2. Virtual Infrastructure
Enables You to:
• Dynamically map computing resources
to the business
• Lower IT costs through increased
efficiency, flexibility and responsiveness
• Provision new services and change the
amount of resources dedicated to a
software service
• Treat your data center as a single pool
of processing, storage and networking
power
Virtual infrastructure brings uniformity to your data center
2
3. Intel Architecture with ESX Server
• Runs directly on hardware
• Manages resource allocations
• Scalable to large virtual
machines and high
performance
VM ware E X S
S erver is virtual infrastructure software for partitioning, consolidating
and m anaging system in m
s ission-critical environments. E X S
S erver and VM ware
Virtual Infrastructure Nodes provide a highly scalable virtual m achine platform with
advanced resource m anagem capabilities, which can be m
ent anaged by VM ware
VirtualCenter.
3
4. Core Building Blocks - Virtual Infrastructure Nodes
• ESX Server
• Hosts multiple virtual machines (VMs)
VMotion • Virtual SMP
• Enables dual virtual CPU VMs
• VirtualCenter
VirtualCenter • Enables centralized management
• VMotion
• Enables migration of VMs between physical
Virtual SMP hosts
• Virtual Infrastructure Node (VIN)
• ESX Server + VC Agent + VSMP + VMotion
ESX Server
4
5. Virtual Infrastructure Solutions Available from IBM
•ESX Server
•SMP
•VirtualCenter Agent
•VirtualCenter Management Server
•VMotion
•Virtual Infrastructure Node
•P2V
•GSX Server
5
6. VMware and IBM Relationship
VMware ESX Server is IBM ServerProven and StorageProven –
2001 - present
• ServerProven: x255, x335, x345, x360, and x440, x445 and BladeCenter
• StorageProven: ESS and FAStT
IBM & VMware Joint Development Agreement (JDA) - February, 2002
• IBM was the first tier-one vendor to sign a JDA with VMware
• Allows future enhanced management in VMware environments
• Allows VMware ESX Server to run optimally on IBM's EXA technology
IBM & VMware Support and Distribution Agreement (SDA) - July, 2002
• IBM resells and supports VMware ESX Server
IBM Launches BladeCenter – October, 2002
• BladeCenter delivers integration, performance, manageability, resiliency and investment
protection
VMware announces ESX Server 2 – July, 2003
• ESX Sever 2.0 pushes Intel computing platform to the next level of virtualization
ESX Server 2 is ServerProven for BladeCenter – November, 2003
• ESX Server 2.0 is Server Proven tested and fully supported by IBM
VMware launches VirtualCenter – November, 2003
• Enterprise-class software that delivers unprecedented management capability
6
7. VMware and IBM Relationship, cont’d.
IBM & VMware expand strategic relationship – February, 2004
• IBM now resells the entire VMware virtual infrastructure line of products, including ESX Server
2.x, VirtualCenter, VMotion, VirtualCenter agents, Virtual Infrastructure Node
IBM announces availability of VMware with BladeCenter – April 2004
• Customers experience dramatically improved hardware utilization by running VMware software
on IBM BladeCenter
• IBM now resells custom blade bundles
IBM announces Virtual Machine Manager – August 2004
• Free IBM Director extension that bridges the gap between the management of physical servers
and virtual machines.
