Transcript Slide 1

Virtualization for
Tier-1 Applications
with VMware vSphere
Presented by: Rick Scherer, VCP-vExpert
Today’s Agenda
• Speaker Bio
• What’s New in VMware vSphere 4.0
– Name Changes
– New Packages
– New Features and Enhancements
• Building a Robust Virtual Environment
– Focusing on the Four Fundamentals
• CPU – Memory – Disk – Network
Speaker Bio
•Worked in IT for 12 years
•Systems Administrator for 10 years
•UNIX Administration (Solaris, AIX, Linux)
•Windows Administration (Windows NT - 2008)
•Network Design and Admin. (Cisco Catalyst, Nexus and MDS)
•Programming (Perl, Shell, PHP, HTML, PowerShell)
•Worked with VMware products for Over 6 Years
•Workstation 3.0, VMware GSX, ESX 2.0
•VMware VCP since 2006
•VMware vExpert in 2009
•Took VMware VCDX Design Exam April 2nd
•Technical Editor of “Mastering VMware vSphere 4.0” Book
•Founder of VMwareTips.com
WHAT’S NEW IN VSPHERE 4.0
VMware Name Changes
Old Name
New Name
VMware Infrastructure VMware vSphere
VMFS VMware vStorage VMFS
VMware Consolidated Backup (VCB) VMware vStorage APIs for Data
Protection
VMware VirtualCenter VMware vCenter
VMware Update Manager VMware vCenter Update Manager
VMware Capacity Manager Vmware vCenter CapacityIQ
VMware Converter VMware vCenter Converter (integrated)
VMware vCenter Converter Standalone
VMware Lifecycle Manager VMware vCenter Lifecycle Manager
VMware Lab Manager VMware vCenter Lab Manager
VMware Stage Manager VMware vCenter Stage Manager
VMware Site Recovery Manager (SRM) VMware vCenter Site Recovery Manager
VMware VI Toolkit VMware vSphere PowerCLI
VADK VMware Studio
VMware vSphere 4.0 Packaging
ESSENTIALS PLUS
Integrated availability
solution for Small
Businesses
($2995 All-in-One for 3
Servers)
ESSENTIALS
Basic management of
free ESXi
($995 All-in-One for 3
Servers)
Scale limited,
Low initial price,
Small officeoriented features
STANDARD
($795 / CPU)
Simple
Consolidation
Basic management
for:
•Simple
consolidation
•Remote offices
•Test labs
Scale unlimited,
Value price, Low TCO
Full featured
ADVANCED
($2245 / CPU)
ENTERPRISE
($2875 / CPU)
ENTERPRISE PLUS
($3495 / CPU)
Simplified
Operations
Availability
Automated
Resource
Management
Automated resource
management for:
•Production
infrastructure
•Mission critical
applications
Large scale
management and
integration for:
•Internal cloud
•Tier 1 applications
High availability for:
•Production
infrastructure
•Mission critical
applications
VMware vSphere 4.0 Packaging
New Features in vSphere 4.0
• VMware VMDirectPath
– Technology that enables Virtual Machines to directly
access underlying hardware devices.
• VMware vNetwork Distributed Switch
– Abstracts the configuration of Virtual Networking
from the Host Level to the Datacenter Level
• VMware vNetwork Third Party Switch
– APIs to allow Third Party network companies to create
externally managed Virtual Switches, example: Cisco
Nexus 1000V
• VMware vStorage Thin Provisioning
– Thin Provisioning Functionality for VMDKs
More New Features in vSphere 4.0
• VMware Fault Tolerance
– Zero Downtime, continuous availability of Virtual
Machines, made possible by VMware vLockstep.
• Hot Add & Hot Plug
– The ability to hot add or remove CPU, Memory, Virtual
Storage or Networking devices in a running Virtual
Machine.
• VMware Data Recovery
– Disk-based backup and recovery of your Virtual
Machines, file and image level full and incremental
backups. Recover an entire VM image or recover
individual files and directories. (Not an upgrade for
VCB)
More New Features in vSphere 4.0
• vShield Zones
– Virtual Appliance that provides a dynamic firewall
capability for applications as they move around a DRS
cluster.
• vApp
– Functionality that enables the construction of a multiVM entity and encapsulates information about the
relationship between VMs and their service level
requirements in OVF.
• vStorage APIs
– Storage Vendor Integration with vSphere, enables
offloading of storage tasks.
More New Features in vSphere 4.0
• VMware vCenter Linked Mode
– New capability in vCenter Server that allows multiple
VCs to share roles, permissions and licensing
information. The true gateway to Cloud Computing.
• VMware vCenter Server Heartbeat
– Provides Continuous Availability for vCenter Server.
• VMware vCenter Chargeback (Late 2009)
– Chargeback mechanism built into vSphere Client
• VMware vCenter AppSpeed (Late 2009)
– Formally B-Hive Conductor, end to end application
monitoring to ensure application performance SLAs.
