0% found this document useful (0 votes)
50 views24 pages

ASE On VMWare Vshpere

asd

Uploaded by

lyrics88
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
50 views24 pages

ASE On VMWare Vshpere

asd

Uploaded by

lyrics88
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

SAP® Sybase® Adaptive Server Enterprise

on VMware vSphere®
Essential Deployment Tips

Bob Goldsand
Partner Architect,
Global Strategic Alliances VMware
Chris Brown
Solution Advisor Expert SAP
White paper
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Table of contents Introduction

With the trend towards large logical CPU and high memory
2 Introduction
systems at cost-effective prices, more and more SAP
3 VMware vSphere
4 VMware vMotion and Storage vMotion customers are looking to deploy their production databases

6 VMware Distributed Resource Scheduler on VMware vSphere®. However, there is a perception


and Storage DRS
that mission-critical, high-performance online transaction
8 SAP Sybase Adaptive Server Enterprise processing databases suffer performance penalties
8 General Purpose Computing Environment
when virtualized.
9 Virtual Machine BIOS Settings

9 Virtual Machine Guest Operating System VMware, in cooperation with SAP, conducted several
10 CPU Considerations extensive tests to compare the performance of SAP®
12 Resource Management Sybase® Adaptive Server® Enterprise (SAP Sybase ASE),
13 Networking
version 15.7 running on VMware vSphere 5.0®, versus
14 Storage
directly on native hardware. The results showed that
15 Multiple Virtual SCSI Controllers
throughput on the high-transaction OLTP database
15 Sybase Direct I/O Parameter
SAP Sybase ASE, running in a virtual machine,
16 EMC VMAX FAST VP
often matched that of an identical physical system.
16 FAST VP Performance Measurement and Data Movement

16 FASTVP and VMware Storage DRS Once the performance myths are dispelled, the next
17 Storage DRS Manual Mode and I/O Metric Inclusion questions are typically centered around support, such
18 Deployment Tips/Guidelines Summary
as “Is the database supported once it is virtualized?”
19 Conclusion and “Do I have to reproduce issues on physical hardware?”
20 Appendix A: Test Results − ASE Native versus vSphere SAP fully supports and certifies SAP Sybase ASE running
21 Resources
on VMware vSphere, and customers will not be required
22 Acknowledgements
to reproduce issues on native hardware.

2
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

• Consolidation. VMware virtualization allows multiple application


With the performance and support barriers removed,
servers to be consolidated onto one physical server, with little
database administrators can focus on delivering ROI to
or no decrease in overall performance. This helps minimize
the enterprise by leveraging the benefits of running SAP or eliminate underutilized server hardware, software,
and infrastructure.
Sybase ASE on VMware vSphere.
• Manageability. The live migration of virtual machines from
This paper describes the advantages and technical server to server and its associated storage is performed with
no downtime using VMware vSphere® vMotion® and VMware
considerations for running SAP Sybase ASE on the
vSphere® Storage vMotion®.
VMware vSphere platform. This document is written

for architects and engineers who are responsible for Figure 1. VMware vSphere

SAP Sybase ASE and VMware virtualized environments.

This paper assumes that the reader has basic knowledge

of SAP database products and the VMware vSphere

virtualization platform.

VMware vSphere
VMware vSphere is an optimal virtualization platform and enabler
for cloud computing architectures. vSphere enables IT to meet
service-level agreements (SLAs) for the most demanding business-
critical applications at the lowest total cost of ownership (TCO).
VMware vSphere delivers control over all IT resources with the
highest efficiency and choice in the industry, as shown in Figure 1.

vSphere transforms industry-standard hardware into a shared,


mainframe-like resilient environment with built-in service-level
controls for all applications. vSphere virtualization and cloud
computing solutions provide for:

3
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

• Automation. VMware automated load balancing takes advantage In addition, VMware vSphere 5.1 enables a virtual machine to
of vMotion and Storage vMotion to migrate virtual machines change its datastore and host simultaneously, even when the hosts
among a set of VMware® ESXi™ hosts. VMware vSphere® do not have a shared storage environment. It allows virtual machine
Distributed Resource Scheduler (DRS) and VMware vSphere
™ ®
migration between clusters in a larger datacenter that might not
Storage DRS™ enable automatic resource relocation and have a common set of datastores between them. vSphere also
optimization decisions for virtual machines and storage. allows virtual machine migration in small environments without

• Availability. High availability can be enabled to reduce requiring access to expensive, shared storage equipment, as

unplanned downtime and support higher service levels for shown below.

applications. VMware vSphere® High Availability (HA) ensures Figure 2. VMware vMotion and Storage vMotion
that, in the event of an unplanned hardware failure, the
affected virtual machines are automatically restarted on
another host in a VMware cluster.

For more information about vSphere, see the “Resources”


section later in this paper.

In addition, VMware vSphere advanced features increase avail-


ability and automation and improve the manageability of SAP
Sybase ASE. The advanced features used in the SAP Sybase
ASE solution described in this paper include VMware vMotion,
Storage vMotion, Distributed Resource Scheduler (DRS),
and Storage DRS. Each of these features is described in
the sections below.

VMware vMotion and Storage vMotion


Together, VMware vMotion and Storage vMotion provide for the
complete virtualization of servers, storage, and networking in vSphere 5.1 allows vMotion and Storage vMotion to be combined
order to move an entire virtual machine running SAP Sybase into one process. This combined migration copies both the virtual
ASE instantaneously, from one physical server to another. The machine memory and its disk over the network to the destination
virtual machine retains its network identity and connections, host. After the memory and disk data are sent over, the destination
ensuring a seamless migration process. virtual machine resumes and the source virtual machine is
powered off.