• IBM eServer xSeries - #1 8-way solution
• IBM eServer BladeCenter™ - #1 blade platform
•VMware License Packs for BladeCenter
• IBM TotalStorage Solutions
• IBM Director, Virtual Machine Manager, and Remote Deployment Manager
• IBM eServer xSeries Lab Services
• IBM Technical Support for VMware
7
9. What is Business Continuity
Business Continuity = “Always-on” uninterrupted
availability of business systems and
applications
• Types of interruptions to
business continuity:
• Planned downtime
• Unplanned downtime
• Disasters
• Components of business
continuity:
• High Availability
• Disaster Recovery
9
10. High Availability with Virtual Infrastructure
• Reduce unplanned downtime
• Redundant network connections
• Redundant storage connections
• Clustering virtual machines and
ESX Servers
• Reduce planned downtime
• Virtual machine migration with
VMotion
10
11. Hardware Redundancy Features in ESX Server
Virtual Machines have the same redundancy features as the best
physical systems
Redundant
Fibre Channel LUN 0
VM VM
S ervice Switches
Console
LUN 1
ESX Server
LUN 2
Controller0
LUN 3
Managed Ethernet VM VM
Service
Console
Switch Environment
LUN 4
ESX Server
LUN 5
Serv ice Controller1
LUN 6
VM VM
Console
Storage
ESX Server Server
NIC Teaming Multi-path Failover
11
12. Best Practices for Redundancy:
Virtual NIC’s and Virtual Switches
Virtual NIC Virtual NIC
VM VM
Service
Console
Named Virtual Switches
Virtual NIC
VLAN B
VLAN A VM
ESX Server
NIC
Teaming
To physical switch
VLAN trunk port
12
13. Clustering: Virtual Infrastructure Adds Options
• Low-cost protection • Protection from software • Lowest-cost protection
from hardware and failures only from hardware and
software failures software failures
• Low-cost availability
• Can add failover improvement • High availability the most
machines as needed flexibility
13
14. Individual Virtual Machine Failover
Linux Windows Windows
Virtual Virtual Virtual
Machine Machine Machine
Faulted Virtual
Machine
restarts
Heartbeat Monitoring Heartbeat Monitoring
VMware ESX VMware ESX
Service Console Service Console
x86 Hardware x86 Hardware
14
15. ESX Server Failover
Linux Windows Linux Windows
Virtual Virtual Virtual Virtual
Machine Machine Machine Machine
Restart
Virtual
Machines
Heartbeat Monitoring Heartbeat Monitoring
VMware ESX VMware ESX
Console OS Console OS
X86 Hardware X86 Hardware
15
16. VMotion: Online Virtual Machine Migration
• Enables real-time online migration
of running virtual machines
• Persistent connection throughout
migration
• Can be automatically initiated when
• Critical alarm is generated for the host
hardware
• Host hardware utilization exceeds
specified level
• It is time for the scheduled maintenance
16
17. VMotion Raises Workload Availability
VMotion enables zero-downtime maintenance and workload
management
• Temporarily free
servers for
hardware
maintenance
• Dynamically
rebalance
workloads to
servers with more
computing
bandwidth
17
18. Benefits of Virtual Infrastructure for High Availability
• Lower cost
• Flexibility in use of physical hardware
• Planned hardware maintenance downtime eliminated by VMotion
• Simpler to implement
• Flexibility in clustering and failover allows choice of best fit for each
environment
• Easy to activate availability features like NIC teaming,
• Greater reliability
• Redundancy features of best physical systems
• Clustering options with virtual machines make it possible to cluster
more applications and systems
18
19. Backup Options for Virtual Infrastructure
1. VM-based backup agent
2. Backup entire VM with backup agent in Service Console
19
20. Backup Agent in Virtual Machine
• Backup agent in each VM
• Functionality identical to
physical machines with
installed backup agents
• Allows for file level and system
Backup
state backup
Server
Tape
Array
20
21. Backup from Service Console
• Backup agent resides in the Service Console
• Capture a complete .dsk or .vmdk file representing a virtual machine
• Reduces overhead on individual VM’s
• Requires virtual disk to be quiesced before reboot
Backup
Server
Tape
Array
21
22. Backup leveraging Raw Disk Mapping
Backup server
LUN 1
VMFS RDM
LUN 4 NTFS or EXT3
22
23. Replication Options
• SAN based replication
(Block-based)
• e.g. RVM, PPRC
• IP based replication (File-
based)
• e.g. Replistor (Legato)
23
24. Recovery Options
• Physical to Physical
• Existing scenario without Virtual
Infrastructure
• Physical to Virtual
• Reduces cost, improves time to
recovery, and increases flexibility
• Virtual to Virtual
• Greatest flexibility, lowest cost, best
time to recovery
24
25. Physical to Virtual Recovery
• Virtual Machine templates with
operating systems and backup agents
are created and archived
• Full system restore occurs into Virtual
Machine using existing recovery agents
25
26. Physical to Virtual Recovery Preparation
P2V Assistant performs all necessary substitutions to transform
a physical system into a production-ready Virtual Machine
• Creates an image of the source
machine using either the VMware
imagining tool or a 3rd party Create
imagining product Source Image File(s)
Physical System
• Performs all necessary HAL and
Virtual Disks
driver substitutions to make the
image bootable
• User can manually modify any
additional settings before having a
production-ready system VM Host Virtual Machine
26
28. Virtual to Virtual Recovery Options
• Same as physical to physical: Replicate backup and
recovery process on all virtual machines…
• OR
• System-level capture and recovery
• Manually copy virtual machine file to a different location
• Use a backup agent installed in the ESX Server console or on the
host OS
• Shared storage
• In case of a hardware failure, recover virtual machine from shared
storage to a different physical system
• Disk / LUN replication
• Run disk replication from inside the VM
• Run storage array based replication
28
29. Recovery with Shared Storage on SAN
• Hosting VMs on SAN
increases system
availability
• If for any reason
hardware hosting virtual
machines fails, VMs can
be restarted on a
different platform.
