© 2009 VMware Inc. All rights reserved vSphere Performance Best Practices Rob Moran Premier Services Engineer – VMware Global Support Services – Cork,

© 2009 VMware Inc. All rights reserved

vSphere Performance Best Practices

Rob Moran

Premier Services Engineer – VMware Global Support Services – Cork, Ireland

2

Global Support Services and Customer Advocacy

Bangalore, India

Tokyo, Japan

Cork, IrelandBurlington, Canada

Palo Alto, CA Broomfield, CO

Support offices

Local language supportSpanish, Portuguese, French, German, Japanese, Chinese

Global Coverage24x7, 365 days/year

6 Support Centers

1000+ Support Engineers

Follow-the-sunSupport for

Severity 1 Issues

Support Relationships with 100% of the

Fortune 100; 99% of Fortune 500

3

Customer Support Day Events

Coming to a location near you: sharing of VMware best practices!

• Support Days are a collaboration between VMware Support, Sales and customers – you learn directly from the experts

• Topics are driven by customer input, and typically include:

• Best practices

• Tips/tricks

• Top issues

• Product roadmaps/demos

• Certification offerings

http://www.vmware.com/go/supportdays

4

Overview

What a performance problem sounds like:

• “My VM is running slow and I don’t know what to do!”

• “I tried adding more memory and CPUs but the problem got worse!”`

• “My VM is slow on one host but fast on another!”

What to look for? Where to start?

We will explore some of the most common performance-related issues that our support centers receive cases for

5

A word about performance….

Troubleshooting methodology must define:

• How to find root cause

• How to fix the problem

Must answer these questions:

1. How do we know when we are done?

2. Where do we start looking for problems?

3. How do we know what to look for to identify a problem?

4. How do we find the root-cause of a problem we have identified?

5. What do we change to fix the root-cause?

6. Where do we look next if no problem is found?

6

Agenda

Benchmarking & Tools

Best Practices and Troubleshooting

The 4 “food groups”

• Memory

• CPU

• Storage

• Network

© 2009 VMware Inc. All rights reserved

BENCHMARKING & TOOLS

8

Benchmarking

Consistent and reproducible results

Important to have base level of acceptable performance

• Expectation vs. Acceptable

Determine baseline of performance prior to deployment

• Benchmark on a physical system if applicable

Avoid subjective metrics, stay quantitative

• “The system seems slower”

• “This worked better last year”

9

Benchmarking

Benchmarking should be done at the application layer

• Use application-specific benchmarking tools and load generators

• Check with the application vendor

Isolate variables, benchmark optimum situation before introducing load

Understand dependencies

• Human interaction

• Other “food groups”

• Compare apples-to-apples

10

Aggregates thousands of metrics into Workload, Capacity, Health scores

Self-learns “normal” conditions using patented analytics

Smart alerts of impending performance and capacity degradation

Identifies potential performance problems before they start

Slide 10

Tools – vCenter Operations

11

Tools – vCenter OperationsSlide 11

12

Tools – esxtop

Valuable tool built in to vSphere hosts

View or capture real-time data

• View or playback data later

• Import data in 3rd party tools

vSphere Client performance graphs get their data from the kernel and VSI

• Presentation/unit may be different (e.g. %RDY)

© 2009 VMware Inc. All rights reserved

MEMORY

14

Memory – Overhead

A VM’s RAM is not necessarily machine RAM

• vRAM + overhead = maximum machine RAM

Source: vSphere 5.1 Resource Management Guide• Note: These are estimated values

15

Memory – Transparent Page Sharing

16

Memory – Host Memory Management

Occurs when memory is under contention

Ballooning

Compression

Swapping

17

Memory – Ballooning

18

Memory – Compression

19

Memory – Swapping

20

Memory – Swapping

21

Memory – VM Resource Allocation

22

Memory – Resource Pool Allocation

23

Memory – Ballooning vs. Swapping

Ballooning is better than swapping

Guest can surrender unused/free pages

Guest chooses what to swap, can avoid swapping “hot” pages

24

Memory – Rightsizing

Generally it is better to OVER-commit than UNDER-commit

If the running VMs are consuming too much host/pool memory…

• Some VMs may not get physical memory

• Ballooning or host swapping

• Higher disk IO

• All VMs slow down

25

Memory – Rightsizing

If a VM has too little vRAM…

• Applications suffer from lack of RAM

• The guest OS swaps

• Increased disk traffic, thrashing

• SAN slow down as a result of increased disk traffic

If a VM has too much vRAM…

• Higher overhead memory

• Possible decreased failover capacity

• Longer vMotion time

• Larger VSWP file

• Wasted resources

26

Memory – Troubleshooting

Wrong resource allocation May not notice a limit, e.g. VM or template with a limit gets cloned