4
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Each of these vSphere advanced features is described in the VMware vMotion:


sections below.
• Moves entire running virtual machines instantly. Performs live
migrations with zero downtime, undetectable to the user.
VMware vMotion
• Manages and schedules live migrations with ease at predefined
VMware vMotion enables the live migration of virtual machines
times without an administrator’s presence, with the reliability and
running SAP Sybase ASE from one physical server to another
manageability that is derived from a production-proven product.
with zero downtime, continuous service availability, and complete
transaction integrity. vMotion is a key enabling technology for • Performs multiple concurrent migrations of a virtual machine
creating the dynamic, automated, and self-optimizing data center. running any operating system, across any type of hardware
This capability makes hardware maintenance possible at any and storage that is supported by vSphere, complete with an
time, and vMotion does not require clustering or redundant audit trail.
servers, as shown below.
• Moves online workloads from one ESXi™ Server host machine to
another in order to maintain service levels and performance goals.
Figure 3. VMware vMotion
• Continuously and automatically optimizes virtual machine
placement within resource pools. Proactively moves virtual
machines away from failing or underperforming servers.

• Performs hardware maintenance without the need to schedule


downtime and disrupt business operations.

5
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

VMware Storage vMotion • Migrates virtual machines running any supported operating
Storage vMotion enables the live migration of virtual machine system on any supported server hardware based on usage
disk files within and across storage arrays, and it provides for and priority policies for tiered storage.
automated storage management. Storage vMotion relocates • Performs live migration of virtual machine disk files across
disk files of virtual machines running on SAP Sybase ASE,
any Fibre Channel, Internet Small Computer System Interface
while maintaining continuous service availability and complete
(iSCSI), Fibre Channel over Ethernet (FCoE), and Network File
transaction integrity, as shown below.
System (NFS) storage system supported by vSphere.

• Dynamically optimizes storage I/O performance by moving


Figure 4. VMware Storage vMotion
virtual machine disk files to alternative LUNs that deliver the
required performance without scheduled downtime.

• Eliminates over-allocation of storage resources in order to


proactively deal with I/O bottlenecks.

• Efficiently manages storage capacity and utilization by reclaiming

unused or “stranded” storage capacity and allocating it to other


virtual machines.

• Moves virtual machines with higher performance needs or critical

workloads to larger capacity storage LUNs as virtual machine


disk files approach their total available LUN size limits.

VMware Distributed Resource Scheduler (DRS)


and Storage DRS
VMware Distributed Resource Scheduler aligns resource usage
with business priority by enabling automated load balancing
vSphere Storage vMotion: across hosts, and optimizes power consumption by turning off
hosts during lower load periods. Similarly, VMware Storage DRS
• Enables the live, automated migration of virtual machine disk
enables ongoing load balancing between datastores to optimize
files from existing storage to different classes of tiered storage
the use of storage resources, and to avoid bottlenecks. Storage
with zero downtime.
DRS provides smart virtual machine placement and load-balancing
mechanisms based on I/O and space utilization.

6
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Each of these vSphere advanced features is described in the VMware Storage DRS
sections below. Storage DRS aggregates the resources of several datastores into
a single datastore cluster to simplify storage management at scale.
VMware Distributed Resource Scheduler
During virtual machine provisioning, Storage DRS ensures that
DRS is an automated load-balancing technology that takes
virtual machine data is assigned the optimal storage location
advantage of vMotion to migrate virtual machines among a set
initially, and it provides intelligent virtual machine placement
of ESXi™ hosts. DRS continuously monitors utilization across
based on the I/O load and available storage capacity. Storage
vSphere servers and intelligently allocates available resources
DRS performs ongoing load balancing between datastores to
among virtual machines according to business needs. DRS
align the storage resources with predefined rules in order to
dynamically aligns resources with business priorities, balances
meet business objectives, as shown below.
computing capacity, and reduces power consumption in the
datacenter, as shown below. Figure 6. VMware Storage DRS

DRS enables automatic initial virtual machine placement on


any of the hosts in the cluster. It also makes automatic resource
relocation and optimization decisions as hosts and virtual machines
are added or removed from the cluster. When DRS is configured
for manual control, it makes recommendations for review and
later implementation only; there is no automated activity.

Figure 5. VMware Distributed Resource Scheduler

vSphere Storage DRS:

• Continuously monitors storage space and I/O utilization across


a preassigned pool of datastores and intelligently aligns storage
resources with business growth objectives.

• Specifies how storage resources are allocated to virtual


machines with rules and policies.

• Assigns dedicated storage infrastructure to business units,


while achieving higher storage utilization through pools of
storage volumes.

7
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

• Empowers business units to build and manage virtual architectures. With the advent of cores and threads, the industry
machines within their storage, while giving central IT has reached a point to where it’s no longer a “chip clock-speed”
control over storage resources. race, but a “core and thread” race.

• Balances storage space usage and storage I/O load, avoids


Previously, SAP Sybase ASE was extremely efficient at utilizing
resource bottlenecks to meet application service levels, and
faster processors, but it did not take full advantage of the additional
increases the manageability of storage at scale.
cores and threads being implemented on Intel and RISC-based
• Easily deploys additional storage capacity and seamlessly processors commonly in use today. With an increasing number of
takes advantage of additional capacity when new storage customers deploying virtualized Intel and RISC-based environments,
is added to a pool of storage volumes. this architectural limitation has become more apparent.