• The only downtime is
time required for VM
booting
29
30. Recovery with Shared Storage on SAN
• Hosting VMs on SAN
increases system
availability
• If for any reason
hardware hosting virtual
machines fails, VMs can
be restarted on a
different platform.
• The only downtime is
time required for VM
booting
30
31. Recovery with Array Based Replication
• Speed up recovery in solutions based on storage replication
• No need to upgrade secondary site server hardware in lock-step with the
primary site
• Easy to automate and no need for bare metal recovery tools
31
32. Recovery with Array Based Replication
• Speed up recovery in solutions based on storage replication
• No need to upgrade secondary site server hardware in lock-step with the
primary site
• Easy to automate and no need for bare metal recovery tools
32
33. Customer Example: Disaster Recovery
Challenge:
• Large chemical company wanted a cost-effective, fast time-to-
recovery solution for their critical applications
Solution:
• Virtualize primary and recovery datacenters using ESX Server
• Use replication software to mirror SAN data
33
34. Business Continuity Customer Example
VMware boot images 17 miles
PRIMARY RECOVERY
SITE SAN SITE SAN
Synchronous /
Asynchronous
Primary Site: Replication Recovery Target:
400 VM’s, 78 ESX 400 VM’s, 50 ESX
Servers Servers
Result: 16.5 minutes to failover!
34
35. Benefits Summary: Virtual Infrastructure
for Better Business Continuity
• Reliable recovery and faster time to recover
• Flexibility to restore to any hardware
• Single stage recovery process
• Lower cost
• Consolidation and pooling of recovery hardware lowers costs
• Simpler to implement
• Standardized procedures
• Application independent backup and recovery
• Interoperability with existing 3rd party tools
35
37. VMware GSX Server 3
Enterprise-Class Virtual Infrastructure for Intel-Based Servers
• VMware GSX Server is virtual infrastructure
for enterprise IT administrators who want to:
• Streamline development and testing operations
• Consolidate departmental workloads
• VMware GSX Server:
• Enterprise-proven across thousands of
customers for the last 3+ years
• Preserves freedom of choice by installing on the
widest variety of Windows and Linux operating
systems
• Offers an upgrade path to datacenter
virtualization
37
38. GSX Server
• Installs like an application – easy
to deploy and manage
• Integrates easily into Microsoft
Windows® and Linux host
environments
• Supports the widest selection of
host and guest operating systems
• Device support inherited from
host operating system
• Portable, hardware-independent
virtual machines
• Can be managed by VirtualCenter
• Upgrade path to ESX Server
38
39. New and Enhanced Features of VMware GSX Server 3
• More platform choices • Centralized Management and
• Supports latest provisioning
Windows/Linux/NetWare OSes • VirtualCenter-based
• Automated virtual test lab customization and provisioning
of server VMs
• Integration with leading test
• Windows integration for
automation solutions
performance monitoring and
• Automatic VM start-up & shutdown event logging
• PXE provisioning of virtual machines • Enterprise-class server VMs
• VMs can use remote client CD • 3.6GB per VM for server-class
drives workloads
• Direct upgrade to ESX Server • Teamed network adapter
• Seamless VM migration to ESX support, SCSI backup devices
Server when highest performance • 10-20% improvement in disk
and scalability needed and networking performance
39
40. Comparison of VMware GSX Server 3 and ESX Server 2
GSX Server 3 ESX Server 2
Computing power GSX Server supports single CPU ESX Server provides virtual SMP to
applications such as Web apps, DNS and address the computing needs of
Active Directory, print and file Servers, and databases, SAP, ERP, and Exchange
custom VB applications
Disk and Network I/O GSX Server supports intermediate disk ESX Server offers higher performance for
and network I/O requirements heavy disk and network I/O applications
Architecture GSX Server is a hosted architecture and ESX Server is a hostless architecture and
installs like an application on Linux and installs directly on the hardware
Windows
