teraflop/s on LINPACK (64 K processors) • Interconnects all compute nodes • 1. 4 Gb/s on all 12 node links (2. 1 GB/s per node) 1 µs latency between nearest neighbors 5 µs to the farthest 11/06/2021 IBM Blue-Gene/P 11 Source: http: //en. wikipedia. org/wiki/Blue_Gene

Cluster computing • Due to availability of internet organizations where able to share or aggregate geographically distributed resources – Cluster systems – Storage facilities – Computing power ? • Decentralized control differentiates from Grid 11/06/2021 12

Cloud Computing • • Grows from existing technologies (Grid, Web) Dynamic – decentralized control of resources Available via Web interfaces Monitoring of resources Dynamic resource allocation – provisioning Virtualization !! Application are available as services or can be assembled using fragments of software (services) – This App development is different from Grid where users generate an executable and is transferred to Grid for execution – Dedicated software for this • Cloud is characterized by lightweight accessibility and ease of use • Switching between providers is not typically supported by cloud providers 11/06/2021 13

Service Level Agreement (SLA) • Cloud providers sign SLA with customers – Services availability above 95% of the time of operation. – Constant monitoring of local node resources usage, fraudulent access detection – Help desk and Level 2 support Monday to Friday 9 am-5 pm CET – Ticket response before EOB of the following work day for 95% of requests – Ticket resolution time within 2 work days for 95% of requests – Scheduled and Unscheduled maintenance and incident handling 11/06/2021 14

SLA Metrics • • • • • CPU capacity CPU speed for Virtual Machines (VMs) Memory size Cash memory size for VM Boot time Time for VM to be ready for use Storage size of data Scale up Max of VMs for one user Scale down Min number of VMs for one user Scale up time Time to increase number of VMs Scale down time Time to decrease number of VMs Availability Uptime of service in specific time 11/06/2021 15

Grid vs. Cloud computing • προβλήματα μεγάλης κλίμακας δεν μπορούν να λυθούν με έναν υπολογιστή… 11/06/2021 http: //markusklems. wordpress. com/2008/06/19/cloud-vs-grid/ 16

Χαρακτηριστικά Cloud Computing Κοινά με άλλα συστήματα: Massive Scale Elastic Computing Homogeneity Geographic Distribution Virtualization Service Orientation Low Cost Software Advanced Security Oυσιώδη: On Demand Self-Service 11/06/2021 Broad Network Access Rapid Elasticity Resource Pooling Measured Service 21

Cloud Service Models Software as a Service (Saa. S) Platform as a Service (Paa. S) Infrastructure as a Service (Iaa. S) Sales. Force CRM IBM Smart. Cloud Google App Engine Adopted from: Effectively and Securely Using the Cloud Computing Paradigm by peter Mell, Tim Grance 11/06/2021 22

Cloud Service Models • Public Clouds – Υπηρεσίες μέσω του Internet • Private Clouds – Ιδιωτικά δίκτυα • Hybrid Clouds – ποικιλία από δημόσια και ιδιωτικά cloud με πολλαπλούς παρόχους 11/06/2021 25 Source: http: //www. dummies. com/how-to/content/comparing-public-private-and-hybrid-cloud-computin. html

11/06/2021 27 http: //pleasediscuss. com/andimann/wp-content/uploads/2011/03/NIST-Cloud-Ref-Architecture. jpg

Διάφορα επίπεδα cloud services Application Service (Saa. S) MS Live/Exchange. Labs, IBM, Google Apps; Salesforce. com Quicken Online, Zoho, Cisco Application Platform Google App Engine, Mosso, Force. com, Engine Yard, Facebook, Heroku, AWS Server Platform 3 Tera, EC 2, Slice. Host, Go. Grid, Right. Scale, Linode Storage Platform Amazon S 3, Dell, Apple, . . . 11/06/2021 28

Cloud Computing Service Layers Services Application Focused Application Development Platform Infrastructure Focused Storage Hosting 11/06/2021 Description Services – Complete business services such as Pay. Pal, Open. ID, OAuth, Google Maps, Alexa Application – Cloud based software that eliminates the need for local installation such as Google Apps, Microsoft Online Development – Software development platforms used to build custom cloud based applications (PAAS & SAAS) such as Sales. Force Platform – Cloud based platforms, typically provided using virtualization, such as Amazon ECC, Sun Grid Storage – Data storage or cloud based NAS such as CTERA, i. Disk, Cloud. NAS Hosting – Physical data centers such as those run by IBM, HP, Navi. Site, etc. 29

Virtual Machines • Eπιτρέπει πολλαπλούς εικονικους Η/Υ να τρέξουν σε μια ενιαία φυσική μηχανή. App App App Guest OS (Linux) Guest OS (Fedora) Guest OS (Windows) VM VM VM Virtual Machine Monitor (VMM) / Hypervisor Xen VMWare etc. Hardware 11/06/2021 31

Cloud taxonomies 11/06/2021 33

Amazon EC 2 • Amazon EC 2 περίπλοκο web service. • EC 2 παρέχει ένα API για instantiating computing instances χρησιμοποιώντας διθέσιμα λειτουργικά συστήματα. • Χρησιμοποιεί Amazon Machine Images (AMIs). • Χαρακτηριστικά: S 3, Cloud Management Console, Map. Reduce Cloud, Amazon Machine Image (AMI) • Πολύ καλή κατανομή, load balancing, cloud monitoring tools 11/06/2021 34

Open Cloud Computing Interface • Standard API για διαχείρηση εργασιών σε clouds • OCCI is a Protocol and API for all kinds of Management tasks. 11/06/2021 πηγή: http: //occi-wg. org/about/ 37

Open. Stack • Σύνολο απο services βασιζόμενα σε OCCI standards – για deployment, autonomic scaling and monitoring – Iaa. S cloud frameworks – open source APIs για public και private Iaa. S clouds • APIs: Nova, Glance, Keystone, Quantum • Iaa. S Cloud Services επιτρέπουν τους χρήστες να διαχειριστούν: – VMs, Virtual networks, storage resources 11/06/2021 38

Cloud impact 11/06/2021 40

βιβλιογραφία • Cloud Computing and Grid Computing 360 -Degree Compared – http: //arxiv. org/pdf/0901. 0131. pdf • What is the Grid? A Three Point Checklist – http: //dlib. cs. odu. edu/What. Is. The. Grid. pdf • Above the Clouds: A Berkeley View of Cloud Computing – http: //www. eecs. berkeley. edu/Pubs/Tech. Rpts/2009/EECS 2009 -28. pdf • UC Berkeley CS 10 Fall 2010 Lecture 20, Cloud Computing – http: //www. youtube. com/watch? v=Mro. Ulbi. Ki 0 U 11/06/2021 41

Demo Openstack Current Cloud Services Generic Enablers • Iaa. S • Paa. S – – CPU Memory HD Network & Internet • Virtual Topologies – OS • Images 11/06/2021 – Iaa. S + GE (Development Platform) • Saa. S – Iaa. S + Paa. S – A Future Internet Application 42