Epicor ERP Hardware Sizing Guide WP ENS
Epicor ERP Hardware Sizing Guide WP ENS
Epicor ERP Hardware Sizing Guide WP ENS
ERP 10 Hardware
Sizing Guide
Table of Contents
Introduction...............................................................................................1
Hardware sizing guide for up to 15 user load............................................2
ERP 10 Hardware sizing guide (up to 50 user load)................................4
ERP 10 Application usage Load Expectation (up to 50 user load)..............5
ERP 10 Hardware for up to 200 user load..............................................6
ERP 10 Application usage Load Expectation (up to 200 user load)............8
ERP 10 Application Server/Hypervisor/Reporting Server...........................9
ERP 10 SQL Server.................................................................................10
ERP 10 Hardware sizing for up to 2,000 Users.....................................12
ERP 10 Hardware sizing for up to 4,000 Users.....................................12
Scale Out Configurations.........................................................................12
ERP 10 Virtual Appserver on VMware...................................................13
ERP 10 Virtual Appserver on Hyper-V....................................................14
ERP 10 Virtual SQL Server on VMware..................................................15
ERP 10 Virtual SQL Server on Hyper-V...................................................16
ERP 10 Hardware sizing on Virtual Servers............................................17
Hardware Requirements for Other Epicor Systems...................................17
Other Epicor Systems...............................................................................18
Network...................................................................................................19
Appendix..................................................................................................21
b
Introduction
Welcome to the Epicor ERP 10 Hardware Sizing Guide. The goal of this guide is to
provide a practical approach to sizing your Epicor ERP 10 application and database
server to obtain the best performance possible from your Epicor software. It will
also help you plan for the future growth of your system.
The key to success in getting your hardware sized correctly is to define your
application load correctly and then to match it with the appropriate hardware
resources. Choosing a hardware architecture which can grow as your business
and/or application load grows is also important.
It is very difficult to accurately come up with a hardware recommendation that
matches your unique application usage. Many of our customers find this guide
very useful and accurate after they have followed all the steps described here.
There are cases where the provided recommendations may fall short for whatever
reason. It is therefore very difficult to guarantee or imply that the hardware
purchased using this guide will solve your unique business requirements. To
improve the odds of success, we highly recommend that you work with the Epicor
Technical Solutions group and implement their hardware sizing recommendations.
This guide is meant for customers buying new hardware and software so we
recommend latest software versions supported by Epicor ERP. If you want to
determine if other versions of software are supported then please contact Epicor
technical support.
Please note that you should consult Epicor Technical Services for High Availability
and Business Continuity customer requirements as this sizing guide cannot get
into details of that topic.
ERP 10 Physical Server. This configuration can support standard load of 15 Office + Data
collection users. See table 2 for the definition of standard load.
Operating System & SQL Server Windows Server 2012 Standard Edition SQL Server 2012 Standard Edition (supports up
to 16 cores on Standard SKU)
Applications
Epicor ERP 10, Epicor Web Access (EWA), Enterprise Search (ES),
Epicor Social Enterprise (ESE)
Processors
Memory (RAM)
32 GB ECC
2 x 146GB 15k RPM HDDs in RAID 1, 3 Gbps, 4 x 200GB SSDs in RAID 1 or 10, 3 Gbps
Operating system and page file on HDD SQL Database and reports on SSDs.
SSD (Recommended)
Note - MLCs are cheaper than SLCs. SLCs are more reliable. MLCs are getting better by the
day, if you maintain a warranty coverage on Enterprise MLC SSDs then you can purchase
MLCs.
Storage Option #2 (you can use
this option if your SAN passes
the SQLIO test)
Existing Storage Area
Network (SAN)
To find out if your SAN delivers sufficient performance for SQL database, see the Testing
the performance of an existing SAN section in the Appendix for details on how to run a
storage benchmarking tool (SQLIO).
Required SQLIO results at all times including during peak load from other application
usage on the SAN:
Database drive letter (location of SQL
mdf file):
Must be able to deliver at least 100MB/sec at a latency of less than 5ms.
Transaction log drive letter (location of SQL log file):
Must be able to deliver at least 30MB/sec at a latency of less than 5ms.
Tempdb drive letter (location of SQL temp db file):
Must be able to deliver at least 50MB/sec at a latency of less than 5ms.