Consolidation ratios GSX Server supports about 4 virtual ESX Server supports about 8 virtual
machines per host CPU machines per host CPU
Resource management GSX Server provides static memory ESX Server provides fine-grained,
allocation dynamic resource management enabling
customers to offer guaranteed SLA
VirtualCenter and GSX Server is VirtualCenter-ready to ESX Server supports VirtualCenter and
VMotion enable customers to manage multiple VMotion which allows running virtual
GSX Server instances machines to be moved
40
42. Why VMware on IBM?
IBM eServer products such as the x445 and BladeCenter have industry-
leading hardware reliability and performance and are fully certified with
ESX Server
Only IBM can deliver a 16-way VMware ESX Server solution
Best
of breed storage capabilities with eServer xSeries, FAStT storage,
and VMware ESX Server
IBM supports VMware, OS’s (Windows, Linux, etc.) running in a VM
directly, and selected applications, reducing risk and downtime
IBM management solutions with IBM Director enhance hardware
availability and manageability – IBM Director Agent supported in ESX
service console
IBM ServerProven & IBM Total Storage Proven testing reduces risk
42
43. Scalability with VMware and IBM xSeries
Benefits scale across deployment size, server form factors
VirtualCenter Suite
w/ VMotion
Mix and match
128+ form factors
Number of Physical CPUs
BladeCenters
16-100s+ CPUs
64
Multiple
x445 Servers
ESX Server SAN Storage
32 + V-SMP
16
ESX Server
Multiple
Choose the right
8 2-4 CPU
SMP
x445 Servers SMP Servers platform for each
4
Servers customer
4 8 16 64 128 512+
Number of Virtual Machines
43
44. Scale Up vs. Scale Out
[ Virtual Infrastructure Solutions from IBM ]
x445 BladeCenter
VMotion
ESX 2 or 4-way
Console Blades
ESX Server VirtualCente
r Console ESX ESX ESX ESX
ESX Server on 4-16x SMP System VirtualCenter w/VMotion on BladeCenter
Large single system image Greater resiliency and availability
• Multiple 2X/ blade and ESX instances
4X
Automatic load-balancing of large number of • “Standby” blade within chassis for availability
VMs in same system
• Simplified resource sharing & mgmt Modular scalability
• Increased scaling of memory & I/ capacity
O
Lower entry cost for smaller systems
• SCSI disk can be used Blade form factor advantages
• Increased density, standardization, power
efficiency
Workload balancing with mixed blades
• e.g. temporarily migrate VM to higher
performance blade to complete task more quickly
44
46. Management for VMware Environments
• IBM provides the complete VMware systems management solution
• IBM Director for physical and virtual machine management
• Virtual Machine Manager
• VMM is a free add-on to IBM Director
• “Single glass management” of virtual / physical machines
• Integration point for VMware VirtualCenter and IBM Director
• Drive VMotion through hardware health events
•VMotion provides live VM migration between blades
• Ease-of-administration and self-healing of ESX Server environments
• Remote Deployment Manager 4.20 in supports deployment of ESX
Server hosts on bare metal blades
46
48. Automated VMotion based on Hardware PFA
• Event action plan to run VMotion triggered by hardware PFA reported
by VMware ESX Server host
1. Create a new event action plan
2. Using Event Action Plan Builder wizard, specify the hardware PFA event(s) for
filtering
• Disk error, fan speed, CPU temperature, memory bit errors
1. Associate the hardware PFA event(s) with the event action plan (drag-n-drop)
2. Define the action for migrating VMs from the failing host to the backup host
• Use Customize Action dialog box
1. Associate the action with the event action plan (drag-n-drop)
2. Associate the new event action plan with the host to be monitored for hardware
PFA (drag-n-drop)
• IBM Director Agent on failing host will detect PFA and pass the alert
to IBM Director Server
• IBM Director Server will invoke the enabled event action plan through
VMM
• VMM communicates with VMware VirtualCenter to invoke VMotion of
VMs from failing server to backup server
48
49. VMM Requirements
VM ware VirtualCenter M anagem Sent erver is
virtual infrastructure managem software,
ent
providing a central and secure point of control for
your virtual com puting resources.
