CLoud computing, virtualization, scalability, media computing, course

Sistemi Distribuiti
Corso di Laurea in Ingegneria
Prof. Paolo Nesi
Parte 17: Cloud Computing &Virtualization
Department of Systems and Informatics
University of Florence
Via S. Marta 3, 50139, Firenze, Italy
tel: +39-055-4796523, fax: +39-055-4796363
Lab: DISIT, Sistemi Distribuiti e Tecnologie Internet
nesi@dsi.unifi.it, paolo.nesi@unifi.it
www: http://www.dsi.unifi.it/~nesi

Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2011-2012 1

Structure
G elementi di cloud Computing
 Motivazioni
 Definizioni
G La Virtualizzazione
G Cloud Computing
G High Availability
G vSphere Infrastructure
G Security on the Cloud
G Conversions among VM and physical machines
G vCenter, datacenters and cluster management
G Comparison among virtual computing solutions
G How to work with Virtual Machines
G IaaS solutions
G SaaS Solutions
G PaaS Solutions

Motivations for
Cloud computing and Virtualization
G Physical systems are
 not flexible:
The HW/SW has to be dimensioned for the worst case and not
for the real/typical one
It can be hardly reused for different purposes
 subjected to HW failure:
Software has to be installed again when moving hardware
OS has to be reinstalled in most cases
 Subjected to high costs of setup:
Power, conditioning, network connection, etc.
 non scalable, to scale frequently implies hardware restructuring
duplication of data, distribution of activities among HW,
Balancing or workload, sharing databases, etc.


Final Motivations
G Reduction of costs for HW and operating system SW
maintenance.
 High costs for high availability
 High costs for high reliability
 High costs to follow the HW trends
G Sharing resources
G Needs of High flexibility in terms of features
 The sofware is becoming a services and not anymore a
product.


Motivations for
Applications on the Cloud
G Single tier applications such as small web portals do not
demand a cloud
G Multitier application servers
 High performance web server, high number of users
 SN: Social network
 CMS: content management systems
 CRM: customer relationship management
 CDN: Content delivering network
 P2P torrent tracker: see Piratebay
G …
G All typically parallel applications that may run on grid
G ….


Datacenter, definition
G Datacenter
 A computer factory/farm in which servers/computers are
hosted and organized: power, net maintenance, etc.
In most cases industrial computers, blades
 They can be exploited for private purposes, grid, cloud
computing, renting/hosting, etc.


Clusters, definition
G Cluster is typically intended:
 Several kinds of computers and/or VMs may be used to compose
a cluster, sharing the same domain or not.
 a set of computers/VMs into a datacenter which are dedicated to a
unique problem, for example:
A microgrid……………
A social network……….
A web portal with its multi-tier servers: front-end portals,
balancer, database computers, backofficer microgrid nodes,
etc.


Infrastructure, definition
G A set of datacenter and clusters
G A set of
 NAS: Network Area Storage
 SAN: Storage Area Network
G Etc..


Typical Features for renting a Computer
or a Virtual Machine into a Datacenter
G Requesting to have one or more Computers/Hosts and/or VMs
G Hardware:
 CPU: 32/64 bit, number of cores/CPU, frequency of work, intel/amd, etc.
 RAM memory: size and frequency of work
 Power supply: fault-tolerant or not; with UPS or not, etc..
 HD: space, speed (7.2, 15Kgiri), security level/RAID, type SAS/SATA, SCSI
 Network features:
Number of connections/cards, number of IP addresses, static/DHCP
Transfer rate: minimum guaranteed, maximum possible, down/upload
Maximum transferred bytes: per day, per month, etc.
 NAS/SAN, Network area storage/SAN, fiber/internet: size, RAID, etc.
G Software:
 Operating systems
 Software preinstalled into the Computer/VM, see in the following
G Services:
 Periodic back up: details on HD space
 Access to VM/Computer: remote desktop or KVM tool
 Reboot or not of the Server, for example via Plesk.

Hosting, definition
G HPC: high performance computing
 Solution based on cloud/grid for parallel execution of algorithms
and tools.
G Hosting web portal into a datacenter/cluster
 Renting a web space via some SLA (service level agreement),
contract, monthly rate to publishing web pages: service httpd
 Additional services: mysql, php, asp, ftp, ssh, https, etc..
 Features:
Space on disk, networking
Domain space, etc.
G Hosting a machine (computer/VM) into a datacenter
 According to some SLA, contract…
 Renting a Computer/VM/Cluster into a data center


From Clustering VM to Cloud
G The first step has been the exploitation of unused resources
to provide them as a service to third party.


Motivations for
Cloud computing and Virtualization
G Most of the datacenters capabilities are not exploited at 100%
in every time instant
 They are typically present in large industries/institutions or in large
services: google, amazon, tiscali, dada, ibm, cnr, etc
 Their size is typically defined on the basis of the worst case
G If they are big: the exploitation of the remaining resources for
cloud/hosting is a solution to recover money, since a non
working machine (like a caw) is a costs without any return.
G If they are small: it could be a solution to host machines on a
professional infrastructure to reduce annual costs for HW/SW
 Delegating to cluster owners the costs for maintenance, renovating
hardware, renovating software, network costs, back up costs,
power supply, etc.