• Maintains data stores and improves service levels for


For this reason, VMware and SAP implemented the threaded-kernel
all applications.
option of SAP Sybase ASE while running an industry-standard
online transaction processing (OLTP) benchmark on both bare
SAP Sybase Adaptive Server Enterprise
metal and virtualized hardware. In each case, the transactional
SAP Sybase ASE has a long history of delivering industry-leading
throughput was measured and reported within SAP Sybase ASE
support for high-performance, high-throughput, high-concurrency
by calculating the number of begin/commit pairs over a given
systems. SAP Sybase ASE, version 15.7 represents the latest
period of time. From the operating system perspective, the CPU
generation of the flagship database management system from
load (user/system/idle) and I/O throughput was measured using
Sybase, an SAP company. Its architecture is designed to perform
industry-standard tools such as iostat.
and scale extremely well under varying workloads.
In addition, the threaded-kernel model is the preferred kernel
SAP Sybase ASE, version 15.7:
method going forward, especially for running SAP Business

• Is
Suite on Sybase ASE.
the first version of SAP Sybase ASE to be certified with
SAP Business Suite software.

• Implements a threaded, lightweight kernel that utilizes


General Purpose Computing Environment
When deployed on the vSphere platform, Sybase ASE can be
operating-system resources very efficiently – thereby
used to create optimized, purpose-built computing environments.
improving performance, scalability, and reliability.
The first step in creating this type of environment requires a
A major design principle of the SAP Sybase ASE threaded
careful examination of the BIOS settings, disabling of unnecessary
kernel is to take advantage of modern, multithreaded processor

8
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

processes and peripherals, and the compilation of a monolithic later releases. Plan the operating system installation to ensure
kernel to direct the critical compute resources (CPU, memory, that it takes advantage of virtualization and creates an optimized
network, and I/O) to the databases. computing environment.

Each of these items is described in the sections below. Simply disabling the peripheral components in the BIOS does
not guarantee that these components are fully disabled. It is
also important to ensure that the peripheral components are not
Virtual Machine BIOS Settings
re-enabled during the operating system installation. In addition,
Servers with Intel® micro-architecture (Nehalem class) including
do not install office productivity suites, graphics, sound, video
the Intel® Xeon® 5500 series and later CPUs, offers two other
programs, and so on.
power management options: C-states and Intel Turbo Boost.
Leaving C-states enabled can increase memory latency. This
After the installation, disable unnecessary foreground and back-
option is therefore not recommended for low-latency workloads.
ground processes.

Enable these settings Disable these settings


For example, Linux processes include:

•• Virtualization technology •• Node interleaving


• anacron, apmd, atd, autofs, cups, cupsconfig, gpm, isdn,
•• Turbo mode •• C1E Halt state
•• Hardware based •• Power saving iptables, kudzu, netfs, and portmap
virtualization support •• Unused features: Video
•• Hyperthreading (workload BIOS Shadowable, Video Examples of Windows® processes include:
dependent) RAM Cacheable, on-board
•• Wake on LAN (required for audio, on-board modem,
• Alerter, Automatic Updates, Clipbook Viewer, Error Reporting,
VMware vSphere® Distributed on-board serial ports,
Power Management™) on-board parallel ports, Help and Support, Indexing, Messenger, NetMeeting, Remote
on-board game port, floppy
•• Execute Disable (required Desktop, and System Restore Services
drive, CD-ROM, USB
for vMotion and DRS)
•• Static high performance/ The operating system settings and configurations below are
balanced
recommended to run Sybase ASE in a virtualization environment:

Table 1. Workload used • In the /etc/sysctl.conf file:


kernel.shmmax = <database dependent size value>
Virtual Machine Guest Operating System vm.swappiness = 10
In order to minimize the virtualization overhead, move the virtual vm.nr_hugepages = <database dependent size value>
machines to modern guest operating systems, such as SUSE®
• Add the ‘sybase’ user to /etc/security/limits.conf in order to utilize
Linux Enterprise Server (SLES) 11 Service Pack 1, Red Hat®
HugePages on the system (by default only the ‘root’ user can
Enterprise Linux® (RHEL) 6, Windows Server® 2008 R2, or
use HugePages).

9
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

• For ease of management use filesystem-based devices (not Virtual CPUs


raw partitions). For example, RedHat’s High Performance When configuring the Sybase ASE database on virtual machines
Filesystem Option (XFS). for production environments, ensure that the total vCPU resources

• Change the I/O scheduler: for the virtual machines running on the system do not exceed the

• To ‘noop’ from the default of ‘cfq’. Sybase ASE performs better CPU capacity of the host. It is good practice to under-commit

in a virtual environment. CPU resources on the host. If the host CPU capacity is overloaded,
the performance of the virtual database might degrade.
• Use a 4K page size, since the Linux operating system utilizes
a 4k block size for I/O operations by default.
Do not over commit CPU resources in a production environment.
For additional operating system and version settings and A reasonable ceiling should be 80 percent of consumption; set
dependencies, go to the Sybase ASE release notes at: 90 percent consumption as the limit for generating an alert to
http://www.sybase.com/support/techdocs the virtual infrastructure administrator.