VirtualCenter Management Server
VC Management
Server
Director 4.2 Agent
VirtualCenter Agent
User Director 4.2 Server
49
50. Core Building Blocks - Virtual Infrastructure Nodes
• ESX Server
• Hosts multiple virtual machines (VMs)
• Licensed per physical CPU
• Virtual SMP
VMotion • Enables dual virtual CPU VMs
• Licensed per physical CPU
• VirtualCenter Agent
VirtualCenter • Enables centralized management
• 1 VC Agent per physical CPU on ESX host
• 1 VC Management Server per datacenter
• VMotion
Virtual SMP • Enables migration of VMs between physical hosts
• Licensed per physical CPU of ESX host
• Virtual Infrastructure Node (VIN)
ESX Server • ESX Server + VC Agent + VSMP + VMotion
• Most popular VMware server product
• Convenient bundle price
50
51. IBM Support for VMware
ITS Support Line or ServicePac for VMware provides
• Single contact for support issues
• 24 x 7 coverage
• 1 or 3 Year Contracts
• IBM supports Windows, Linux, firmware, IBM Director, VMware and
many applications in a VMware environment
• IBM interfaces with VMware, guest OS vendor, hardware option
vendors
• All licensed VMware components + HW firmware + Operating System =
ServicePac
• Optional Support Line Contract available
51
52. Support Line Vs. ServicePac – Supported Products
Product ServicePac Support Line
xSeries System Software (BIOS, Drivers, Firmware, Microcode, etc)
xSeries
IBM Director (Server, Agent, RDM, Software Distribution, Plus Pk)
Hardware Clustering
Microsoft Operating Systems (XP, NT, 2000, 2003)
Microsoft
Microsoft Clustering
Microsoft Applications (SQL, Exchange, BizTalk, Office, etc)
Linux Operating Systems (Red Hat, SUSE, Turbo Linux)
Linux HA Clustering
Linux
Linux High Performance Clustering
Linux Apps (Apache, Directory Server, Samba, Developers Kit, etc)
SW Support on non-IBM HW
Other
VMware (ESX, Virtual SMP, VirtualCenter)
52
53. Support Line for Windows / Linux
Support Line Highlights
• Voice support through 1-800-IBM- Supported Products
SERV (option 2) SL for SL for
SL for
Linux
• Electronic support through web Windows Linux
Clusters
• 24x7 coverage for all problems xSeries &
Yes yes yes
IntelliStation
• Unlimited support calls for 12 months
IBM Director Yes yes yes
• “Per Environment” support
High Availability
• SW support on OEM systems Clustering
Yes yes yes
Windows Yes no No
Linux No yes yes
Account Advocate (option) VMware Yes yes yes
• Single Point of Contact High Performance
No yes yes
• Account Management Clustering
• Monthly status call MS Apps Yes no No
• Problem management/escalation Linux Apps no yes yes
53
So let’s get right into what virtual infrastructure is, and what it can bring you. Virtual Infrastructure provides a layer of abstraction between the computing, storage and networking hardware, and the software that runs on it. WE BRING UNIFORMITY TO THE DATACENTER WITH VIRTUAL INFRASTRUCTURE. WE SMOOTH OVER THE DIVERSITY. Now we can do management of hundreds of machines using an approach with built-in consistency. Now we can share resources to increase efficiency and utilization. Now IT can have a data center that’s more responsive to business needs. Virtualization at the data center level, across many physical machines – what we call “Virtual Infrastructure” simplifies the data center so companies leverage their resources while controlling costs.