Structure
 emulation, para-virtualization
 virtual resources
 snapshots
G Cloud Computing
G High Availability
G IaaS solutions, SaaS Solutions, PaaS Solutions


Virtual Machine, Virtualization
G Virtual Machine:
 An image of an operating system that can be put in execution into a real
host/computer creating a virtual computer that exploits a part of all of the
host resources
G E.g.: Host may be Linux-like while the VM may be Window, Mac, Linux, …

G Virtualization
 Transforming a physical computer into a VM, virtual machine, hosted on
some Host computer
G Hypervisor (VM Monitor) to manage the several VMs on the
host
Computer 1
Virtual Virtual
Machine Machine

Hypervisor


Benefits of Virtual Machine
G Main benefits:
 Separation of OS/SW with respect to the needed HW
 Exploiting legacy solutions which can be wrapped into a VM and
protected with a physical firewall without reinstalling and recompiling
old applications.
G For example:
 An old Linux Server hosting several web portals with a old versions
of: MySQL, PHP, etc.. and many configuration aspects: users,
mailing lists, etc. very time consuming to port on a new server
 An old Cobol application running only on an old Windows 2000
Server, which cannot be recompiled into a new Windows Server
2008 at 64 bits without spending months of work.
 A Cobol application running on machine based on an IBM system
36….
 Etc.


VM on PC workstations
G On a single Computer it is possible to put in execution a VM by
using a standalone Hypervisor
G VMware Workstation can play on Windows:
 VMware Virtual Machines with: MS Win, Linux, Mac, ChromeOS,
etc.
G VMware Player can play on Windows or Linux
 Free of charge
 VMware Virtual Machines with: MS Win, Linux, Mac, ChromeOS,
etc
G VMware Fusion can play on MAC:
 VMware Virtual Machines with: MS Win, Win 7, ChromeOS, etc.
G Microsoft Virtual PC on Windows 7:
 Free of charge
 Create VM with Win XP, Linux Ubuntu,


Cloud Computing with VMs
Several Hypervisors on a Clusters
internet

Computer 1 Computer 2
……………

Virtual Virtual Virtual Virtual
Machine Machine Machine Machine

Hypervisor Hypervisor

Virtual cluster manager
Virtualized Infrastructure --- control servers


Software/Components preinstalled
into rented Computers/VMs
G Several kinds of software components/tools that can be
accessible on rented VM and/or Computers.
G Availability of HW/SW: for example at 98% or 99.999%
G Typical components that could be requested:
 DB: MySQL
 FTP: server and client
 Web Server: Apache, IIS
Add-on: PHP, Perl, Python, cache tools on several levels
 SMTP address, antispam
 Antivirus
 Web Application Server: TomCat, ….
G A full server can be customized, so that any other tool
can be installed as well from the user


Approaches for Virtualization
G Hardware Emulation
 Fully emulation of Hardware devices and features
 It is possible to use an original Operating system without changes, may be
with some drivers installed, but not kernel changes
 Higher isolation among VMs, strong robustness, limited efficiency.
 Used by VMware
 Typically 10% of overhead
G Paravirtualization (total or only on some devices)
 The execution is performed via specific API and the hosted Operating system
has to be modified to use them instead of the original HW
 Lower isolation, higher efficiency,
 Lower robustness: VM crashes may crash the whole system
 Used by: HP-VM, Xen (both of them which can also go in emulation mode)
 Used by VMware: VMXNET (100 Gbps net), PVSCSI in vSphere4
 Typically 2% of overhead


Para-virtualization
G It is intended as a solution to cut out a part of the I/O stack,
for example for the HD access or network.
G The following example is related to HP VM v. 3.5


Implementing Host Profiles
G Host Profile
 Memory Reservation
 Storage
 Networking
 Date and Time
 Firewall
 Security
 Services
 Users and User
Groups
 Security

I Reference Host I Cluster

D1


“Speeds and Feeds”
Optimization for the Highest Consolidation Ratios

Virtual Machines VM Scale Up 8-way vSMP and 255 GB of
Virtual hardware scale out RAM per VM
APP APP
OSAPP OSAPP APP
OS OS OS
64 cores and 512GB of
ESX Hardware Scale Up physical RAM

Hardware Assist
Lowest CPU overhead
Purpose Built Scheduler

CPU Hardware Assist
Maximum memory efficiency
Page Sharing

Memory Ballooning

VMXNET3 Wirespeed network access
VMDirectPath I/O
Networking
Greater than 360k iops per second
Storage stack optimization Lower than 20 microsecond latency
VMDirectPath I/O
Storage
Current NEW


ESX 4.0 Performance with
SQL Server 2008
Relative Scaling Ratio G ESX achieves 90%
VM 147.24
of native performance on
Native 133.12 4.0 vCPU VM

94.04 G Workload transaction
79.88
latency unchanged
51.08
between ESX 4.0
45.22 and Native

1 vCPU 2 vCPU 4 vCPU


Virtual Resources, 1/2
G The idea of Virtual Resources (CPU, Mem, HD, net.) consists in
providing a number of resources larger than those that physically
available and manage them virtually and/or dynamically
G For example, one Host may have 2 VM and HW resources:
 2 cores at 1300 Mhz
VM1:
 1.8 Gbyte RAM
I 1 CPU 1300 Mhz, 400 Mhz reserved
 2 network cards
I 1 Gbyte RAM max, 400 Mbyte
 Best case: reserved
CPU=400+800 Mhz I 2 network cards, 2 IPs

RAM=400+400 Mbyte
2 Network cards shared VM2:
 Worst case  no resources enough:
CPU=3 cores at 1300 Mhz reserved
RAM=2 Gbyte I 2 network cards, 2 IPs
2 Network cards shared


Virtual Resources, 2/2
G The server hosting the VMs VM1:
 Min Mhz:
400Mhz + 800Mhz reserved
 Max Mhz: I 2 network cards, 2 IPs

1300Mhz + 2*1300Mhz
VM2:
 Min RAM: I 2 CPU 1300 Mhz, 800 Mhz reserved
400Mbyte + 400Mbye I 1 Gbyte RAM max, 400 Mbyte

 Max RAM: reserved
I 2 network cards, 2 IPs
1000Mbyte + 1000Mbye
 Network:
No limits on the number of
virtual IP addresses/cards


Limiting VM Resources
G VM Resources (CPU, Mem, HD, net.) consists also in providing
support for:
 Dinamically providing resources over the reserved values that can
be negotiation into the SLA/ contract.