Performance Enhancements in
VMware ESX 4.0
VMware ESX 3.5
VMware ESX 4.0
vCPU per Virtual Machine
4
8
RAM per Virtual Machine
64GB
255GB
NICs per Virtual Machine
4
10
Physical CPUs per Host
32
64
Physical RAM per Host
256GB
512GB
Virtual Machines per Host
128
256
Max Network Throughput
9Gb/s
40Gb/s
100,000
400,000+
Physical CPUs per Cluster
1024
4096
Physical RAM per Cluster
8TB
32TB
IOps per Host
BUILDING A ROBUST
VIRTUAL ENVIRONMENT
Building a Robust Virtual Environment
• Plan like you would for a physical
implementation.
– Build redundancy into your servers, storage
infrastructure and network infrastructure
– Separate your Capture and Retention Data
• Capture Data (DB, Exchange Mailboxes, etc.) should be on
faster SAS/FC Disks
• Retention Data (O/S, Backups, Applications) can be loaded
to slower SATA Disks
Building a Robust Virtual Environment
– Get State of the Art
• Multi-Core CPU Architecture is Everywhere!
• Enable Hardware Assisted Virtualization (Intel-VT & AMD-V)
• Maximize your Memory Investment – Memory is Cheap!
– FC, iSCSI or NFS Storage – It Doesn’t Matter
• Basic VMDK Traffic requires low latency, not high bandwidth
• For Large Deployments you can utilize NFS and iSCSI
– If your worried about throughput for large data loads,
10GbE is becoming more cost effective
– Verify your Service Console Settings
• Allocate Maximum RAM to the Service Console (ESX only) –
800MB
• Enable NTP and make sure its in sync
• Make sure DNS is functioning (forward and reverse lookups)
• Make Your Service Console network is redundant
– vCenter Services and HA rely on your SC being connected
Building a Robust Virtual Environment
• VMware Tools
– For all the great vCPU Co-Scheduling and Memory
Sharing Capabilities built into ESX, VMware Tools on
your Virtual Machines MUST always be Up To Date!
• Common CPU
Performance Issue
– Caused by vCPU
Over-Subscription
• When a vCPU needs to be scheduled,
the VMkernel maps a vCPU to a
“hardware execution context.”
H.E.C.
• A hardware execution context is a
processor’s capability to schedule one
thread of execution.
H.E.C.
H.E.C.
– A core or a hyperthread
– VMkernel load balances
• All the vCPUs in a VM must be
simultaneously scheduled.
• Check CPU %RDY on your VM to see if
it is waiting for a physical core or
hyperthread.
H.E.C.
H.E.C.
H.E.C.
H.E.C.
• Resolving CPU Performance Issues
– Caused by vCPU Under-Utilization
• Virtual Machines assigned 2 or 4 vCPUs but are not
actually using them.
• In this scenario you’re basically wasting potential CPU
cycles for other Virtual Machines that may need them.
• Check CPU %WAIT to see if your Virtual Machines vCPUs
are just sitting there doing nothing.
• High CPU %WAIT with Low CPU Used means you’ve
assigned too many vCPUs. Reduce to relieve possible
contention.
– Add additional cores
• Common Memory Performance Issues
– Caused by Memory Over-Commit
• The VMkernel and balloon driver (vmmemctl) do a great
job delivering and controlling memory to Virtual Machines
• Check Swap Used for your Virtual Machine in vCenter or
ESXTOP
• Lower Limits & Reservations on non-critical
Virtual Machines
• Increase Physical RAM
– Caused by Low VM Memory Assignment
• Check your Guest O/S Swap to make sure it has low
utilization.
• Check Memory Consumed in vCenter and ESXTOP
• Increase Virtual RAM
• Raise Limits and Reservation on Virtual Machine
Disk
• Resolving I/O Performance Issues
– Separate VMDKs based on performance needs
• High I/O Virtual Machines with Low IOps HDD do not mix
• More Spindles equals better performance
– iSCSI and NFS Based Datastores
• Utilize Jumbo Frames, LACP 802.11ad Link Aggregation
with multiple Targets (iSCSI) or Exports (NFS)
• Watch Network Throughput on ESX servers and also
Storage Array
– FCP
• Watch Latency Counters in vCenter
Network
• Virtual Network Design – Best Practices
– Build Redundancy into your vSwitches
• VMware HA is dependant on your Service Console
and it’s network gateway
– Segment VMotion Traffic
• To optimize available NIC ports, utilize the standby NIC port on
the vSwitch containing your Service Console
– Verify all physical ports, their associated VLANs and
utilize host profiles
– Utilize 802.1q VLAN tagging to maximize utilization of
your available physical NICs
– VMware Fault Tolerance will require its own
independent network port for vLockstep traffic.
– Monitor Network Throughput in vCenter or ESXTOP
VMware VMDK Hint
– I/O Intense Virtual Machines will benefit when their
Starting Partition is divisible by 4096.
– Misalignment can result in degraded performance. The
recommended starting value is 32768, typical VMs default
setting is 32256.
– The best option is to fix your template:
•
•
•
•
•
Prior to O/S installation boot with a WinPE CD
Run diskpart
Select Disk 0
Create Partition Primary Align=32
Reboot and Install your O/S as normal
•Vizioncore Has a Utility called vOptimizer Pro to
automate This with no downtime!!! – Even if an O/S is
already installed!
THANK YOU – QUESTIONS?