And of course, what our Virtualization software is actually doing is acting as a “traffic cop” – working with the operating systems of each guest, or virtualized machine, to ensure that each gets the appropriate time using the actual hardware. In other words, the operating system layer of each “virtual” machine still believes it sees and controls the hardware – but in truth, what it believes to be the hardware layer is actually software: our virtualization layer, the virtual machine monitor to be precise, is exporting the same interface as hardware. Now, this addition of VMware software comes with benefits and costs. The benefits are that we can create flexible partitioning, isolation, and encapsulation. The costs are that there is some overhead, and thus a small inefficiency in consolidation. Let’s do a quick review of the benefits, and then chat about capacity planning.
VMware server virtualization product line is made up of several products. ESX Server is a key product which enables virtualization, including all the key benefits of Hardware abstraction, encapsulation, and partitioning. Virtual SMP is an add-on to ESX Server which enables customers to deploy virtual machines with 2 virtual CPUs. VirtualCenter is the new technology which facilitates the move from mere virtualization and consolidation to true Virtual Infrastructure. It enables centralized management of both virtual machines and physical servers they reside on. VMotion is an add-on to VirtualCenter which enables migration of a “live” stateful virtual machine to a different physical platform without interrupting user sessions
Virtual Machines have the same redundancy features as the best physical systems ESX Server supports multiple connections to both a standard network (NIC teaming) and to a SAN (multi-pathing) for increased redundancy. Physical NICs can be teamed together to create a bond that provides automatic network load balancing, redundant network connectivity and increased throughput. By default, ESX Server balances the load on the adapters in the bond. However, it can also be configured in fail-over mode, in which traffic is routed through one adapter unless that adapter fails, at which point another adapter in the bond immediately takes over. Need for NIC fail-over is monitored in multiple ways, not just based on link status: Software watchdog makes sure the HW is healthy Inter-VMNIC beacons Best practices: Use bonds for redundancy and increased throughput Bind only adapters with similar capabilities. Reason: Although ESX Server supports acceleration features such as VLAN tag handling, checksum calculations, and TCP segmentation offloading, it can only use features supported by all NIC’s in a bond. ESX server also supports multi-pathing. Implementing multi-pathing requires 2 Fibre Channel HBA’s per ESX Server and redundant fibre channel switches to provide multiple paths to the shared storage. Best practices: Redundant HBAs connected to redundant fibre channel switches to ensure availability Shared VMFS’s should be set to ‘public’ mode
Virtual Machines have Virtual NICs. Virtual NICs plug into Virtual Switches which are named. You can give virtual switches common names to identify the network that the virtual machines attached to this switch are connecting to. For example you could name the switch “corpnet” eluding to the fact that the physical NIC attached the this virtual switch is attached to a corporate network. Here are some best practices for use when configuring networking in your virtual infrastructure: Attach a virtual NIC to a Virtual Switch that has multiple physical NICs connected to it. This is extremely valuable for mission critical applications and applications that require a large amount of network bandwidth. For virtual infrastructure nodes VMware requires a minimum of 3 network adapters: 1 dedicated to the virtual machines, 1 dedicated to the service console and 1 dedicated to VMotion migrations. Use High Performance (vmxnet) network driver
Application clustering is a common way to increase availability. How Clustering Works: Application fails Backup application identifies failure and takes over With virtual infrastructure you have additional clustering options that enable more flexible, less expensive clustering for high availability. Physical to Virtual Clustering (ESX only) Physical machine is clustered with a Virtual Machine (VM) Physical machine is primary, virtual machine is backup Shared storage required (SAN) Cluster using SAN LUN's set up as RAW disks Heartbeat across VMNIC interfaces Benefits: Protects against both hardware and software failures. Adding additional fail-over VM’s is easy—just provision a new VM from VirtualCenter “ Cluster in a box”: one VM on the physical system is the primary, a different VM