See the “Resources” section later in this paper for more information. A multithreaded application such as Sybase ASE can benefit from
using multiple CPUs. However, for latency-sensitive applications,
do not over commit virtual CPUs compared with the number of
CPU Considerations
physical CPUs. In order to ensure that the physical CPUs are
For the best performance of latency-sensitive applications in
not oversubscribed, limit the total number of vCPUs to the total
guest operating systems, schedule all virtual CPUs (vCPUs) on
number of physical CPUs, minus 1.
the same Non-Uniform Memory Access (NUMA) node. Fit and
allocate all virtual machine memory from the local physical
Configuring virtual Sybase ASE database with excess vCPUs
memory attached to the NUMA node. Set processor and
can impose a small resource requirement on vSphere because
memory affinity as described below.
unused vCPUs continue to consume timer interrupts. vSphere
co-schedules virtual machine vCPUs and attempts to run the
When the processor affinity for vCPUs is scheduled on a specific
vCPUs in parallel to the extent possible. Unused vCPUs impose
NUMA node, set the processor affinity using the vSphere client
scheduling constraints on the vCPU being used and can
as shown in the table below.
degrade its performance.
Processor Affinity Memory Affinity
Workload CPU Optimization - CPU Scheduling
For the virtual machine For the virtual machine
settings, go to : settings, go to : In some use cases, the application owner may want to ensure
•• Resources tab •• Resources tab that the VMkernel does not de-schedule the virtual machine
•• Advanced CPU •• Advanced CPU when the vCPU is idle. In order to determine that there is opti-
•• Check the Scheduling •• Check the NUMA Memory
Affinity box Affinity box mal database performance, add the configuration below:

Table 2. Workload used

10
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

• Go to Virtual Machine Settings in vCenter Server, Options tab, • %RDY – The percentage of time the world was ready to run,
Advanced General, Configuration Parameters, and add: but is not scheduled to a core. A world in a run queue is wait-
• monitor_control.halt_desched (with the value of False) ing for the CPU scheduler to let it run on a PCPU. When
%RDY is greater than 10 percent, it could be an indication of
Performance is normally monitored through vSphere vCenter.
resource contention.
However, it is a best practice to periodically collect additional
• %CSTP – The percentage of time the world is stopped from
statistical measures of the host CPU usage. This can be done
running to allow other vCPUs in the virtual machine to catch
through the vSphere Client, or by using esxtop or resxtop. CPU
up, co-deschedule state. If %CSTP is greater than 5 percent,
usage tips are listed below.
this usually means the virtual machine workload is not using
Work with your VMware administrator to interpret esxtop data: VCPUs in a balanced fashion.

By using esxtop, DBAs can gain additional performance insight


• If the load average listed on the first line of the esxtop CPU
with respect to CPU resource contention. DBAs should also
Panel is equal to, or greater than, the number of physical
work with their VMware administrator to fully understand and
processors in the system, this indicates that the system
interpret esxtop statistics.
is overloaded.

• The usage percentage of physical CPUs on the PCPU line can Memory Considerations
be another indication of a possibly overloaded condition. For memory, the best practice is to configure memory reservations
equal to the size of the Sybase ASE max memory setting. Do not
In general, 80 percent usage is a reasonable ceiling in production
over commit memory in a production environment. When con-
environments. 90 percent usage should be used as an alert to
solidating Sybase ASE database instances, vSphere presents
the VMware administrator that the CPUs are approaching an
the opportunity for sharing memory across all virtual machines
overloaded condition that needs to be addressed. However,
that may be running the same operating systems, applications,
decisions concerning usage levels are typically made based
or components. In this case, vSphere uses a proprietary, trans-
on the criticality of the Sybase database being virtualized,
parent page-sharing technique to reclaim memory that allows
regarding the desired load percentage.
databases to run with less memory than physical. Transparent
When using esxtop, three critical statistics to interpret are: page sharing can also be used to over- commit memory, without
any performance degradation.
• %RUN – The percentage of total time the “world”1 is running
on the processor. If %RUN is high, it does not necessarily Hardware-Assisted Memory Virtualization
mean that the virtual machine is resource-constrained. (See Some recent processors include a feature that addresses over-
description of %RDY below.) head due to memory management unit (MMU) virtualization by
providing hardware support to virtualize the MMU. Without
1 Esxtop uses “worlds” and “groups” as the entities to show CPU usage.
A “world” is an ESX VMkernel schedulable entity, similar to a process
or thread in other operating systems. A “group” contains multiple worlds.

11
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

hardware-assisted MMU virtualization, VMware ESXi™ maintains Large (Huge) Memory Pages
“shadow page tables” that directly map guest virtual memory to Large (huge) pages can potentially increase TLB access efficiency,
host physical memory addresses. thereby improving database performance. The use of large pages
can significantly improve the performance of SAP Sybase ASE
These shadow page tables are maintained for use by the processor
databases on vSphere, compared to running the workload using
and are kept consistent with the guest page tables. This allows
small pages. Large-page support is enabled by default in
ordinary memory references to execute without additional overhead,
VMware ESX versions 3.5 and later. The use of Large Pages
since the hardware translation lookaside buffer (TLB) caches
is supported for SAP Sybase ASE version 15.0.3 and greater.
direct guest virtual memory to host physical memory address
translations that are read from the shadow page tables. However, Using HugePages at the operating system level enables SAP
extra work is required to maintain the shadow page tables. Sybase ASE to achieve approximately a 10 percent performance
improvement. It is recommended for any 64-bit production system.
Hardware assistance eliminates software memory virtualization
Furthermore, in larger memory configurations (that is, configurations
overhead. In particular, hardware assistance eliminates the
where SAP Sybase ASE utilizes more than 256 GB of RAM,
overhead required to keep shadow page tables in synchronization
HugePages are required in order for SAP Sybase ASE to boot.
with guest page tables, although the TLB miss latency is significantly
higher. This means that hardware assistance provides workload
benefits depending primarily on the memory virtualization overhead Resource Management

that is caused when using software memory virtualization. When using vSphere advanced workload management features
such as VMware vMotion and VMware DRS, the database is freed
If a workload involves a small amount of page table activity, from the resource limitations of a single host. With vMotion, virtual
such as process creation, mapping the memory, or context machines running SAP Sybase ASE can be moved from one
switches, software virtualization does not cause significant physical vSphere to another to balance the available resources.
overhead. However, workloads having a large amount of page DRS can dynamically allocate and balance computing resources
table activity, such as workloads from a database, are likely to by continually monitoring the utilization of resource pools associated
benefit from hardware assistance. with virtual machines in a VMware cluster.