G Controlling and limiting access and the exploitation of HW
resources:
 A limit on the number of CPUs
A limit on the number of Clocks, over of a reserved number of
clocks
 A limit on the maximum size of the RAM, over of the reserved
number of Mbytes
 A limit on the size of the HD, SAN/NAS access
 A limit on the number of network cards, number of Mbps, etc.


Virtual Machine: Snapshots
G Working on VM Snapshots
 Creating a Snapshot:
A point from which it is possible to reboot, restart
Consuming HD space
Making back up since the core image of the VM is not
changed, changes are confined in the files representing the
last status “you are here” and not the previous conditions
 Restarting from a past snapshot
Losing current point: “you are here”, to avoid this do
another snapshot!!
 Deleting a past snapshot
Recovering HD space, removing a past restarting point
 Removing all snapshots
G Defragmenting images of the HDs into the VM


VM Snapshots
G VM Snapshots can be at VM Off or ON
 Snapshots of running VM have implications…
G Removing Snapshots
 Defragmenting images of the HDs into the VM
 Consolidating the changes in the previous version

G VMware WS has an automated Snapshot model to plan the
periodic snapshotting of the VM, for example:
 every hour, day, week, …

G A way to make back up
 A different way can be to clone the VM on different host or
NAS. In most cases, the cloning implies the lost of performed
snapshots


START Virtual Machine: Snapshots, 1/3

YOU are
here



YOU are
here


Structure
G Cloud Computing
 cloud vs grid
 goals of cloud computing
 soluzioni as a Service
G High Availability


Cloud Computing
G A Cloud: a type of parallel and distributed system consisting
of a collection of interconnected and virtualized computers,
VMs
 They are dynamically provisioned and presented as one or more unified
computing resources
 based on service-level agreements, SLA, established through
negotiation between the service provider and consumers
G Subset of grid computing where the allocated process are
virtual computers
G Implies
 An alternative way to have local servers or GRIDs
 Outsourcing: HW and SW tools, they may be grid elements
 Outsourcing: network, CPU, memory, HD, etc.
 Definition of some service agreement, monthly rate, minimum network
capability, kind of HW, mem space, minimum level of CPU/mem, etc.


Abstraction of Cloud Computing, (Buyya & Yeo)


Cloud vs GRID computing
G Cloud computing is an evolution of GRID computing
G Renting a GRID service means:
 Parallelize the algorithm, demand the execution, wait for the results, etc.
 To do no know where the processes are executed. In most cases batch
processing such as on Globus
G Renting a VM/Computer into a Cloud means:
 Simple contracts, remote access to servers
Get access to virtual or physical resources at 100%
 Privacy of the allocated processes in your VM/Host
 Processes running on your preferred OS and do not have to be recompiled
as in most Grid solutions.
 Scalability in terms of number of CPU, Mem, network performance,
computers, etc. OR you have to change your OWN architecture
Simple creation and reconfiguration of multitier solutions
• Creating fault tolerant solutions
• Balancing load of CPU, Memory, network, etc.


Cloud vs SuperComputers
G Cloud Computing is based on a set of Hosts for hosting VM or
providing access to single/multiple Servers fully at disposal of
the customers accessing to the cloud via remote access (see
later).
 They are MIMD computers
 Hosts may contains hypervisors to host VM of diff. OSs.
 Hosts may be single computers with Linux, Windows,…
G Super Computers, such as Blue Gene/P:
 Processor PowerPC 450 with 4 cores, 850 Mhz
 Each board: 32 CPU processors
 Each Rack: 32 boards
 Jülich Research Centre in Germania has a Blue
Gene with 65536 processors  167 teraFlops,
the strongest computer in the world!


Comparing Computing Services


Comparing Computing Services
G In the previous table, different solution for computing
services in the network are compared.

G They are mainly: cloud computing, grid services,
application services,

G The solutions taken have been selected as
representative of their category
 Please see the slides on GRID for more details and a wider
comparison on the grid solutions
 Please see the slides regarding the general distributed
systems for multitier applications


Advantages of Cloud Coumputing
G Grids are difficult to use and to maintain:
 GRID customers have too many different needs that make the
creation of fully open grid very difficult.
G Cloud computing
 HW/SW is hosting virtual computers that can be moved to other
solutions with low costs
 Lower costs since the HW/SW is seen as a service
No maintenance, centralized services such as back up,
scaling, etc.
 Lower costs for Small Business:
Reduction of costs since the admortment can be performed by
who is exploiting the computer
 Scalability similar to grid:
Horizontal scaling
Parallelization as MIMD


Goals of Cloud Computing
G Scalability.
 scaling with workload demands so that performance and compliance
with service levels remain on target
G Availability.
 users of Internet applications expect them to be up and running every
minute of every day, i.e.: h24, 24/7
G Reliability
 physical system components rarely fail, but it happen. So that, they
can be replaced without disruption.
 Today, reliability means that applications do not fail and most
importantly they do not lose data, and the service is not stopped.
G Security.
 Applications need to provide access only to authorized, authenticated
users, that need to be able to trust that their data is secure.


Goals of Cloud Computing
G Flexibility and agility:
 Adapt rapidly to changes of business conditions by increasing
the velocity at which applications are delivered into customer
hands. E.g.: more CPU, more clock, more memory, more
network cards, etc.
G Serviceability:
 In the past this meant using servers that could be repaired
without, or with minimal, downtime.
 Today it means that an application’s underlying infrastructure
components can be updated or even replaced without
disrupting its characteristics including availability and security.
G Efficiency:
 differentiates the cloud computing. The process allocation and
costs have to be very effective with respect to the investment.