is the backup Can be placed on a SAN with shared LUN storage (disks are ‘shared’) Heartbeat across VMNET or VMNIC interfaces Benefits: Only protects against software (application and OS) failures. Low-cost way to get some additional availability. Cluster Across boxes: cluster nodes (VM’s) are on separate ESX Servers. VM on one ESX Server is primary, VM on second ESX Server is backup Can be placed on a SAN with shared LUN storage (disks are ‘shared’) Clustering Support: Clustering is done via clustering software (e.g. Microsoft Cluster Server, Veritas Cluster Server, Steeleye), and is not inherent in the VMware ESX Server Applications running in virtual machines need to be cluster-aware Requires VMFS virtual disks to be on a shared volume on a SAN Can refer to VMware ESX Server Administration Guide for clustering examples
Virtual machine failover is a means to provide higher availability for applications that are not cluster aware and/or cluster aware apps for which you can’t justify the cost of implementing an application cluster. Virtual machine failover is OS and application independent. Do you prefer to pay for the additional OS licenses, application licenses, and additional maintenance costs for failover virtual machines unnecessarily? This can all be eliminated because you can restart the same virtual machine with its OS, applications, etc. on the failover server—a benefit of the encapsulation of virtual machines. Virtual machine failover is available via Veritas Cluster Server (VCS) for VMware or can also be implemented via scripting. VCS monitors virtual machines for failure as if they were applications, and can restart them locally or remotely using the same failover paradigm as traditional clustering. Why not just cluster the virtual machines using a standard cluster technology such as Microsoft Cluster Server? The difference between clustering and VM failover is that we are automating the recovery of a system from a crashed state for a non-cluster-aware application. A cluster-aware application will maintain availability and not drop transactions a failed VM, recovered on a secondary system.
You can also cluster the ESX Server hosts If physical ESX host fails, the same Virtual Machines will restart on surviving node in cluster N + many implementation Shared storage (SAN) required. Heartbeat across ethernet0 within ESX Server Service Console VCS monitors virtual machines for failure as if they were applications, and can restart them locally or remotely using the same failover paradigm as traditional clustering.
One powerful capability that you have available in a virtualized environment is VMotion—a VirtualCenter component that enables you to move live, running virtual machines between physical systems – with 100% uptime and no dropped connections. VMotion can be automatically initiated in response to critical alarms, utilization above thresholds, and scheduled maintenance.
Ensuring workload availability, i.e. that workloads remain active and available to users, is another key element of ensuring high availability. Using VMotion you can eliminate most planned downtime for hardware maintenance. Rather than needing to bring down a physical server and the applications running it in order to do maintenance, you can use VMotion to move running virtual machines off of the physical host, perform maintenance, restart the host, and then use VMotion to return the running virtual machines to their original host hardware. You can also use VMotion to proactively manage workloads to increase availability: moving virtual machines off of over-taxed systems or moving them off of systems encountering potential problems. You can also repurpose hardware without requiring application downtime.
(Read slide)
VMware virtualization software has been tested an certified with leading backup products, so you don’t need to change your backup plans. We’ll look at specific backup options in the following slides.
VM-based backup agent (ESX Server and GSX server): agent is installed in each VM. This provides file-level backup and restore but is usually the slowest method.
Service-console based backup agent (ESX Server only): with Linux agent installed directly in service console, can back up complete virtual machines. However, you can only restore complete virtual machines and may need to create a new VM configuration file when restoring to a different system.
Replication reduces the time to recovery because data does not need to be restored from backup. Replication can be either block-based, which is the case for SAN-based replication, or file-based. SRDF and SnapMirror are examples of block-based replication software toolds that can be used with ESX Server. Legato Replistore is an example of a file-based replication software package.