For these reasons, enable vSphere to Choose the Best Virtual Over-commitment is perfectly acceptable for SAP Sybase ASE in
Machine Monitor based on the CPU and Guest Operating nonproduction environments such as Dev, Test, QA, and others,
System Combination. where predictable and consistent performance is not as critical.

12
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Networking VMXNET 3 by default also supports an adaptive interrupt


For Sybase ASE, the networking configuration includes virtual coalescing algorithm for the same reason that physical NICs
distributed switch, VNXNET3, and network interface card (NIC) implement interrupt coalescing. Virtual interrupt coalescing
optimization. Each of these components is described in the helps drive high throughput to the virtual machines with multi-
sections below. ple vCPUs with parallelized workloads such as multiple threads.

Virtual Distributed Switch NIC Optimization


For a database-as-a-service with SAP Sybase ASE and complex Most NICs also provide a mechanism, usually via the ethtool
infrastructure implementation, the best practice is to use a Virtual command and/or module parameters, to disable interrupt
Distributed Switch (vDS). Prior to vSphere 4.0 virtual infrastructure coalescing. In order to determine if disabling the physical
administrators configured standard switches on each host. The vDS NIC interrupt moderation on the VMware® ESXi™ host is
simplifies virtual machine networking and enables virtual machine needed, perform the following command:
networking that spans multiple vSphere hosts in a data center
# esxcli system module parameters set -m ixgbe -p “InterruptThrottleRate=0”
to be managed as a single virtual switch from a centralized
VMware vCenter Server through vSphere Client. This example applies to the Intel®10 GbE driver called ixgbe. In
order to find the appropriate module parameter for the NIC, first
When a host is added, networking for that host does not require
find the driver using the ESXi command:
configuration. Instead, the host is added to a defined port
# esxcli network nic list
group that is dedicated to Sybase ASE traffic or other specific
application traffic.
Then find the list of module parameters for the driver used:

In addition to supporting Private VLANs (PVLANs), the vDS also # esxcli system module parameters list -m <driver>

provides the ability to shape both inbound and outbound network


Large Receive Offload (LRO) is another feature of VMXNET 3
traffic. VMware Standard Switches can be easily migrated to vDS
that helps deliver high throughput with lower CPU utilization.
in a nondisruptive manner with the vCenter Server management
LRO aggregates multiple received TCP segments into a larger
user interface.
TCP segment before delivering it to the guest TCP stack.
VNXNET 3 However, for latency-sensitive applications that rely on TCP, the
The best practice is to use VMWare VMXNET 3 virtual NICs time spent aggregating smaller TCP segments into a larger one
for virtual machines running the SAP Sybase ASE database. adds to the latency. It can affect TCP algorithms such as TCP
VMXNET 3 is the latest generation of paravirtualized NICs delayed acknowledgment (ACK) that cause the TCP stack to
that are designed from the ground up for performance. delay an ACK until the two larger TCP segments are received.
This also adds to the end-to-end latency of the application.

13
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

In order to determine if disabling LRO benefits application stack applications and is analogous to provisioning dedicated LUNs
requirements, reload the vmxnet3 driver in the Linux guest: in the physical world. However virtualized customers can use

# modprobe -r vmxnet3
vSphere features such as VMware vSphere® Storage I/O Control,
Storage vMotion, and Storage DRS to prioritize and manage
Add the following line in /etc/modprobe.conf data store workloads. For production SAP Sybase ASE data-
(Linux-version dependent): bases, use dedicated data stores. For nonproduction or less
demanding environments, use consolidated data stores.
options vmxnet3 disable_lro=1

Then reload the driver using:


VMFS
VMware vSphere® VMFS provides high-performance, clustered
# modprobe vmxnet3
storage virtualization that is optimized for virtual machines. With
VMFS, each virtual machine is encapsulated into a small set of
Storage files. VMFS is the default storage management interface used to
For Sybase ASE, the storage configuration includes datastores, access those files on physical SCSI disks and partitions. VMFS
Virtual Machine File System (VMFS), Paravirtualized SCSI allows multiple ESXi™ instances to access shared virtual
Adapters, and VMware Storage DRS. Each of these components machine storage concurrently. It also enables virtualization
is described in the sections below. based, distributed infrastructure services such as vMotion,
VMware DRS, and VMware High Availability to operate across
Data Stores
a cluster of ESXi™ hosts.
vSphere uses data stores to store virtual disks. Data stores
provide an abstraction of the storage layer that hides the In order to balance performance and manageability in a virtual
physical attributes of the storage devices to the virtual environment, it is an accepted best practice to deploy databases
machines. VMware administrators can create data stores using VMFS. Raw Device Mappings (RDMs) are sometimes
that can be used as a single consolidated pool of storage, erroneously selected to provide increased performance. In contrast,
or many data stores that can be used to isolate various the two dominant workloads associated with databases, random
application workloads. read/write and sequential writes, have nearly identical performance
throughput characteristics when deployed on VMFS or using RDM.
In traditional Storage Area Network (SAN) deployments, it is
a generally accepted best practice to create a dedicated data Paravirtual SCSI Adapters
store if the application has a demanding I/O profile, and data- It is a best practice to create a primary adapter for use with a disk
bases fall into this category. The creation of dedicated data stores that hosts the system software and Sybase ASE binaries, and
allows DBAs to define individual service-level guarantees for different separate paravirtual SCSI (PVSCSI) adapters for the Sybase ASE