Definizioni
G Classificazione NIST
 Software as a Service (SaaS)
 Platform as a Service (PaaS)
 Infrastructure as a Service
(IaaS)

G Business Process as a Service
(BPaaS)
 Aggiunto in seguito

G Everything as a Service (XaaS)


Infrastructure as a Service (IaaS)
G erogazione di servizi infrastrutturali relativi a capacità
elaborativa, storage, rete e altri elementi di base
assolutamente indipendenti da servizi applicativi di
qualunque tipo.
G Si utilizza quindi l’infrastruttura messa a disposizione dal
provider per eseguire la propria applicazione,
 pagamento in base al consumo dell’infrastruttura
 lasciando sotto la responsabilità dell’utente la gestione del
sistema operativo, dell’eventuale middleware e della parte
di runtime, oltre che dell’applicazione stessa.
G Amazon EC2 è un esempio di servizio IaaS.


Platform as a Service (PaaS)
G erogazione di servizi applicativi di base come sistemi
operativi, middleware, linguaggi, tecnologie di base dati e
l’ambiente runtime necessari per eseguire l’applicazione,
G L’applicazione rimane l’unica cosa sotto la responsabilità
dell’utente, oltre alla definizione del modello (e.g.,
numero e dimensione dei server, datacenter,
caratteristiche del networking) da utilizzare per
l’esecuzione dell’applicazione.
G Google AppEngine è un esempio di Platform as a
Service.
G A livello PaaS viene anche collocato l’insieme dei servizi
MaaS, Middleware as a Service.


Software as a Service (SaaS)
G erogazione di servizi applicativi di qualunque tipo, accessibili
indipendentemente dalla collocazione e dal tipo di device
utilizzato.
G Non è eseguita un’applicazione proprietaria del cliente, ma il
cliente stesso paga il diritto (mediante licenza o canone di affitto)
di utilizzo di un’applicazione messa a disposizione dal provider,
senza preoccuparsi di come essa venga realizzata e gestita nel
cloud.
G L’unica preoccupazione del cliente in questo caso, oltre
ovviamente alla scelta della corretta applicazione che soddisfi le
sue necessità, è gestire il numero di licenze richieste in funzione
del numero di utenti.
G SalesForce.com Customer Relationship Management (CRM) è
un esempio di soluzione in cui il software è venduto in modalità
as a service.

Business Process as a Service (BPaaS):
G erogazione di servizi non esclusivamente riferiti ad ambiti
applicativi ma direttamente alle funzionalità di business o
di processo, potenzialmente trasversali rispetto alle
piattaforme applicative.
G un processo di business mappato interamente nel cloud
(composto da servizi, applicazioni web, applicazioni
legacy, servizi di integrazione, etc.).
G Il processo di business è un pattern di servizi ed include
problemi di sicurezza, costi, scalabilità connessione fra
local e cloud bidirezionale, il cloud può essere un burst
per l’azienda, e può sgravare i costi nel momento del
bisogno.


Modello Generale


Definizioni
G Private Cloud. abilitata per operare soltanto per un’organizzazione.
può essere gestita dalla stessa organizzazione o da parte di terzi.
G Community Cloud. condivisa da più organizzazioni a supporto di una
singola community che ha interessi e obbiettivi comuni. Questa può
essere gestita dalle stesse organizzazioni o da terzi in modalità on-
premise e off-premise.
G Public Cloud. resa disponibile in maniera pubblica ed è di proprietà di
un organizzazione che vi gestisce la vendita di servizi cloud.
G Hybrid Cloud. Infrastruttura composizione di due o più cloud (siano
essi private, community o pubblici), rimangono entità separate, ma
comunque accomunate da standard o tecnologie proprietarie che
abilitano un certo livello di portabilità di dati e/o applicazioni di
migrazione e/o bursting.


Structure
G Cloud Computing
G High Availability
 Workload Balancing
 RAID on HD
 SAN/NAS


High Availability
G The high availability has to be guaranteed only
by the integration of features of:
 High Reliability
 High Serviceability
 Fault tolerance
 Migration of VM to different HW
 Disaster recovering


High Availability
G High Availability,
available 99.999 %
(called “Five Nines”)
percent of the time.

G Five Nines is the term
for saying a service or
system will be up almost
100 percent of the time.

G In case of failure:
 the path changes to
guarantee the service


High Availability
G May be achieved by redundancy of :
Servers and Services in hot spare or balancing
• VM based architectures
HW/SW: power supply, network connections, etc.

 Load Balancing/Balancer according to server traffic/requests
Server Clustering, multitier solution
 Hot Spare:
Server: cloned server to be used when the main is not
functioning: heartbeat to detect the server availability and thus
failover. (heartbeat signal to communicate the correct running
of a process/CPU)
HD: Raid based on SAN/NAS, LUN, in host
 Mixed: balancing and hot spare


High Availability: Load Balancing
G Service distributed/cloned on more servers
 According to different policies
Round robin
Network traffic
G Single transfer rate capability on the front-
end of the load balancer
G Sensing the availability of the server
on balancer +
 heartbeat solutions to
understand if the servers are
alive or not
G Common NAS/SAN

G That is a Cluster of servers on the cloud !


