The document discusses maximizing database performance within a software-defined datacenter, highlighting key concepts, advantages, and challenges associated with database integration in VMware's vCloud environment. It outlines best practices for resource allocation, storage management, and ensuring quality of service, while also addressing common pitfalls. The importance of cloud performance management and the role of vCenter operations in monitoring and optimizing database interactions are emphasized.

1. Maximize Database Performance in Your Software-
Defined Datacenter
Mark Achtemichuk, VMware
Michael Webster, VMware
VAPP4683
#VAPP4683

2. 2
Introductions
 Michael Webster – websterm@vmware.com
• VCDX-066, vExpert 2012 - 2013, VMware
• Strategic Architect, APJ – Business Critical Apps
• Twitter: @vcdxnz001
• Blog: http://longwhiteclouds.com
 Mark Achtemichuk – machtemichuk@vmware.com
• VDCX-050, vExpert 2013, VMware
• Sr. Technical Marketing Architect, Performance Specialist
• Twitter: @vmMarkA
• Blog: http://virtualizationeh.ca

3. 3
Agenda
 Introduction
 Database Performance Basics
 Software-Defined Datacenter Concepts
 Resource Allocation Models
 Storage Abstraction and Acceleration
 Ensuring Quality of Service
 Cloud Performance Management
 Common Pitfalls, Gotchas and Roadblocks
 Summary

4. 4
Database Advantages in vCloud
 Faster Development and Testing Lifecycles
 Faster Defect Resolution
 eg: Passing vApp Links to Developers
 Concurrent Isolated Testing
 eg: Multiple Copies of Same vApp
 Increased Testing Validity
 eg: Reduced Time to Validate Environment
 Combined Testing of Functionality with Network Security

5. 5
Database Advantages in vCloud (Continued)
 No Need to Manually Manage Per VM Reservations
 Declarative Service Defined Architecture
 Portability Between Private, Semi-Private and Hybrid/Public Cloud
 Rapid Elasticity, Easily Burst Resources For Peak Workloads
 Multi-tiered Application Simulated in Single vApp
 Self-service Load Balancing
 vApps Automatically Monitored with vCenter Operations

6. 6
Database Challenges with vCloud
 3rd Party Vendor Backup and Data Protection Integration with
vCloud Director
 Site Recovery Manager Integration for vCloud Director
 No Support for RDM
 No Support for Shared VMDK’s
 No Support for Multi-writer flag
 All Virtual Disks on Same Tier of Storage
 All Virtual Disks Contained within Single Datastore (by default)
 No Control Over Number of Datastores
 No Control Over NUMA Node Size

7. 7
Database Performance Basics

8. 8
Maximum Scalability and Performance With vSphere 5.x%ofApplications
95% of Apps
Require
IOPS
Network
Memory
CPU
< 10,000
<2.4 Mb/s
< 4 GB at peak
1 to 2 CPUs
VMware
vSphere 4
300,000
30 Gb/s
256 GB per VM
8 VCPUs
VMware Inf.
100,000
9 Gb/s
16/64 GB per VM
4 VCPUs
VMware
vSphere 5.5
1mm+ per VM
>36Gb/s
1 TB per VM
64 VCPUs
ESX 2
7,000
.9 Gb/s
3.6 GB per VM
2 VCPUs
ESX 1
<5,000
<.5Gb/s
2 GB per VM
1 VCPUs
3.0/3.5

9. 9
CPU & Memory
 HW v8, v9 or v10?
 Rightsize vCPU config
 Respect NUMA, vNUMA
 vCPU’s are a single thread
 Memory is cheap performance, direct relationship to read IO
 Server Side Cache can reduce VM memory requirements

10. 10
Approx DB Memory Architecture (high level)
Client sessions /context
Main DB memory
Oracle – SGA
SQL Server – Buffer Pool
Operating System
Large DB memory and OS Cache => caches more data blocks => can minimize access to disk
So set memory reservation to protect this
VM Configured
Memory
OS File System Cache
VM Reserved
Memory