Network
For good network performance, use 1 Giga bit (or faster) network cards.
To ensure that network latency is as low as possible between the Epicor application
server and the SQL database server, add at least two network cards:
One dedicated network card that is directly connected to the SQL database
(Cable from network card to network card).
One dedicated network card that is connected to the LAN (and is processing all network
traffic except SQL
database traffic).
Table 2
Load expectation for up to 15 user load (Small ERP)
Application Usage
Load Description
Typical application usage by end-users using ERP 10 via Smart Client and/or
browser and / or mobile devices.
If you are processing more than 25K transactions (insert, update or delete) in any
one hour period of the working day then schedule this activity to less busy time of
the day. 25K transaction is the sum total of all sources of automatic processing.
If this activity cannot be scheduled to less busy time of the day then please discuss
the server sizing with Epicor technical services.
Heavy and concurrent reporting should be scheduled to less busy time of the day or
on a separate reporting appserver.
No (If yes then discuss the increase in usage with Epicor technical services)
MRP regeneration will be scheduled during less busy times. For daily MRP runs use
MRP net change.
If your business requires you to run more than 2 processes and 2 schedulers for
MRP run then discuss the increase usage with Epicor technical services.
No
users
Seasonal peaks
No
ERP 10 Physical Server. This configuration can support standard load of 50 Office +
Data collection users. See table 4 for the definition of standard load.
Applications
Epicor ERP 10, Epicor Web Access (EWA), Enterprise Search (ES),
Epicor Social Enterprise (ESE)
Processors
Memory (RAM)
64 GB ECC
To find out if your SAN delivers sufficient performance for SQL database, see the
Testing the performance of an existing SAN section in the Appendix for details on
how to run a storage benchmarking tool (SQLIO).
Required SQLIO results at all times including during peak load from other
application usage on the SAN:
Database drive letter (location of SQL mdf file):
Must be able to deliver at least 100MB/sec at a latency of less than 5ms.
Transaction log drive letter (location of SQL log file):
Must be able to deliver at least 30MB/sec at a latency of less than 5ms.
Tempdb drive letter (location of SQL temp db file):
Must be able to deliver at least 50MB/sec at a latency of less than 5ms.
Network
For good network performance, use 1 Giga bit (or faster) network cards.
To ensure that network latency is as low as possible between the Epicor application
server and the SQL database server, add at least two network cards:
One dedicated network card that is directly connected to the SQL database
(Cable from network card to network card).
One dedicated network card that is connected to the LAN (and is processing all
network traffic except SQL database traffic).
Load Description
Typical application usage by end-users using ERP 10 via Smart Client and/or
browser and/or mobile devices.
If you are processing more than 50K transactions (insert, update, or delete) in any
one hour period of the working day then schedule this activity to less busy time of
the day. 50K transaction is the sum total of all sources of automatic processing.
If this activity cannot be scheduled to less busy time of the day then please discuss
the server sizing with Epicor technical services.
Heavy and concurrent reporting should be scheduled to less busy time of the day or
on a separate reporting appserver.
No (If yes then discuss the increase in usage with Epicor technical services)
MRP regeneration will be scheduled during less busy times. For daily MRP runs use
MRP net change.
If your business requires you to run more than 3 processes and 3 schedulers for
MRP run then discuss the increase usage with Epicor technical services.
No
Seasonal peaks
No
ERP 10 Physical Server Machine. This configuration can support standard load of 200
Office + Data collection users. See table 6 for the definition of
standard load.
Windows Server 2012 Standard Edition SQL Server 2012 Standard Edition (supports
up to 16 cores on Standard SKU)
Applications
Epicor ERP 10, Epicor Web Access (EWA), Enterprise Search (ES),
Epicor Social Enterprise (ESE)
Processors
Memory (RAM)
96 GB ECC
PCI-Express based card i.e. Fusion Fusion-IO should hold SQL database (MDF file), SQL transaction log (LDF file) and SQL
IO (SAS / SATA based) storage
tempdb files.