Balancing multi-tier, 3-tier


High Availability: Hot spare
G Fully cloned servers to be
alternatively used when the
running one fails
G Internal Network
 Heartbeat to detect the server
availability and thus failover.
 [To keep servers aligned on
context and data]
G Shared data storage is a
simplification and optional.
 A different solution may be to
have a cloned storage to keep
aligned among the two servers


High Availability: Hot spare, hw
G Three separate networks cards
 Front end
 Heartbeat
 Database NAS/SAN
G UPS/APC solutions with
 2 UPS, each of which with
network card
G NAS/SAN
 Raid 5 or 6, 60
 Fiber connection


Network and Virtual Networks

G The same VM with access
to 2 different network via
real network adapters

G A virtual network


Partitioning the database
G The database may be
 stored in a single NAS
 stored into a distributed solution by
partitioning the tables horizontally
or vertically

G Horizontal partitioning of a table:
 According to a certain primary key
 Distributing the table on different
servers


HD RAID

I RAID 0: HD0+HD1

I RAID 1: HD0 cloned on HD1

I RAID 5: gestione del guasto
di un disco
I valido fino a 14
I per la doppia parità
 RAID 6


An Example of RAID 50

G RAID 50: A1,A2,A3,A4,A5,A6 and three Parities
 Min 9HD, can support failures of three HDs
 improves upon the performance of RAID 5
G In this case:
 Size: 6 over 9 HD, you lose a HD per R5
 Read/Write speed up: n(m − 1)XHD
 n=3, m=3  6x in this case


SAN vs NAS
G In most cases they are wrongly considered the same staff
 It is true that: NAS/SAN typically are HD in Raid connected/shared
to/by Servers
G NAS: Network Area Storage
 Network HD sharing content at level of File via HTTP, NFS, CIFS,
etc., protocols
 Multiple servers may access to them
 Reduced performances
G SAN: Storage Area Network
 HD system segmented in LUN, mounted by the Server/VM
operating system at level of disk block
 The Operating System has to format with it own file system
 Possible protocols: low level IP, iSCSI, Fiber, etc.


Structure
G Cloud Computing
G High Availability
 Vmotion
 Power Management
 Resource Scheduling
 Fault Tolerance


vSphere 4 infrastructure of VMware
G High level features:
 HA: high availability
 DRS: Distributed Resource Scheduling
 Creating Fault Tolerance architectures
 DPM: datacenter power management, based on VMotion
 Converting VM into VM for infrastructure, from physical to VM
 vApp: are Virtual Application Services
 Cloning and Moving VM
 Making Templates for VM
 Backup VM and thus virtual servers
G Low level features:
 VMotion: VM moving among Hosts
 Dynamic increment of: CPU ck, MEM, net…


Summary of
VMware
vSphere 4.0


VMotion of
VMware vSphere
G you need:
 VM without snapshots
 VM must be powered off to
simultaneously migrate both
host and datastore
 Compatibility among host CPU
and VMs
 Dedicated virtual network
 The VMs can be ON
G Steps:
1. moving HD images
2. aligning OS and CPU status
3. off-the old and then on-new


VMware DPM, Power Management
G DPM consolidates workloads to reduce power consumption
 Cuts power and cooling costs
 Automates management of energy efficiency
 optimizing host resources
G Based on VMotion


VMware DRS, Distrib. Res. Scheduling
G DRS is used to balance the workload among Hosts
G Moving VMs is a tools for balancing the workload
on Hosts
G ..


VMware Fault Tolerance, FT
Single identical VMs running in lockstep on separate hosts
Zero downtime, zero data loss failover for all virtual machines in case
of hardware failures
Single common mechanism for all applications and Operating
systems
Need to have a storage shared
by the same VM
The VM itself can be
stored in the SAN APP APP APP
OS OS OS

VMware vSphere™

SAN/NAS

How VMware FT Works

Primary Secondary
Virtual Machine Virtual Machine

VMkernel VMM VMM VMkernel
Log Update? Log Read?
Record Logs
Log Buffer Log Buffer
Heartbeat?

Read/Write Read

Single Copy of Disks on Shared Storage (SAN)


HA: High Avalability of vSphere
G When a host fails, the
running VM on the host
may be turned ON on
another hosts
 Just the time to turn on
again the host
G HOT Spare solution:
 It is also possible to
keep aligned 2 distinct
hosts to make a faster
switch OFFON of the
VM on the faulty host
 implies to have
duplicated resources:
Host, CPU etc.


Structure
G Cloud Computing
G High Availability


Security on the cloud
G Protecting VMs from external access
G Protecting VMs each other in the cloud
G Technologies:
 Accessing to other VM via dedicated virtual
networks, using Virtual Networking, Virtual Switch
 Avoiding shared disk, at least using authenticated
connections
 Using Firewall
 Communicating with other VMs via protected
connections: protected WS, HTTPs, SSL, SFTP,
etc.


Virtual Networking
G Per isolare meglio dei guest
OS da altri che non sono
nello stesso trust domain
G isolare i virtual switch dei
singoli trust domain.
G Solo i guest che condividono
lo stesso domain hanno
schede di rete virtuali sullo
stesso virtual switch.
G virtual switch su porte logiche
del sistema host che non
hanno indirizzo ip
configurato.


Vmware Vsphere VMsafe
Security VM
• HIPS
• Firewall
• IPS/IDS
Security API
• Anti-Virus
ESX Server

 Creates a new, stronger layer of defense – fundamentally changes
protection profile for VMs running on VMware Infrastructure
 Protect the VM by inspection of virtual components (CPU, Memory,
Network and Storage)
 Complete integration and awareness of VMotion, Storage VMotion, HA,
etc.
 Provides an unprecedented level of security for the application and the
data inside the VM


VMware vSphere
Data Recovery
Backup Only


Structure
G Cloud Computing
G High Availability


VM Converter, the migration
G Conversion possibilities, migration possibilities

G P2V, from Physical  Virtual machine
 Reusing legacy servers into stronger and new HW machines
 From ISO CD of an OS  VM

G V2V, from Virtual  to Virtual
 Import/export a VM from/to different standards
 From VM Workstation  Infrastructure VM
 From Infrastructure VM  template for VM with some parameters
 From Infrastructure VM VM Workstation
 From Infrastructure VM  Infrastructure VM changing parameters
 …..