14
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

data and log devices. PVSCSI adapters are high performance SAP Sybase ASE – Direct I/O Parameter
storage adapters that can result in greater throughput and lower In Sybase ASE version 15, the directI/O parameter is used to
CPU utilization. PVSCSI adapters are best suited for high configure Sybase ASE to transfer data directly to disk, bypassing
performance storage environments . 2
the operating system buffer cache. This directI/O parameter acts
similar to a “raw device” in legacy UNIX. In fact, raw partitions

Multiple Virtual SCSI Controllers in Linux are simply direct I/O to a file. Direct I/O performs I/O

VMware recommends creating multiple virtual SCSI controllers in the same manner as raw devices and provides the same

to distribute the I/O associated with database workloads. When performance benefit as raw devices (significantly better than

creating multiple SCSI controllers, map these controllers to the dsync), but it has the ease of use and manageability of file

database or operating system workload profile. Ensure that the system devices.

operating system and Sybase ASE binaries reside on one SCSI


The default settings for any filesystem-based device in Linux is
controller, and the Sybase ASE data files and log files reside on
to use DirectI/O, although it requires a filesystem that supports
separate SCSI controllers. In analytic or decision support data-
it, such as ext4 or xfs, as described in the test scenario. Do not
base systems, ensure that the temporary database files reside
use Direct I/O or dsync options for tempdb devices, as recovery
on a separate SCSI controller.
is not important for tempdb. The directio and dsync parameters

The primary purpose for using multiple virtual SCSI controllers are mutually exclusive. If a device has dsync set to true, you

is to parallelize the units of work in a database transaction. In cannot set DirectI/O to true for this device. In order to enable

this case, consider the implications when using multiple SCSI DirectI/O for a device, first reset dsync to false.

controllers to parallelize a single unit of work within a transaction.


File System Alignment
For instance, creating several SCSI controllers for data files
As in the physical world, file system misalignment can severely
increases throughput, but it may also increase the latency.
impact performance. File system misalignment not only manifests
Virtual Virtual Type Virtual Pool itself in databases, but with any high I/O workload. VMware makes
SCSI Driver Device
the following recommendations for VMware VMFS partitions:
LSI Logic 0:0 ASE binaries and RAID 5,
operating system Fibre Chanel
• Similar to other disk-based file systems, VMFS suffers a penalty
Paravirtual 1:0 ASE data files RAID 5,
Fibre Chanel when the partition is unaligned. Use VMware vCenter to create
Paravirtual 1:1 ASE data files RAID 5, VMFS partitions, since it automatically aligns the partitions
Fibre Chanel
along the 64 KB boundary.
Paravirtual 2:0 ASE log files RAID 1,
Fibre Chanel • In order to manually align your VMware VMFS partitions,

Table 3 Workload used


check your storage vendor’s recommendations for the partition
starting block (for example, EMC VNX uses 128k offsets).
2 To configure PVSCSI adapters for use with the SAP Sybase ASE database, go to the VMware
Knowledge Base Article: Configuring disks to use VMware Paravirtual SCSI (PVSCSI) adapters.

15
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

VMware Storage DRS In doing so, it combines the advantages of virtual provisioning
Historically, virtual machine storage provisioning has imposed with automatic storage tiering at the sub-LUN level to optimize
operational challenges. Monitoring datastore capacity and I/O performance and cost, while radically simplifying storage
load has proven to be very difficult, and as a result it is often management and increasing storage efficiency. Data movement
neglected. The Storage DRS is a feature introduced in VMware between tiers is based on performance measurement and user-
vSphere 5.0 provides smart virtual machine placement and
®
defined policies, and it is executed automatically and
load balancing mechanisms based on I/O and space capacity. nondisruptively using FAST VP.
It helps decrease the operational effort associated with the
provisioning of virtual machines and the monitoring of the
FAST VP Performance Measurement and Data Movement
storage environment.
With FAST VP, entire devices are promoted or demoted between
tiers based on overall device performance. FAST VP works at the
EMC VMAX FAST VP
sub-LUN level, introducing finer granularities of both performance
EMC® Fully Automated Storage Tiering for Virtual Pools (FAST VP)
measurement and data movement, and can spread the data
challenges the conventional wisdom associated with data
from a single thin device across multiple tiers.
placement and layout. In the past, the database architect
needed to work closely with storage administrators to make sure
The metrics collected at the sub-LUN level for thin devices under
data and log files were placed on the highest-performing drives,
FAST VP control contain measurements that enable FAST VP to
and to closely monitor the tracks on which data was placed.
make separate data movement requests for every 7,680 KB unit
of storage that makes up the thin device. This unit of storage
With FAST VP data placement is not static. FAST VP
is called an extent group and it consists of 10 contiguous thin
provides support for sub-LUN data movement in thin
device extents.
provisioning environments. FAST VP continuously analyzes
devices at the sub-LUN level. This enables it to identify and
relocate the specific parts of a LUN that are the most active and FASTVP and VMware Storage DRS
would benefit the most from being moved to higher-performing Auto-tiering technologies such as FAST VP migrate LUN segments
storage such as EFD. It also identifies the least active parts of (chunks) to different disk types, based on the use pattern. Hot
a LUN and relocates that data to higher-capacity, more cost- (frequently accessed) segments typically move to faster disks,
effective storage such as SATA, without altering performance. whereas cold segments move to slower disks. Depending on the
array type and vendor, there are different policies and thresholds
for these migrations.