11. 11
Network
 Use vmxnet3
 Jumbo Frames
 Evaluate latency settings:
 http://www.vmware.com/resources/techresources/10220

12. 12
“Your database is an extension of
your storage”

13. 13
Storage
 Use top tier storage policy
 Watch fast-provisioning disk chain length
 Disk Alignment
 Use multiple vmdks on multiple vSCSI adapters
 Use multiple data files or log files per vCPU
 Queues are everywhere

14. 14
Typical Storage Sizing and Performance
SIOC Enabled when more than one VM share datastore - vCloud

15. 15
Oracle Storage Sizing and Performance

16. 16
Oracle Storage Sizing and Performance

17. 17
Simple Sample SQL Server Database & Storage Layout
Application
SQL Server
datafile1.mdf
datafile2.ndf
datafile3.ndf
datafile4.ndf
datafile5.ndf
datafile6.ndf
datafile7.ndf
datafile8.ndf
logfile.ldf
tempfile1.mdf tempfile2.ndf tempfile3.ndf tempfile4.ndf
LUN1 LUN2 LUN3 LUN4 LUN5
D: E: F: G: L:
LUN6
templog.ldf
T:
4 vCPU, 32 GB RAM (an example)
ESX Host
DataStore1 DataStore2 DataStore3 DataStore4 DataStore5 DataStore6
VMDK1 VMDK2 VMDK3 VMDK4 VMDK5 VMDK6
Can be mount points under a drive as well
Best Practices
• Enable traceflag 834 in SQL Server for Large Page Support and 2371 for dynamic automatic statistics update.
• Service account that SQL Server is running under must have the “Lock Pages in Memory” privilege set in Local Policy
Partition:NTFS–64kclustersize
PVSCSI1 PVSCSI3
Can be Shared under
one drive and VMDK
PVSCSI2

18. 18
Software-Defined Datacenter Concepts

19. 19
Site Recovery
Manager
VMware vSphere
VMware
vCenter Server
VMware
vCenter Server
Site Recovery
Manager
VMware vSphere
Site A (Primary) Site B (Recovery)
Servers Servers
SDDC - Automating Security, Compliance, DR
1 Configure primary site for security and
compliance services in the software-
defined data center.
Replication
Software-Defined
Data Center
Software-Defined
Data Center
2 Automated replication – not only
for workloads but also for
services
3 Test subset of DR workloads
and services
4 Test full DR scenario
Replication
Test
Test

20. 20
Resource Allocation Models

21. 21
Organization Virtual Datacenter – Allocation Models
 Reservation pool resources
allocated to the organization
virtual datacenter are
completely dedicated
 Allocation pool is a pool of
allocated resources with a
certain percentage guaranteed
 Pay As You Go model provides
the illusion of an unlimited
resource pool – Can be capped
Consumer #1
Guaranteed Resources
Consumer #2
Guaranteed Resources
Reservation = Limit
| | |
100% Reservation 100% Reservation
Consumer #1
Guaranteed Resources
Consumer #2
Guaranteed Resources
25% Overcommit
| | |
75% Reservation 75% Reservation
Burstable
Resources
| |
Consumer #1 Consumer #2
Unlimited Resource Pools
| | |
VM % Reservation VM % Reservation
VM VM VM VM VMVM VM
|

22. 22
Storage Abstraction and Acceleration

23. 23
Storage Profiles
• The organization administrator can
offer different classes of storage
within a single virtual datacenter
• Virtual datacenters, vApp
templates, and media can define a
default storage profile to be used
• VMware vSphere Storage DRS™
functionality can be used within
VCD, allowing for Storage vMotion
between datastores within a cluster
based upon metrics defined
(available space and I/O latency)
• Do not use I/O Load Balancing
when using Array Auto Tiering
• VM allocated only one tier of
storage
Gold
High Speed/Low
Latency
Silver
Moderate
Speed/Latency
Bronze
Low Speed/ High
Latency
Virtual
Datacenter 1

24. 24
Profile Driven Storage & vCloud Director
 Storage Profiles are now represented in vCloud Director
 This allows storage to be managed on a per ORG vDC
 If the Storage Profile associated with a vApp is changed, the vApps
are automatically Storage vMotioned to a compliant datastore