With virtual infrastructure, physical to physical recovery is your only recovery option. However, physical to physical recovery has significant problems: Hardware bottlenecks Require one-to-one duplication with identical hardware at recovery site. Thus, need to find a separate target recovery server that exactly matches each primary server if a disaster occurs. Long, lengthy process requiring manual intervention Many steps necessary before one can start “single-step automatic recovery” from backup server Hard to train personnel Complex processes, limited equipment availability complicate training
Virtual infrastructure gives you the option of recovering physical systems to virtual machines. Steps in implementing physical to virtual recovery: Use P2V tool from VMware to create a virtual machine that matches the physical machine. Note that P2V needs to be re-run whenever the physical system changes. Locate recovery target hardware (doesn’t have to match primary site hardware). Install ESX server on recovery target machine. When primary site goes down, start up virtual machines, restore from backup, and restart applications. This recovery plan has significant benefits: Can use any hardware as recovery target—doesn’t need to match primary hardware. Can recover multiple physical systems to one physical system. Greatly improved time to recovery. Recovery processes that took many hours or days can be reduced to a few hours or less.
As an aside, here’s an overview of how the P2V tool works. (Read slide)
In a virtualized datacenter, virtual to virtual recovery provides you a fast, reliable, and simple way to implement disaster recovery. Benefits: Can restore to any hardware No need to install and configure OS, applications, etc—just copy over virtual machines and data (if not already using replication or mirroring) and boot VM’s. Significantly improved time to recovery.
There are several options for disaster recovery in the virtual-to-virtual recovery scenario. You have the option of replicating the processes from physical-to-physical recovery in your virtual infrastructure environment: install backup/recovery agents on each virtual machine and initiate restore from backup for each. However, you also have several other options to consider: (Read list). Let’s look at these options in more detail.
When virtual infrastructure is deployed in conjunction with shared storage, such as SAN, recovery becomes much easier and faster. If a physical server hosting VMware software fails, virtual machines can simply be re-started on different hardware—the virtual machine disk already exists. (CLICK to show failure, click again twice to show recovery on other virtual system)
Those of you who are in charge of mission critical applications are probably familiar with image based recovery. You can use SAN vendors’ software to replicate individual volumes and whole storage arrays to a different physical location bit-by bit. This SAN based solution is great for guaranteeing data and application recovery, particularly in a virtual infrastructure environment. Unlike a physical server environment, where servers targeted for recovery need to replicate the primary datacenter one to one in quantity, configuration, make and model, and even purchase date, virtual infrastructure frees you from lock-step upgrades. Recovery can now be automated. For example, one of VMware’s manufacturing customers reduced the time to recover the complete datacenter in a different city to 16 minutes from when the disaster is declared! (CLICK to several times to step through failover)
Let’s look at a disaster recovery implementation scenario taken from a real customer, a large US chemical company. In the primary datacenter: Many virtual machines running on ESX Server, booting from the SAN. Connection to a SAN for both data and boot images. At the disaster recovery site we have machines available as targets for recovery as well as another SAN. These machines have ESX Server installed. Synchronous or asynchronous replication software is used to replicate data to the recovery site SAN. In this configuration, if a disaster occurs at the primary site, our recovery consists of: Recover the virtual machine template files to the recovery targets. Boot the virtual machines. Restart applications (depending on type of replication used, may need to recover or repair application data).
Key message is that VMware Virtual Infrastructure allows IBM to sell whichever platform provides their customers with the required amount of scalability and flexibility immediately and into the future. There is no limit to the scalability of this solution: you can start a customer on the smallest platforms, add bigger systems as they start to grow and eventually add complete Blade Servers to the overall solution giving up to 100s of physical CPUs.
VMware server virtualization product line is made up of several products. ESX Server is a key product which enables virtualization, including all the key benefits of Hardware abstraction, encapsulation, and partitioning. Virtual SMP is an add-on to ESX Server which enables customers to deploy virtual machines with 2 virtual CPUs. VirtualCenter is the new technology which facilitates the move from mere virtualization and consolidation to true Virtual Infrastructure. It enables centralized management of both virtual machines and physical servers they reside on. VMotion is an add-on to VirtualCenter which enables migration of a “live” stateful virtual machine to a different physical platform without interrupting user sessions