16
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Storage DRS generates migration recommendations that prevent user by automating the selection of an individual datastore, leaving
out-of-space situations or resolve imbalances in performance. In the user with the much smaller scale decision of selecting a
order to prevent hot spots in the datastore cluster and decrease datastore cluster.
the overall latency imbalance, Storage DRS I/O load balancing
By default, Storage DRS is invoked every eight hours and requires
uses device modeling and virtual machine workload modeling:
performance data over more than 16 hours to generate I/O load
• Device modeling helps Storage DRS analyze the performance balancing decisions. FAST VP uses different time cycles to collect
characteristics of the devices backing the datastores. and analyze the workload before moving LUN segments than
Storage DRS. The misalignment of Storage DRS invocation and
• Virtual machine workload modeling analyzes workloads
FAST VP algorithm cycles makes it unpredictable when LUN
of virtual machines running inside the datastore cluster.
segments might be moved, potentially conflicting with Storage
Both device and workload modeling help Storage DRS
DRS recommendations.
assess improvement in I/O latency achieved after a virtual
machine migration. For these reasons, the best practice is either to turn off Storage DRS
or configure Storage DRS in manual mode with I/O Metric disabled.
In order to characterize the datastore, device modeling opens
and reads random blocks. Storage DRS does not detect active
Storage DRS Manual Mode and I/O Metric Inclusion
FAST VP structures and might open blocks located on disks of
In manual mode, Storage DRS makes recommendations when
similar speed, either slow or fast. The datastore can be distributed
the configured thresholds for latency or space utilization have
throughout multiple tiers of disk types, thereby leading to an
been exceeded. It also provides recommendations for placement
incorrect performance characterization. Therefore, the assessment
during the provisioning process of a virtual machine or a virtual
of migrating the workload can be incorrect. This can generate a
disk. In other words, when Storage DRS is set to manual mode,
recommendation to move into an underperforming datastore. It
it monitors the environment and makes recommendations on
also can cause a recommendation to be withheld while there is
destinations for virtual disk migrations. Manual mode is the default
sufficient performance available in other datastores.
setting when configuring a Storage DRS datastore cluster.
By aggregating datastores backed by FAST VP technology into
In order to disable I/O Metric inclusion, go to Advanced Options,
a datastore cluster, Storage DRS provides simplified storage
and uncheck the I/O Metric box in the Storage DRS Runtime
management on the datastore cluster level, rather than on the
Rules configuration window.
datastore level. Initial placement speeds up the process for the

17
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Deployment Tips/Guidelines Summary

Component Settings Component Settings

Operating system •• XFS file system vSphere •• Scale vCPUs for each ASE engine
•• NOOP I/O Scheduler (Example: 4 ASE engines = 4 vCPUs).
•• Hyperthreading enabled – •• Allow vSphere to Choose the Best
workload dependent Virtual Machine Monitor based on
the CPU and Guest Operating
Server BIOS •• Enable: System Combination.
•• Virtualization Technology •• Set memory reservations to equal
•• Turbo mode to the size of the Sybase ASE max
•• Hardware based virtualization support memory setting.
•• Hyperthreading •• Use VMXNET 3 virtual NICs.
•• Wake on LAN (Required for DPS) •• Hardware Assisted Memory Virtualization
- CPU/MMU Virtualization Option =
•• Execute Disable Automatic
(Required for vMotion/DRS
•• Use Virtual Distributed Switch for ease
•• Static high performance/balanced of management and to isolate database,
•• Disable: application, vMotion traffic on separate
•• Node interleaving VLANS.
•• C1E Halt state •• Determine if disabling interrupt
•• Power saving coalescing is optimal.
•• Unused features: Video BIOS •• Determine if disabling LRO is optimal.
Shadowable, Video RAM Cacheable, •• VMFS 5.0:
on-board audio, on-board modem, •• Paravirtualized SCSI driver for
on-board serial ports, on-board SAP Sybase ASE devices.
parallel ports, on-board game port, •• LSI Logic SCSI driver for operating
floppy drive, CD-ROM, USB system – SAP Sybase ASE binaries.
Sybase ASE •• Use large (huge)pages •• Align file system offset – Ex: VNX
(supported for 15.0.3 and greater). 128K offset.
•• Use direct I/O = true for data devices, •• Use dedicated datastores for SAP
set to false for temp devices: Sybase ASE production databases.
•• DSYNC turned off •• Use multiple Virtual SCSI Controllers.
•• AIO Lib loaded •• Leverage New Storage Architectures
•• Tabular Data Stream (TDS) packet (EX: EMC FASY VP).
size 16K (LOBs)
Table 4. Sybase ASE on vSphere: General Guidelines for Settings
•• Page Size 4K typical – 16K for SAP

18
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Conclusion
As shown by our joint testing, even the most demanding
SAP Sybase ASE database workloads can be virtualized using
VMware vSphere. This is an important data point for customers
who are considering running not only their existing SAP Sybase
ASE database on VMware vSphere, but SAP Business Suite on
SAP Sybase ASE as well. These results were achieved for the
most part out-of-the-box. There is minimal tuning or specialized
configurations needed to run SAP Sybase ASE on VMware
vSphere. This paper is a guideline database administrators
can use to easily virtualize their environments, without having
to learn or relearn their duties.