P2V, vConverter vSphere
Virtual to be
converted

Converter on
vSphere Client
Physical To be virtualized
with VMware workstation

Helper
….is becomes
the new VM

Physical to be virtualized Converter
Hosting the: Convert Stand Alone, Client
Server

Data, disks, …


P2V, OpenCD vConverter vSphere

Boot from CD with
OpenCD Converter
containing the: Convert Converter on
Stand Alone Client vSphere Client
and Converter Server

Helper
….is becomes
Data, disks, … the new VM
Physical
To be virtualized

Cloning HD physically:
-No resize (+-)
-No partitions
-RAID drivers needed


P2V conversion
G How to convert, typically:
 From VM Converter Stand Alone on Windows or Linux
it includes a Server side and a client consol,
• server side has to be installed on the P machine to be
converted.
• Client side is a consol
From Linux to VM on ESX Host
From Windows to CM on ESX Host
 From VM converter as plugin of vCenter, via vCenter Client
From Windows to VM on ESX Host
 From OpenCD, with included VM Converter and alone
 From VMware Workstation


P2V conversion
G Hypothesis:
 Converter server installed on Physical machine
 Converter in third machine or on the physical (windows)
 All snapshots are removed and the disks re-compacted.
G First step:
 The Converter Client is launched
 The Converter Client creates an Helper VM on the vCenter Host
G Second step:
 The Helper VM works independently to transfer the data from VM
Converter Server to the new VM via the Helper VM, under the
control of Client.
 The Converter Client provides commands to close and destroy the
Helper VM and put in execution the new VM with the final shape


P2V conversion
G Third step:
 Reboot the new machine as Virtual machine
Then install the VMware tools on the VM from the vCenter or
from the VMware Workstation hypervisor.
Revise the general config, network, etc.
See if it possible to use paravirtualized drivers instead of
physical drivers.
 If the converted VM is not booting
Verify the VM setting: disk, operating system, disk drivers, etc.
Regenerate the VM tools on the VM and try again


Structure
G Cloud Computing
G High Availability


vCenter vSphere
G The so called Inventory
G Datacenter
 Cluster00
Host….
Host….
Application:
• VM1
• VM2
VM3
VM4
 Cluster01
H2..
…


Some Snapshots of
vSphere


Storage View


A Larger Datacenter


Partially Collapsed with Separate Physical
Trust Zones


Monitoring the Solution
G Monitoring and assessing performance at level of:
 Datacenter, Cluster, Host
 Virtual Machine from outside

 Virtual Machine inside:
this has to be performed by using tools inside the VM
operating system
Windows:
• System monitoring hosts, detailed performances
Linux:
• Top or other tools


Performance Analysis, Cluster/DC


Datacenter space consumed again time


Performance Analysis, Single VM


Performances Analysis of VM on the Host


Structure
G Cloud Computing
G High Availability


Virtualization Solutions
G VMware vSphere
 Datacenter and VM management, a large range of OS
G Microsoft Hyper-V
 Based on Microsoft Windows Server 2008
G HP Integrity
G Datacenter and VM management, x86, linux, etc.
G No GUI for monitoring
G XEN
 Virtual machine monitor, hypervisor: Suse, RedHat, Sun Solaris, Debian
 X86, 64, PowerPC 970
 GPL licensing
G others
 SWsoft with its Virtuozzo
 IBM Power VM
 VirtualBox of Sun, open source


Comparison (www.ctistrategy.com)


Comparison, 2


Standards in Virtual Machine formats
G Virtual Machine formats on the HD

G OVF: Open Virtualization Format, on the push of VMware
 Format for VM

G VM disk format of VMware is a standard which is
supported by:
 VMware workstation
 vSphere VMware
 VirtualBox of SUN


Structure
G Cloud Computing
G High Availability


How to WORK with VMs in the Cloud
G KVM solutions
 Local access via local KVM
 Local server access via HTTP
G Windowing Terminal
 MS Windows Remote Desktop
 X Terminal to linux
G Remote Solutions:
 VNC, Radmin, etc.
 VNC: Also possible via HTTP port
G Telnet, char based consol, SSH
 VT100 terminal for example


Windows Remote Desktop
I Remotemachine
I Windows 2003

I Localmachine
I Windows Vista


VNC remote connection
I Remotemachine
I Windows XP

I Localmachine
I Windows Vista


Structure
G Cloud Computing
G High Availability
G vSphere Infrastructure, Security on the Cloud
G IaaS solutions
G SaaS Solutions
G PaaS Solutions


Modello Generale


Infrastructure as a Service (IaaS)
G erogazione di servizi infrastrutturali relativi a:
 CPU come clock, storage as HD, rete, mem
 altri elementi di base assolutamente indipendenti da servizi
applicativi di qualunque tipo.
G infrastruttura messa a disposizione dal provider per
eseguire la propria applicazione:
 Affitto di Server e/o macchine virtuali
 pagamento in base al consumo dell’infrastruttura
 lascia sotto la responsabilità dell’utente la gestione del
sistema operativo, dell’eventuale middleware e della parte di
runtime, oltre che dell’applicazione stessa.
G Amazon EC2 è un esempio di servizio IaaS.