Epicor Reports 2 x 15K HDD in RAID 1
Total
4 15K RPM HDD (3 Giga bits /seconds SATA or SAS)
Quantity 1, at least 785 GB ,
Fusion-IO card
Storage Option #2 - SSD
(Recommended)
SQL transaction log (LDF file): 2 x Solid State in RAID 1. (Only Epicor SQL Log file on
this drive)
SQL tempdb: 2 x Solid State Drive in RAID 1. (This will house at least 8 tempdb files)
Epicor Reports 2 x Solid State Drive in RAID 1
Total drives - 12 Solid State Drives (3 Giga bits /seconds SATA or SAS)
Note - MLCs are cheaper than SLCs. SLCs are more reliable. MLCs are getting better
by the day, if you maintain a warranty coverage on Enterprise MLC SSDs then you can
purchase MLCs.
To find out if your SAN delivers sufficient performance for SQL database, see the
Testing the performance of an existing SAN section in the Appendix for details on
how to run a storage benchmarking tool (SQLIO).
Required SQLIO results at all times including during peak load from other application
usage on the SAN:
Database drive letter (location of SQL mdf file):
Must be able to deliver at least 100MB/sec at a latency of less than 5ms.
Transaction log drive letter (location of SQL log file):
Must be able to deliver at least 30MB/sec at a latency of less than 5ms.
Tempdb drive letter (location of SQL temp db file):
Must be able to deliver at least 50MB/sec at a latency of less than 5ms.
Network
For good network performance, use 1 Giga bit (or faster) network cards.
To ensure that network latency is as low as possible between the Epicor application
server and the SQL database server, add at least two network cards:
One dedicated network card that is directly connected to the SQL database
(Cable from network card to network card).
One dedicated network card that is connected to the LAN (and is processing all
network traffic except SQL database traffic).
Load Description
Typical application usage by end-users using ERP 10 via Smart Client and / or
browser and / or mobile devices.
If you are processing more than 100K transactions (insert, update or delete) in any
one hour period of the working day then schedule this activity to less busy time of
the day. 100K transaction is the sum total of all sources of automatic processing.
If this activity cannot be scheduled to less busy time of the day then please discuss
the server sizing with Epicor technical services.
Heavy and concurrent reporting should be scheduled to less busy time of the day
or on a separate reporting appserver.
No (If yes then discuss the increase in usage with Epicor technical services)
MRP regeneration will be scheduled during less busy times. For daily MRP runs use
MRP net change.
If your business requires you to run more than 6 processes and 6 schedulers for
MRP run then discuss the increase usage with Epicor technical services.
No
Seasonal peaks
No
Operating System
Applications
Epicor ERP 10, Epicor Web Access (EWA), Enterprise Search (ES),
Epicor Social Enterprise (ESE)
Processors
Memory (RAM)
64 GB ECC
Network
For good network performance, use 1 Giga bit (or faster) network cards.
To ensure that network latency is as low as possible between the Epicor application
server and the SQL database server, add at least two network cards:
One dedicated network card that is directly connected to the SQL database
(Cable from network card to network card).
One dedicated network card that is connected to the LAN (and is processing all network
traffic except SQL database traffic).
Applications
Epicor ERP 10, Epicor Web Access (EWA), Enterprise Search (ES), Epicor Social Enterprise (ESE)
Processors
SQL Server 2012 Standard Edition (supports up to 16 cores on Standard SKU, you may want
to consider enterprise edition for high availability and online indexing features)
Intel Xeon E5-2667 v2 3.3GHz, 25M Cache, 8.0GT/s QPI, Turbo, HT, 8C, 130W
It is recommended to leave Hyper-Threading enabled (check BIOS settings).
Disable Power Saving options and C-States in BIOS. Enable Turbo Boost.
Memory (RAM)
128 GB ECC
Storage Option #1
Operating system and page file: 2 x 15K HDD in RAID 1.
Fusion-IO (Recommended) Fusion-IO should hold SQL database (MDF file), SQL transaction log (LDF file) and
PCI-Express based card
SQL tempdb files.
i.e. Fusion IO (SAS / SATA
Epicor Reports 2 x 15K HDD in RAID 1
based) storage
Total
4 15K RPM HDD (3 Giga bits /seconds SATA or SAS)
1 Fusion-IO card
Storage Option #2 SSD
Solid State Drive (SAS /
SATA based) storage
10
To find out if your SAN delivers sufficient performance for SQL database, see the Testing the
performance of an existing SAN section in the Appendix for details on how to run a storage
benchmarking tool (SQLIO).