19
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Appendix A: Test Results − ASE Native versus vSphere Figure A-2 results show that this configuration exceeded
Appendix A provides summarized, sample validation test results for physical performance.
Sybase ASE running on native physical hardware versus running
on the VMware vSphere virtualization platform. These tests were Figure A-2. 4 SAP-Sybase ASE engine threads 6 vCPUs
run on standard x86 hardware using an OLTP transactional
benchmark. The results below show consistently less than
5 percent transactional overhead as the number of users
are increased.

(The graphs below are part of a presentation delivered by


Chris Brown at SAP TechEd in Las Vegas in 2012.)

Figure A-1 results show an average of less than 5 percent overhead.

Figure A-1. 11 Sybase ASE engine threads and 12 vCPUs

Figure A-3 results show a significant increase in transactional


throughput with EMC FAST VP enabled.

Figure A-3. Sybase ASE on vSphere with EMC FAST VP

20
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Resources • VMware Support and Downloads web site: http://www.vmware.


Customers can find more information about the products and com/support/product-support/vsphere/index.html
technologies described in this paper using the links listed below. • Performance Best Practices for VMware vSphere 5.0:
http://www.vmware.com/pdf/Perf_Best_Practices_vSphere5.0.pdf
VMware
Deployment References General Information
• Press Release: VMware Cloud Infrastructure Validated by SAP • VMware web site: http://www.vmware.com/
for Virtualization of SAP’s Database Portfolio: • Featured VMware Documentation Sets:
http://www.vmware.com/company/news/releases/vmw-sap-
http://www.vmware.com/support/pubs/
database-05-14-12.html
• VMware Licensing Help Center:
• VMware Global Alliances Blog: SAP-Sybase Makes Running
http://www.vmware.com/support/licensing/
Business Critical Databases on vSphere Hassle Free:
• VMware Product Podcasts:
http://blogs.vmware.com/alliances/2012/06/vmware_sybase_
http://www.vmware.com/technical-resources/podcasts/
sap.html
• VMware Technology Network (Community) web site:
• Performance Characterization of VMFS and RDM Using
http://communities.vmware.com/community/vmtn
a SAN: http://www.vmware.com/files/pdf/performance_char_
vmfs_rdm.pdf • Community, VMware Knowledge Base:
http://communities.vmware.com/community/vmtn/resources/
• Configuring Disks to Use VMware Paravirtual SCSI (PVSCSI)
knowledgebase
Adapters: http://kb.vmware.com/kb/1010398
• VMware Support Insider: http://blogs.vmware.com/kbtv/
• VMware vSphere® Distributed Switch Best Practices:
http://www.vmware.com/files/pdf/techpaper/vsphere-distrib- • VMware TV: http://www.youtube.com/user/vmwaretv
uted-switch-best-practices.pdf • VMworld TV: http://www.youtube.com/user/VMworldTV
• VMware vSphere: http://www.vmware.com/products/datacenter- • VMware KB TV (external):
virtualization/vsphere/overview.html
http://www.youtube.com/user/VMwareKB

21
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

Acknowledgements
The following individuals contributed to the creation of this
technical paper:

• Peter Thawley, Vice President Global Database Solutions, SAP

• Todd Muirhead, Performance Staff Engineer, R&D, VMware

• Tim Harris, Director of Validation Services, Alliances, VMware

• Biswapati Bhattacharjee, ISV Validation Engineer,


Alliances VMware

• Jon Catanzano, Senior Technical Writer/Editor, consultant

22
SAP Sybase Adaptive Server Enterprise on VMware vSphere®

23
© 2013 VMware, Inc. All rights reserved. © 2013 SAP AG. All rights reserved.

VMware Global, Inc. SAP AG - Walldorf


3401 Hillview Ave. Dietmar-Hopp-Allee 16 (früher: Neurottstraße)
Palo Alto, CA 94304 69190 Walldorf
USA Germany

This product is protected by U.S. and international SAP, R/3, SAP NetWeaver, Duet, PartnerEdge, ByDesign, SAP BusinessObjects Explorer,
copyright and intellectual property laws. VMware StreamWork, SAP HANA, and other SAP products and services mentioned herein as well as
products are covered by one or more patents listed at their respective logos are trademarks or registered trademarks of SAP AG in Germany and
http://www.vmware.com/go/patents. other countries.

VMware is a registered trademark or trademark of VMware, Inc. in the United States and/or Business Objects and the Business Objects logo, BusinessObjects, Crystal Reports, Crystal
other jurisdictions. All other marks and names mentioned herein may be trademarks of their Decisions, Web Intelligence, Xcelsius, and other Business Objects products and services
respective companies. mentioned herein as well as their respective logos are trademarks or registered trademarks of
Business Objects Software Ltd. Business Objects is an SAP company.

Sybase and Adaptive Server, iAnywhere, Sybase 365, SQL Anywhere, and other Sybase
products and services mentioned herein as well as their respective logos are trademarks or
registered trademarks of Sybase Inc. Sybase is an SAP company.

Crossgate, m@gic EDDY, B2B 360°, and B2B 360° Services are registered trademarks
of Crossgate AG in Germany and other countries. Crossgate is an SAP company.

All other product and service names mentioned are the trademarks of their respective
companies. Data contained in this document serves informational purposes only.
National product specifications may vary.

These materials are subject to change without notice. These materials are provided by
SAP AG and its affiliated companies (“SAP Group”) for informational purposes only, without
representation or warranty of any kind, and SAP Group shall not be liable for errors or
omissions with respect to the materials. The only warranties for SAP Group products and
services are those that are set forth in the express warranty statements accompanying
such products and services, if any. Nothing herein should be construed as constituting an
additional warranty.

3/13

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy