Intro - why cloud? internet always depicted as a cloud in pictorial representation formation of clouds and breaking up
How? - realized that h/w is underutilized
Change from traditional model Scalability a problem in traditional computing that consisted of a H/W (processor, ram, hard disk), OS and S/W Hardware costs Accessibility Upgrades Business continuity and disaster management
Distributed Model in a signle computer with multi core processor, multiple ram and multiple hard drive - it is a distributed model Client Server Model (communication over n/w using IP address and messages to communicate) Virtualization - multiple partitions in your compute
- gmail server (multiple processor virtualized to look like one)
Case study- online money transfering using web application- net bankingfrom ICICI to SBI Client machine request - ICICI server once done ICICI server act as client and talk to SBI server Client and srver and interchangable
grid is a network hence grid computing is n/w computing is a form of ditributrd computing distribute work to different computers in the n/w client server? yes but server sends the request to clients (distribute work), client process and sends respose back. say 50 formulas to be calculated like in case of weather forecast, server distrubute the 50 formulas to different clients in the n/w. the server then aggregates the responses from the clients and orders it to make sense. In short server does the resource management
- clients use their own hardware and other resources
-server needs to know how fast each client is, connection strength, type of commncn etc to distribute wisely
-server doesn't care how it is processed only the response - server gathers and orders responses
Demo on BOINC - berkeley open Infrastructure network computing SETI - search for extra terrestrial intelligence - listen for radi signals, and look for intelligent waves not a proffesional group, intrest driven team. They cant have their resource so they use BOINC. SETI at Personal computer act as the server and sends request, the priority is always less.
MilkyWay, similar app trying to map whole of milky wayalso needs high computational power. takes images and image processing. Cient side comp run diagnostic and send report to server. server send request to client, client process (100% perfomance by client machine in task manager), but doesn't interfeare or effect the work in client machine, coz priority is low. can manage the time of running this applications.
Cloud Concepts renting instead of buying - pay only for what u use. can be used to connect your computer (thin client) to a computer with better and stronger configuration (thick clients) through a browser.
diff b/w cloud and grid
transparency cloud - opaque to resources, cannot figure where what resource is coming from grid - resources are exposed
federation cloud - provider make their on architecture, federation provided only through client interface grid - middleware approach (middleware is that s/w that makes it look like a single h/w)
case study - forbes.com stock market trade - transaction (buy and sell) only from 9 am to 3.30 pm on weeke days. hesvy load during this period. hence they don't have to buy heavy duty servers, and spend on maintanance. solution can be to keep few servers with them and hire the rest on cloud. Advantages:
1.pay only for wht u use 2.scalabiliy, with few clicks they can commission or decommission more servers, as and when transactions increase or decrease. To set up a server IT team can take months, on cloud its almost instant. 3.Forbes job is not to provide It services, but to do stock transactions, amazon will provide that. relying on amazon for IT service, forbes can not worry about IT services and concentrate more on their job. 4.Rent jusy .10 cent for an hr
5.cn create private one for a company, more security.
How is it possible to do for 10 cents? The providers have identified unused space in their webhosting servers. so they use only the servers they already had but unused. 10 cents is thus extra money
Android Phones - contacts saved, sync with gmail, all ur contacts in gmail, even if u loose ur contacts, u can have it in your gmail, next phone, link it with same gmail account - u have all ur contacts
Setup and demo
2 approaches of setting up client Cloud provider view point Blue gene approach - using main frame - provided by IBM clustering approach- google
Client Perspective Infosys - needs entire desktop for an employee, use VPN or RSA connection rediff - doesn't need desk top only servers social n/w (orkut/facebook) - doesn't need desk top only servers
All the above users can come in and can leave, by hiring from cloud, the problem of wasting machine when user reduce is not there.
Cloud Offerings Infrastructure (IaaS) - servers (Rackspace, Joyent) Platform (PaaS) - azure (microsoft), SDK, force.com Application (SaaS) - program written on SDKs, google docs, sales force, zoho - accounting collaboration - cutomer relation ship management - Enterprise Resource Planning - Human Resource Management - Content Management etc
will not have to install on all user PC, cost of individual licence
IaaS Deployment Models Public Cloud Community Cloud Hybrid Cloud Inter cloud - network of networks Cloud Storage -
Disadvantages Privacy/security - data is away from you- control lost? its possible for your server can get hacked. cloud provider can access it.
Availability and perfomance internet connection down
sustainability and siting cannot cope up with the maintanance
business of thin client manufactures can reduce? can gain in thick client production. Moving from laptop to tablets. even with less config, they can connect to cloud to leverage its power. enterprise has the required internet connectivity, but individuals don't. indutry is more on increasing the net speed. individuals using cloud in a way they don't realize
Compliance - standardization? breaks down softwares for interface, Eucalyptus, nymbus etc has s/w to make their architecture in line with amazon
Internet Ecosystem HCL - manage Carbon - software used to analyse and manage its carbon emissions. migrated to windows azure, using cloud this application is offered on - premise maintanance cost is reduced, and cost of entry, and cost of ownership deployment to 1/4 of the time
Day 2 EyeOS - Cloud based OS install and modify apache server extension=php_curl.dll extension=php_shmop.dll memory_limit = 128 to memory_limit = 256 restsrt apache in xammp
To run office as daemon (daemon - small s/w that run to listen signals) asking it to accept connection and listen to port 8100
soffice.exe -headless -noloop - orestore accept=socket,host=localhost,port=8100;urp;StartOffice.ServiceManager
Port - s/w and services listening to different port default to them; eg apache 8080 Kaazing WebSocket Gateway, the world’s first high-performance Web communication platform
http://localhost/eyeos/index.php http://192.168.6.5/eyeos/index.php --> to logn to server give the IP address or the computer name intead of localhost, if on the n/w
exif tool for cloud indexing metadata has to be indexed for easy search eyeOS - Destop as a service (all three Iaas, Paas, SaaS together)
Virtualization For running multiple OS
Why? - better utilization, reduce e- waste
·1-1 mapping b/w h/w and os ·underutilized h/w - 140 billon dollar in excess server capacity today ·conflicts b/w apps ·unsustainable power, cooling req - 29 billion $ in power and cooling industry wide, 20:1 admin ratio for maintanace
instead of OS controlling H/w , Virtualiztion layer controls eg: KVM, XEN, (no OS) VM ware
multiple projects machine can run in isolation in different virtual machines on same physic direct access to h/w - gives better perfomance allocate resource as and when required