Required SQLIO results at all times including during peak load from other application usage
on the SAN:
Database drive letter (location of SQL mdf file):
Must be able to deliver at least 100MB/sec at a latency of less than 5ms at all times.
Transaction log drive letter (location of SQL log file):
Must be able to deliver at least 30MB/sec at a latency of less than 5ms at all times.
Tempdb drive letter (location of SQL temp db file):
Must be able to deliver at least 50MB/sec at a latency of less than 5ms at all times.
Network
For good network performance, use 1 Giga bit (or faster) network cards.
To ensure that network latency is as low as possible between the Epicor application server
and the SQL database server, add at least two network cards:
One dedicated network card that is directly connected to the SQL database (Cable from
network card to network card).
One dedicated network card that is connected to the LAN (and is processing all network
traffic except SQL database traffic).
11
Quantity
Appserver
1 x APP01
SQL Server
1 x SQL01
1 X APP01
Quantity
Appserver
2 x APP01
SQL Server
1 x SQL01
1 X APP01
Load balancer
12
VMWAREAPP01
Host software
Guest OS
4 virtual CPUs
Create a full CPU reservation to ensure that the virtual machine
has dedicated access to physical CPU cores.
It is not recommended to overcommit CPU resources.
Memory (RAM)
12 GB (reserved RAM)
It is not recommended to overcommit memory resources.
If needed, disable Memory Ballooning so that the virtual
application server has a dedicated amount of RAM.
Ensure that the RAM on the host supports ECC.
See APP01
Network
Load balancing
VM Backup software
13
HYPERVAPP01
Host software
Guest OS
4 virtual CPUs
Create a full (100%) virtual machine reserve to ensure that the virtual machine has
dedicated access to physical CPU cores.
It is not recommended to overcommit CPU resources.
Memory (RAM)
12 GB (reserved RAM)
Select Static memory when you allocate memory to the virtual machine.
It is not recommended to overcommit memory resources.
Ensure that the RAM on the host supports ECC.
See APP01
Use the IDE Controller option for the virtual hard disk that contains the OS.
Use the SCSI Controller option for all other virtual hard disks.
Network
Use the Network Adapter virtual network card (and not the Legacy Network
Adapter).
Load balancing
Physical load balancing appliance recommended. Software based works too. Use
Multicast Mode if you are using Microsoft Network Load Balancing (NLB). This is
recommended for stability reasons.
Backup software
14
VMWARESQL01
Host software
Guest OS
Memory (RAM)
24 GB (reserved RAM)
It is not recommended to overcommit memory resources.
If needed, disable Memory Ballooning so that the virtual application server has a
dedicated amount of RAM.
Ensure that the RAM on the host supports ECC.
See SQL01
(host requirements)
Virtual hard disk
Network
Ensure that the physical host has multiple physical network cards which are all
utilized by the hosted virtual machines.
Use the VMXNET3 virtual network card.
VM Backup software
VM level snapshot backup works for Epicor ERP 10. When you virtualize SQL
Server be sure to test and make sure differential backup works correctly in
your case.
File-level backups are supported.
15
HYPERVSQL01
Host software
Guest OS
Memory (RAM)
24 GB (reserved RAM)
Select Static memory when you allocate memory to the virtual machine.
It is not recommended to overcommit memory resources.
Ensure that the RAM on the host supports ECC.
See SQL01
Use the IDE Controller option for the virtual hard disk that contains the OS.
Use the SCSI Controller option for all other virtual hard disks.
Network
Backup software
VM level snapshot backup works for Epicor ERP 10. When you virtualize SQL Server
be sure to test and make sure differential backup works correctly in your case.
File-level backups are supported.
16
Quantity
Appserver
SQL Server
Load Balancing
Processor
2.8 GHz or higher. Recommended dual core 32-bit or 64-bit processor and SATA
hard drives. SSD is especially beneficial if the workstation runs multiple applications
at the same time.