IaaS
G Una serie di Host, VM e NAS/SAN
G Un gestore di macchine virtuali
G Un sistema per il deploy basato su templates
G Un sistema per l’accounting e il deploy automatico
G Alcuni servizi per l’accesso diretto a Host e VM


IaaS Solutions
G Based on Virtualization
G A set of Datacenters, Hosts, NAS/SAN, networks, etc.
G IaaS commerciali come
 Amazon’s Elastic Compute Cloud (EC2) and Simple Storage Service
(S3). Amazon EC2 è il leader del mercato dei provider IasS.
G Come Technology provider di soluzioni IaaS si hanno:
 Oracle,
 VMware: Vcenter, Vsphere, etc.;
 Nimbus;
 OpenNebula;
 HP;
 Eucalyptus; Ubuntu Enterprise Cloud (basato su Eucalyptus);
 GoGrid;
 Flexiscale;
 UNISYS,
 Enterprise Cloud Manager, ECM, on Hybrid.


Service Model IaaS

Hosting.com

OpenNebula
NephoScale
BitRefinery

Eucalyptus
Rackspace

ReliaCloud

Terremark
OpSource
GoDaddy

Softlayer
Amazon

Nimbus
GoGrid
Infrastructure

EC2
Feature

Pay As You Go x x x x x x x x
Dynamic Service Level
Agreement
Certificazioni (e.g., PCI
x x x x x x x x
o SAS 70)
Scale Up x x x x x x x x
Scale Out x x x x x x x x x x x x x x
Live Support x x x x x x x x
Monitoring Tools x x x x x x x
APIs x x x x x x x x x x x
Free Tier x x x x
Highly customizable
x x x x
istances
Cloud Burst x


IaaS confronto (legenda)
G Pay per use – se si paga a consumo, in realtà molti provider hanno una filigrana di offerte
più ampia, includendo anche piani mensili, sconti e promozioni ecc.
G Dynamic SLA – se viene offerta la possibilità di ridefinire gli SLA.
G Certifications – se il provider offer certificazioni sulla compliance/sicurezza come PCI o
SAS 70.
G Scale Up – se è possibile lo scale up di single istanze di server, tramite l’aggiunta di
memoria, extra CPU o storage.
G Scale Out – se è possibile fare il deploy veloce di nuove istanze dei server.
G Live Support – può essere diviso in:
 Poor (n) – solo forum di supporto for free; in alternativa a pagamento.
 Average (y) – supporto 24×7 gratis (telefono, chat, forum).
 Extensive (y) – offerte di supporto multiplo per ogni soluzione proposta.
G Monitoring – può essere diviso in:
 Poor (n) – nessuna soluzione di monitoring/alert integrata, sono necessari strumenti di terze parti
da acquistare separatamente.
 Average (y) – strumenti di monitoring minimali e senza servizi di alert.
 Extensive (y) – soluzioni complete e integrate di strumenti di monitoring compresi nel prezzo.
G APIs – se vi sono API per interagire con i server.
G Free Tier – se vengono offerte soluzioni di prova per il test dei servizi


Structure
G Cloud Computing
G High Availability
G IaaS solutions
G SaaS Solutions
G PaaS Solutions


Platform as a Service (PaaS)
G erogazione di servizi applicativi di base come:
 sistemi operativi, middleware, linguaggi,
 tecnologie di base dati
 ambiente runtime necessari per eseguire l’applicazione,
G rimane sotto la responsabilità dell’utente,
 applicazione
 la definizione del modello (e.g., numero e dimensione dei
server, datacenter, caratteristiche del networking) da utilizzare
per l’esecuzione dell’applicazione.
G Google AppEngine è un esempio di Platform as a Service.
G A livello PaaS viene anche collocato l’insieme dei servizi
MaaS, Middleware as a Service.


PaaS
G Middleware per accesso ai servizi di base
 Middleware as a Service, MaaS
 API per accesso a questi servizi
G Sistema di sviluppo per applicazioni che possono usare
le API
G Servizi e risorse base

G Business Process Modeling:
 permette definire come allocare e configurare le
macchine virtuali ed i servizi sul cloud.
G Processo di lavoro


Appliant Process Modeler


Apache Hadoop
G framework che supporta applicazioni distribuite
G applicazioni distribuite con elevato accesso ai dati
G Hadoop può lavorare direttamente con qualsiasi file system
distribuito, che possa essere montato da un sistema
operativo sottostante semplicemente usando un file:// URL


Alcuni esempi di PaaS
G Google App Engine (https://appengine.google.com/) detto
anche GAE
G Windows Azure Platform
(http://www.microsoft.com/windowsazure/)
G Force (http://www.salesforce.com/platform/ )
G Oracle Fusion Middleware
(http://www.oracle.com/it/products/middleware/index.html)
G Eccentex AppBase
(http://www.eccentex.com/platform/features.html)
G 3Tera AppLogic (http://www.ca.com/us/cloud-platform.aspx)


Service Model SaaS

Salesforce
Dynamics

RightNow
Microsoft
NetSuite
Aplicor
Software

.com
Feature

Campaign Management
x x x x x
Lead management x x x x x
Event Management x x
Pay Per Click (PPC) integration
x x x
E-mail con trackable link e click-through
x x x x x
Account Management x x x x x
Contact Management x x x x x
Opportunity Management x x x x x
Competitive Intelligence x
Sales Analytics x x x x x
Data Deduplication x
Ticket/case/incident management x x x x x
Email to case creation x x x
Routing x x x x x
Escalation x x x x x
Customer Surveys x x x x
Knowledge base x x x
Self Service Portal x x x x x
Dashboard x x x x x
Ad Hoc Report Writer x x x x x
Data Warehouses x x
OLAP x x x
Groupware Integration x x x x x
Partner relationship management
x x x
E-commerce Suite x
Sales Order Processing x x
Workflow Designer x x
Offline Edition x x x
Enterprise Resource Planning x x x
Certifications ISO, NIST SAS70 SAS70
SLA with Guarantee x


Structure
G Cloud Computing
G High Availability
G IaaS solutions
G SaaS Solutions
G PaaS Solutions


Software as a Service (SaaS)
G erogazione di servizi applicativi accessibili indipendentemente
dalla collocazione e dal tipo di device utilizzato.
G Non sono eseguite applicazioni del cliente
 il cliente paga il diritto (mediante licenza o canone di affitto) di
utilizzo di un’applicazione messa a disposizione dal provider, senza
preoccuparsi di come essa venga realizzata e gestita nel cloud.