Custom share values

Ballooning or swapping at the host level

• Ballooning is a warning sign, not a problem

• Swapping is a performance issue if seen over an extended period

Swapping/paging at the guest level

• Under-provisioned guest memory

Missing balloon driver (Tools)

27

Memory – Best Practices

Avoid high active host memory over-commitment

• No host swapping occurs when total memory demand is less than the physical memory (Assuming no limits)

Right-size guest memory

• Avoid guest OS swapping

Ensure there is enough vRAM to cover demand peaks

Use a fully automated DRS cluster

• Use Resource Pools with High/Normal/Low shares

• Avoid using custom shares

© 2009 VMware Inc. All rights reserved

CPU

29

CPU – Overview

Raw processing power of a given host or VM

• Hosts provide CPU resources

• VMs and Resource Pools consume CPU resources

CPU cores/threads need to be shared between VMs

Fair scheduling vCPU time

• Hardware interrupts for a VM

• Parallel processing for SMP VMs

• I/O

30

CPU – esxtop

31

CPU – esxtop

Interpret the esxtop columns correctly

%RDY - The percentage of time a VM is ready to run, but no physical processor is ready to run it which may result in decreased performance

%USED – Physical CPU usage

%SYS – Percentage of time in the VMkernel

%RUN – Percentage of total scheduled time to run

%WAIT – Percentage of time in blocked or busy wait states

%IDLE – %WAIT- %IDLE can be used to estimate I/O wait time

32

CPU – Performance Overhead & Utilization

Different workloads have different overhead costs (%SYS) even for the same utilization (%USED)

CPU virtualization adds varying amounts of system overhead

• Direct execution vs. privileged execution

• Non-paravirtual adapters vs. emulated adaptors

• Virtual hardware (Interrupts!)

• Network and storage I/O

33

CPU – vSMP

Relaxed Co-Scheduling: vCPUs can run out-of-sync

Idle vCPUs incur a scheduling penalty

• configure only as many vCPUs as needed

• Imposes unnecessary scheduling constraints

Use Uniprocessor VMs for single-threaded applications

34

CPU– Scheduling

Over committing physical CPUs

VMkernel CPU Scheduler

35

CPU– Scheduling

Over committing physical CPUs

VMkernel CPU Scheduler

X X

36

CPU– Scheduling

Over committing physical CPUs

VMkernel CPU Scheduler

X XX X

37

CPU – Ready Time

The percentage of time that a vCPU is ready to execute, but waiting for physical CPU time

Does not necessarily indicate a problem

• Indicates possible CPU contention or limits

38

CPU – NUMA nodes

Non-Uniform Memory Access system architecture

Each node consists of CPU cores and memory

A CPU core in one NUMA node can access memory in another node, but at a small performance cost

NUMA node 1 NUMA node 2

39

CPU – Troubleshooting

vCPU to pCPU over allocation

• HyperThreading does not double CPU capacity!

Limits or too many reservations

• can create artificial limits.

Expecting the same consolidation ratios with different workloads

• Virtualizing “easy” systems first, then expanding to heavier systems

• Compare Apples to Apples

• Frequency, turbo, cache sizes, cache sharing, core count, instruction set…

40

CPU – Best Practices

Right-size vSMP VMs

Keep heavy-hitters separated

• Fully automated DRS should do this for you

• Use anti-affinity rules if necessary

Use a fully automated DRS cluster

• Test that vMotion works

• Use Resource Pools with High/Normal/Low shares

• Avoid using custom shares

© 2009 VMware Inc. All rights reserved

STORAGE

42

Storage – esxtop Counters

Different esxtop storage views

• Adapter (d)

• VM (v)

• Disk Device (u)

Key Fields:

• DAVG + KAVG = GAVG

• QUED/USD – Command Queue Depth

• CMDS/s – Commands Per Second

• MBREADS/s

• MBWRTN/s

43

Storage – Troubleshooting with esxtop

High DAVG: issue beyond the adapter

• bad/overloaded zoning, over utilized storage processors, too few platters in the RAID set, etc.