25. 25
Storage DRS & vCloud Director
 vCloud Director Interoperability/Support for Linked Clones
• vCloud Director will use Storage DRS for the initial placement of
linked clones during Fast Provisioning
• vCloud Director will use Storage DRS for managing space utilization
and I/O load balancing

26. 26
Increasing Database Performance in vCloud Director
 Queue depth
 Larger # of datastores
 Storage DRS Anti-Affinity Rules – Separate VMDK
 No Fast Provisioning for Production VM’s
 Server Side Caching

27. 27
Ensuring Quality of Service

28. 28
Enabling QOS
 CPU and Memory
 Use DRS for auto migration
 Use the right allocation model
 Size Application for VM Allocation Model
 Network
 Use NetIOC at lowest layer “carefully” if necessary
 What about Load Based Teaming? Not with VXLAN!
 Storage
 Use storage profiles appropriately
 Leverage array based hot block migration
 Storage IO Control
 Storage DRS

29. 29
Cloud Performance Management

30. 30
Slide 30
How vCenter Operations Delivers a New Model for Operations
 Tightly integrated with vSphere
 Self-learns “normal” conditions using
patented analytics
 Aggregates underlying metrics into
Workload, Capacity, Health scores
 Powerful visibility and drill down from
datacenter to component level
 Smart alerts of impending performance
degradation
 Integration of 3rd party monitoring tools
An integrated approach and patented analytics to transform
how IT ensures service levels in dynamic environments
!

31. 31
vSphere
vCOps High Level Architecture
OpenVPN
Postgres DB
vSphere
WebApp
Custom
WebApp
Admin
WebApp
vCenter Operations Manager vApp
UI VM
Rolled up
capacity
data
Capacity Analytics
FSDB
Postgres DB
Collector
ActiveMQ
Performance Analytics
Analytics VM
Metric
Data
vSphere
VMware Cloud / vCenter
vSphere
vC Ops Mgr vSphere UI
vCenter
Configuration
Manager
3rd Party Data
Sources
vCenter
Communications
over SSL
vC Ops Mgr Custom UI

32. 32
vCOps vCloud Mapping

33. 33
Oracle Database – Dashboard Example

34. 34
Common Pitfalls, Gotchas, Roadblocks

35. 35
Third Party Clustering in Virtual Machines
 For clustering with VMFS: Disabling simultaneous write protection provided
by VMFS using the multi-writer flag, shows how to share virtual disks –
http://kb.vmware.com/kb/1034165
 * Windows Server 2012 Failover Clustering – http://kb.vmware.com/kb/1037959
Cluster Software RDM VMFS /
vMotion
Supported by
Vendor
Microsoft Failover Clustering Yes* No Yes*
Symantec Veritas Cluster
Services
Yes Yes
(6.0.x)
Yes
SUSE Clustering (HAE) Yes Yes Yes
Red Hat Cluster Yes Yes Yes
(Red Hat 5.7)
Oracle Clusterware and RAC Yes Yes Yes
(RAC 11.2.0.2)

36. 36
Important Architecture Traps to Avoid
 Don’t disable DRS
 vMotion Slot Size and Host Sizing when using HA % Admission Control
Impacts Maintenance and HA Failovers
 Max # of LUN’s per Host / Cluster Impact on Size of Cluster
 Host NUMA Node Too Small for Optimal Performance and
Efficient Utilization
 NUMA Node Size Not Known in Cloud
 Size Hosts and Clusters to Allow Maintenance and Failure Resource,
Consider Sacrificing Optimal Performance During Failure

37. 37
Summary

38. 38
Questions

39. 39
Other VMware Activities Related to This Session
 HOL:
HOL-SDC-1304
vSphere Performance Optimization
HOL-SDC-1317
vCloud Suite Use Cases - Business Critical Applications

40. THANK YOU

42. Maximize Database Performance in Your Software-
Defined Datacenter
Mark Achtemichuk, VMware
Michael Webster, VMware
VAPP4683
#VAPP4683