RAM
2 GB
Monitor
SVGA or higher adapter Color SVGA monitor with 1024 X 768 resolution
DVD-R
Other
17
Hardware Description
1 X VMWAREAPP01 OR 1 X HYPERVAPP01
EPM
AFR
1 X VMWAREAPP01 OR 1 X HYPERVAPP01
Handheld Equipment
Device Requirements
Screen size minimum of 240 x 320
Device must support MS Terminal Services client
IEEE 802.11 Radio Frequency standards support
Access Points
IEEE 802.11 Radio Frequency standards support
Barcode Equipment
MES Workstation
PioneerPOS Stealth M5
Same configuration as Epicor workstation (Table 12)
Printer
At least one laser printer required for standard reports and forms
Must have current Windows drivers
Communications
18
Network
Hardware Recommendations
A 1 Gbit (or faster) network is recommended. In some cases a 10 Gbit network connection will provide additional benefit.
Fiber optic cable recommended in environment with electrical noise.
When designing network, try to minimize the latency (one way to measure latency is ping time) between the Epicor ERP 10
client workstations and the Epicor ERP 10 application server box. For example users having a 125ms ping time will see slower
response than the users having a ping time of 25ms between the workstation and the application server box.
If you have users in remote sites, consider placing a Remote Desktop Services (RDS) server on the same location (same LAN)
as the Epicor ERP 10 server. Instead of running the Epicor ERP 10 client locally and connecting to the Epicor ERP 10 server over
a WAN, users can access the client on a Remote Desktop session. This improves performance as the Epicor ERP 10 server and
Epicor ERP 10 client are both on the same LAN.
WAN accelerator - Consider implementing WAN accelerators on both sides of a WAN if you got environments with high
latency and/or low bandwidth.
Network Bandwidth
The bandwidth of a network defines the amount of data that can be transferred across the network at once. If the bandwidth
of the network is insufficient for the amount of information being transmitted (usually because too many users are on a single
network segment), performance drops noticeably. In these cases, the network will need re-configuring to lower the amount of
traffic or increase the bandwidth of the network. Traditional hubs are being replaced by high-speed data switches, which further
help alleviate network bottlenecks. Multiple server network cards and automated load balancing may be considered too.
Network bandwidth is not the only factor that determines the speed of a network as perceived by the end user. The other
key element of network performance is latency. While it would often appear that data is transmitted instantly between one
point and another (that is, with no delay at all), this is typically not the case. Network latency may be caused by:
The time it takes for a packet to travel from point to point
The transmission medium (optical fiber, wireless, etc.) may introduce some delay because larger packets typically take longer to
move from place to place
Routers and other processing points each take time to examine and possibly make changes to packet header
Intermediate devices such as switches and bridges may cause delays
It is often difficult to diagnose a network bandwidth or latency issue and this is best left to professional network analysts.
As a general rule, the Epicor client has a bandwidth requirement of 10 Kbps (kilobits per second) to 500Kbps. This will
vary depending on the volume of data transferred between the client and the server. Additionally, file attachments require
increased bandwidth.
A simplified rule of thumb for the number of clients to connect per network segment is to determine the worst-case
acceptable throughput for each network client, then divide that amount into the selected network throughput. For example,
if each client should have no less than 1.5 Mbps of available bandwidth, and the network supports 100 Mbps, the segment
could possibly support 66 clients. Unfortunately this simple calculation does not take into account the network latency,
network media characteristics, or network traffic in addition to Epicor, but it does provide for a good starting point for the
maximum amount of clients per network segment.
19
20
Appendix
Testing Performance of an Existing SAN
If you have an existing SAN and want to ensure that it can deliver good performance, then you can run a Microsoft storage
benchmark utility
named SQLIO.
The following tests are designed to test various aspects of an I/O disk subsystem i.e. bandwidth (Mega Bytes/second i.e. MB/
Sec), Latency (milliseconds), performance of your I/O system with desired block size (64KB) and file size and type of I/O read
or write and sequential v/s random writes. The parameters described in previous statement have a great impact on IOPS and
hence they are specified exactly as needed here for testing using SQLIO. On the same machine you will get different IOPS
number if you change any one parameter. So testing with Epicor recommended parameters is highly recommended.
Download SQLIO from:
http://www.microsoft.com/download/en/details.aspx?id=20163
Instructions:
The default install path on a 64-bit OS for SQLIO is C:\Program Files (x86)\SQLIO\
In the SQLIO install path, the configuration file called param.txt determines which drive letter SQLIO will test.