High KAVG: issue in the kernel storage stack

• Driver issue

• Full queue

Aborts: GAVG exceeding 5000 ms

• Command will be repeated, storage delay for the VM

44

Storage – Benchmarking with iometer

45

Storage – Storage I/O Control

Allows the use of Shares per VMDK

Throttling occurs when datastore reaches latency threshold

• Higher share VMDKs perform IO first

vCenter monitors latency across all hosts

• Not effective if datastore shared with other vCenters

46

Storage – Storage DRS

Datastore clusters

• Maintenance mode

• Anti-affinity rules

vCenter monitors for latency and disk space

• Migrate VMDKs for better performance or utilization

Not effective with automated tiering SANs

• Check HCL to confirm these features are compatible

47

Storage – Troubleshooting

Snapshots

Excessive traffic down one HBA / Switch / SP can cause latency

• Consider using Round Robin in conjunction with ALUA

• Always be paranoid when it comes to monitoring storage I/O

Consider your I/O patterns

• Peak time for storage IO?

• Virus scans, database maintenance, user logins

Always consult with array vendor

• They know the best practices for their array!

48

Storage – Best Practices

Use different tiers of storage for different VM workloads

• Slower storage for OS VMDKs

• Faster storage for databases or other high-IO applications

Use the Paravirtual SCSI adapter

• Reduced overhead, higher throughput

Use path balancing where possible, either through 3rd party plugins / Round Robin and ALUA, if supported.

Use Storage DRS with SIOC

• Balance for both free space and latency

• Simplified datastore management

© 2009 VMware Inc. All rights reserved

NETWORK

50

Network – Load Balancing

Load balancing defines which uplink is used

• Route based on Port ID

• Route based on IP hash

• Route based on MAC hash

• Route based on NIC load (Load Based Teaming)

Probability of high-bandwidth VMs being on the same physical NIC

Traffic will stay on elected uplink until an event occurs

• NIC link state change, adding/removing NIC from a team, beacon probe timeout…

51

Network – Troubleshooting

Check counters for NICs and VMs

• Network load imbalance

• 10 Gbps NICs can incur a significant CPU load when running at 100%

Ensure hardware supports TSO

• Use latest drivers and firmware for your NIC on the host

For multi-tier VM applications, use DRS affinity rules to keep VMs on same host

• Same vSwitch / VLAN, rules out physical network

If using Jumbo Frames, ensure it is enabled end-to-end

52

Network – Best Practices

Use the vmxnet3 virtual adapter

• Less CPU overhead

• 10 Gbps connection to vSwitch

Use the latest driver/firmware for the NICs on the host

Use network shares

• Requires Virtual Distributed Switch 4.1

Isolate vMotion and iSCSI traffic from regular VM traffic

• Separate vSwitches with dedicated NIC(s)

• Most applicable with Gigabit NICs

53

How to measure the network?

scp from/to ESXi host is not valid check!

With scp we will involve underlying storage on source and destination VM/host

CPU can affect the test, scp will encrypt/decrypt the network flow

Copy to ESXi host can give false result as the management interface has very limited resources

54

How to check network performance?

VM – VM on same ESXi host. This will exclude physical network problems

VM –VM on different ESXi host. This will involve physical NICs and switch as well

Physical – VM. Will also test physical devices but we can focus on one VM

Physical – Physical: this will give us some number about what to expect

Use iperf/jperf/netperf. Free tool for network test

55

Iperf

56

Iperf

Windows and Linux version

Will not use storage

We can use different option for test (UDP/TCP)

Automatically calculates bandwith

57

In conclusion…

58

Key Takeaways – Performance Best Practices

Understand your environment

• Hardware, storage, networking

• VMs & applications

Advanced configuration values do not need to be tweaked or modified

• In almost all situations

Use fully automated DRS

Use Paravirtual hardware

59

Important Links

60

Important Links