G Il cliente deve scegliere la corretta applicazione che soddisfi le
sue necessità

G SalesForce.com Customer Relationship Management (CRM) è
un esempio di soluzione in cui il software è venduto in modalità
as a service.


Architetture SaaS
G Interfacce esterne
G L’insieme di API, rappresentano i metodi con cui le
applicazioni possono venir relazionate con altri servizi o
software,
 Cloud support for mobile, come Apple
 Push service come Apple
G Middleware è lo strumento che permette all’applicazione
di sfruttare tutti i servizi e le risorse di livello più basso


Soluzioni SaaS
G Google Apps (Google) - Google offre gratuitamente ai privati i
propri servizi applicativi come:
 Google Documents, GMail, Calendars, Sites.
G Oracle CRM On Demand in grado di supportare tipologie di
offerta sia in multi tenancy sia in single tenancy
G SalesForce.com (Salesforce) - Customer Relationship Manager.
G Zoho Office (Zoho) - Propone un insieme di applicazioni mirate
alle aziende. Il portfolio di servizi prevede più di 20 applicazioni
tra cui ad esempio posta elettronica, office app CRM
G IBM LotusLive (IBM) - Consiste di una raccolta di soluzioni di
collaborazione aziendale online e servizi di social networking.


GAP: Google App Engine


GAP: Google App Engine
G Datastore: un database non relazionale distribuito e scalabile,
accessibile attraverso un set di API e gestibile attraverso il pannello di
controllo dell’applicazione. Il servizio di basa su BigTable, un sistema
proprietario per l’archiviazione di dati strutturati.
G Google accounts: un set di API che permette l’autenticazione di utenti
attraverso account Google.
G URL Fetch: API che permette di ricavare risorse sul web.
G Mail: API che permette l’invio di e-mail.
G Memcache: sistema di caching di tipo chiave-valore. I contenuti di tale
cache sono condivisi tra le varie istanze in esecuzione dell’applicazione.
G Image manipulation: API utilizzabile per la manipolazione di immagini.
G Tasks: Lo sviluppatore ha inoltre la possibilità di schedulare task
secondo determinati orari o utilizzare una coda di task per permettere
alla propria applicazione di eseguire delle operazioni di background
mentre delle richieste web vengono servite.


SalesForce Java, .NET, Ruby, SOQL, SOSL REST
Force.com PHP, Python, Perl,
129

Adobe Flex e AIR
Red Hat Java, Java EE, Multi IaaS Git, MySQL, REST x
OpenShift Python, Perl, PHP, Provider SSH, MongoDB,
Ruby (AWS, etc.) rsync MemBase,
Memcache
Rackspace PHP, ASP, .NET Multi-tenant Rackspa FTP/S MySQL 5 & MS object x
Cloud Sites PaaS ce Cloud FTP SQL Server 2008 storage/
Files/Aka CDN
mai
Microsoft C#, Java, PHP,
Windows REST
Ruby
Azure
Jelastic Maria DB, MySQL,
Sistemi Distribuiti, Univ. Firenze, Paolo Nesi 2011-2012
Mave
PostgreSQL,
Java n, Git, x
MongoDB,
SVN
CouchDB
Heroku
Amazon RDS
(MySQL),
Ruby
MongoDB, Redis,
CouchDB
PaaS

Google App Google Cloud JDBC,
Java, Python
Engine SQL DB-API
CumuLogic Amazon
EC2,
OpenStack,
Git, MySQL, MySQL
Java, Spring CloudStack, REST x
SVN cluster, MongoDB
Eucalyptus,
VMware
vSphere
Cloudify deploy su
Java, .NET,
qualsiasi Cassandra, CLI,
Groovy, Ruby, C++,
IaaS e BYON Git MongoDB, REST, x
Node.JS, Spring
(bring-your- MySQL, HSQL Web
(more soon)
own-nodes)
Java Spring,
CloudFound Groovy Grails, MongoDB,
Git CLI
ry Ruby Rails Sinatra, MySQL, Redis
Node.js
Mave
REST,
CloudBees Java n, Git, MySQL x
CLI
SVN
AWS
Elastic Java AWS Cloud
Beanstalk

Dashboard/C
Repositories

centralizzata
Tecnologie
Service

Databases
Model

Platform

Network
Delivery
Content
Feature

hosting

onsole
(CDN)
Cloud

APIs

References
G VMware: http://www.vmware.com
G HP: www.hp.com/go/integrityvm
G Microsoft Hyper-V: http://www.microsoft.com/hyper-v-
server/en/us/default.aspx
G “Windows Server 2008 Hyper-V Technical Overview”-
http://download.microsoft.com/download/4/2/b/42bea8d6-9c77-4db8-b405-
6bffce59b157/Hyper-V%20Technical%20Overview.docx
G Comparison of Hypervisors by VMware -
http://www.vmware.com/technology/whyvmware/architectures.html#c132894
G “VMware vSphere 4” Datasheet - http://www.vmware.com/products/vsphere/

G All the pictures representing VMware staff have been taken from Vmware library and
are under rights of VMware. These slides represents only the view of the author and not
that of VMware or of other companies mentioned in the slides such as HP, microsoft or
others. The slides are only provided for didactical purpose and no for profit.


CLoud computing, virtualization, scalability, media computing, course

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to CLoud computing, virtualization, scalability, media computing, course (20)

Recently uploaded (20)

CLoud computing, virtualization, scalability, media computing, course