If you want to test the D-drive letter, then change the param.txt content to the following (80 Gigabyte size is important):
21
Description
-o
Number of outstanding I/O requests per thread. When attempting to determine the capacity of a given volume
or set of volumes, start with a reasonable number for this and increase until disk saturation is reached (that is,
latency starts to increase without an additional increase in throughput or IOPs). Common values for this are 8,
16, 32, 64, and 128.
Keep in mind that this setting is the number of outstanding I/Os per thread.
-LS
Instructs SQLIO to capture disk latency information. Capturing latency data is recommended when
testing a system.
-k
Specify either R or W (read or write). Both reads and writes should be tested. Keep in mind that the storage
array cache may absorb a good amount of write activity and potentially skew results. Ensure that the duration is
long enough to saturate the cache in order to get a true measure of storage capacity.
-s
Duration of test (in seconds). For initial tests, running for 5-10 minutes per I/O size is recommended to get a
good idea of I/O performance.
-b
Size of the I/O request in kbytes. For random I/O, 8 KB and 64 KB are as many values as are need to be tested.
The most common random I/O size for SQL Server is 8 KB. For random I/O, pay close attention to the number
of I/Os per second and latency. For sequential I/O, test a range of sizes (4, 8, 16, 32, 64, 128, 256) and pay
attention to throughput and latency.
-f
-F
Name of the file that will contain a list of the test file(s) used by SQLIO.
22
About Epicor
Epicor Software Corporation is a global leader delivering business software
solutions to the manufacturing, distribution, retail, and service industries. With
more than 40 years of experience, Epicor has more than 20,000 customers in over
150 countries. Epicor solutions enable companies to drive increased efficiency and
improve profitability. With a history of innovation, industry expertise and passion
for excellence, Epicor inspires customers to build lasting competitive advantage.
Epicor provides the single point of accountability that local, regional, and global
businesses demand. For more information, visit www.epicor.com.
These HW sizing guidelines are designed to assist you in determining your hardware requirements for a specific solution and to help you zero in on a solution between upper and lower estimates for your hardware requirements. These guidelines
represent the culmination of our combined experiences in the field and in testing of our software in various environments and offer our reasonable estimates and approximations of the hardware resources necessary for different ERP software
implementation scenarios. However, each Customer project, situation and system is unique and these general Hardware sizing guidelines and recommendations are not absolute nor are they intended to be your sole source of information pertaining
to the process of sizing the Hardware for your system. Hardware Sizing estimates also very much depend on such variables as the architecture of the project, the complexity of the installation and implementation, expected usage and traffic and
the project requirements. It is difficult to accurately arrive at a hardware recommendation that matches your unique application usage. There are instances where the suggested guidelines and recommendations fall short for whatever reason and it
is very difficult to guarantee or imply that the hardware purchased using this guide will solve or meet your unique business requirements. Thus we do not guarantee that the Hardware you acquire based on these guidelines will solve or meet your
requirements, nor do we guarantee or warrant the software utilization or performance that you will ultimately experience based upon such Hardware. To improve the odds of success, we highly recommend that you work with the Epicor Technical
Solutions group and allow them to work with you directly in determining your hardware sizing recommendations.
info@epicor.com
www.epicor.com
Asia
238A Thomson Road #23-06
Novena Square Tower A
Singapore 307684
Singapore
Phone: +65.6333.8121
Fax: +65.6333.8131
This document and its contents, including the viewpoints, recommendations, dates and functionality descriptions expressed herein are believed to be accurate as of its date of publication, April 2014. However, Epicor Software Corporation does not
make any guarantee, representations or warranties with regard to the enclosed information and specifically disclaims the implied warranties of fitness for a particular purpose and merchantability. All information contained herein is subject to change
without notice. The usage of any Epicor Software shall be pursuant to an Epicor license agreement and the performance of any consulting services by Epicor personnel shall be pursuant to Epicors services agreements. Any hardware purchased shall
be subject to its own equipment purchase agreement. Epicor, the Epicor Logo, and Business Inspired are trademarks of Epicor Software Corporation, registered in the United States and certain other countries. All other trademarks acknowledged.
Copyright 2014 Epicor Software Corporation.
This guide is meant for customers buying new hardware and software so we recommend latest software versions supported by Epicor ERP. If you want to determine if other versions of software are supported then please contact Epicor technical support.