2010 Samples Branch Cache

89 Fifth Avenue, 7th Floor
New York, NY 10003

www.TheEdison.com
212.367.7400
White Paper
Enhanced Branch Office Productivity

and WAN Bandwidth Savings with
Microsoft BranchCache™
Printed in the United States of America.
Copyright  2010 Edison Group, Inc. New York. Edison Group offers no warranty either
expressed or implied on the information contained herein and shall be held harmless for errors
resulting from its use.
All products are trademarks of their respective owners.

First Publication: March 2010
Produced by: Howard Marks, Sr. Analyst; Kelly Strand Anderson, Sr. Analyst; Barry Cohen,
Editor-in-Chief
Table of Contents
Executive Summary ..................................................................................................................... 1

Introduction .................................................................................................................................. 4
The Remote Office Dilemma ..................................................................................................... 6
Testing BranchCache Performance ........................................................................................... 9
Test Methodology ................................................................................................................... 9
Laboratory Configuration ................................................................................................... 13
Test Results ................................................................................................................................. 16
Summary of Results ............................................................................................................. 16
Scenarios ................................................................................................................................ 19
 Test Scenario 1 – CIFS/SMB File Accesses .................................................................. 19
 Test Scenario 2 – HTTP File Downloads..................................................................... 23
 Test Scenario 3 – SCCM Application Package Download ....................................... 27
 Test Scenario 4 – MOSS 2007 File Access.................................................................... 32
Conclusions and Recommendations ...................................................................................... 36
Appendices .................................................................................................................................. 37
Test Environment ................................................................................................................. 37
Test Results ............................................................................................................................ 39
CIFS/SMB Test Results......................................................................................................... 39
HTTP Test Results ................................................................................................................ 45
MOSS2007 Test Results........................................................................................................ 51
SCCM Test Results ............................................................................................................... 55
Executive Summary
Organizations seeking to lower operating expenses are investigating technology

solutions that can directly lower costs and improve productivity. One area in which
these organizations are seeking solutions is in the IT infrastructure supporting remote or
branch offices. Remote office workers frequently and repeatedly download the same
files from the central office. Increasingly, these files are composed not only of text but
also of mixed content types that include photographs, illustrations, animations, and
videos. Such documents are much larger than those in the past, demanding very high
Wide Area Network bandwidth utilization levels and slowing file access with adverse
impact on productivity.
With the release of Windows 7, Microsoft has delivered BranchCache™, a solution to

address the remote office challenge. Working in conjunction with Windows Server 2008
R2, Windows 7 clients with BranchCache capability enabled can expect to see significant
performance improvements when downloading files from the central office.
The concept behind BranchCache is straightforward: when BranchCache is enabled, a

copy of data accessed from an Intranet web, SharePoint, or file server is cached locally in
the branch office. When another client on the same network in the branch office requests
the same file, the file is transferred from the local cache rather than from across the
WAN. BranchCache improves performance when copying files between central and
remote or branch offices using the SMB, HTTP, and HTTPS protocols.
Microsoft commissioned Edison Group to examine WAN bandwidth savings with

BranchCache. In order to quantify the extent of the performance improvement that
organizations can expect, Edison analysts created a set of use case scenarios and ran
comparison tests over a range of typical WAN network types. These tests compared the
performance of Windows 7 clients with BranchCache capability enabled against client
PCs running the older Windows operating systems Windows XP and Windows Vista.
Edison’s results showed savings that, in many cases, exceeded our expectations. The
results were consistent across all the network bandwidth conditions and for all the
operating systems and BranchCache settings. These savings included:
 CIFS/SMB file downloads for a PowerPoint presentation were as much as 69 percent
faster while utilizing an average of up to 59 percent less bandwidth than with
Windows Vista or Windows XP.
Edison: Enhanced Branch Office Productivity and WAN Bandwidth Savings with BranchCache Page 1
 User download times with the HTTP protocol were 63 percent faster than with
Windows Vista and Windows XP, while requiring 49 percent less Wide Area
Network bandwidth.
 Downloading an application or patch using Microsoft System Center Configuration
Manager (SCCM) and the BITS protocol was 76 to 79 percent faster and utilized 53 to
58 percent less bandwidth.
 When transferring a file from a Microsoft Office SharePoint Server 2007 (MOSS
2007), clients running Windows 7 with BranchCache were 65 to 68 percent faster and
utilized 66 percent less bandwidth than Windows XP or Windows Vista.
The chart below (Figure 1, Page 2) illustrates the bandwidth savings for the three
operating systems tested across four business use scenarios. These levels of bandwidth
savings can have a direct and immediate effect on operating expenses, especially in
organizations where WAN costs are growing quickly.
BranchCache Savings
Bandwidth Utilization
35
30
Average MB Transferred
25
20 Windows 7
15 Windows XP
Windows Vista
10
5
0
HTTP MOSS SCCM SMB
Figure 1 – BranchCache Savings for Bandwidth Utilization.

(Smaller is better.)
For businesses looking to improve the productivity of workers in remote offices,

downloading files in less than one third the time previously required can be directly
translated into greater worker productivity and job satisfaction. (See Figure 2, Page 3 for
a chart illustrating potential time savings.)
BranchCache Savings
File Download Time
Average Seconds to Download File 250
200
150 Windows 7
Windows XP
100 Windows Vista
50
0
HTTP MOSS SCCM SMB
Figure 2 - BranchCache Savings - File Download Time

(Smaller is better)
Based upon the evidence of our tests, Edison believes that for some organizations, the
costs of upgrading to Windows 7 can be significantly offset by bandwidth savings alone.
If organizations are already contemplating PC upgrades to these remote offices,
upgrading to Windows 7 with BranchCache should be considered the new standard.
Introduction
Objective
Since the release of Windows 7 on October 22, 2009, many organizations have been
testing the new operating system. Among the new features that organizations should
strongly consider when making IT purchase or upgrade decisions is BranchCache.
BranchCache is a new feature in Windows 7 and Windows Server 2008 R2 that enables
the acceleration of file download times for users in remote offices over a variety of
networks without requiring additional bandwidth or installation of accelerator
appliances.
Microsoft has published several documents that provide executive and technical
overviews of BranchCache that are available at www.branchcache.com. This white
paper does not duplicate the contents of those documents. Instead, it is intended to
provide third-party validation of Microsoft’s performance claims for BranchCache,
while also providing some hard metrics that readers can use in determining how
BranchCache can lower bandwidth costs and improve the productivity of workers in
satellite offices.
Audience
This white paper should be read by technically savvy business decision makers who are
concerned with Wide Area Network costs and remote office productivity. It will also be
useful for those organizations that are evaluating Windows 7 and assessing strategies for
upgrading their installed base of Windows XP and Vista clients.
Contents of this Report
This white paper contains the following sections:
The Remote Office Dilemma

This section describes the effect on productivity experienced by remote workers because
of long file transfer delays when accessing commonly used files or documents from a
centrally located server share, web server, SharePoint Server, or Configuration
Management Server.
Testing BranchCache Performance
This section describes the test scenarios and testing methodology Edison developed and
used. It includes a high-level discussion of the lab environment, the configurations, and
test methodologies.
Test Results
This section provides a summary of our test results, showing how BranchCache affects
file transfer performance within the scenarios tested.
Conclusions and Recommendations

This section provides Edison’s analysis of the benefits of BranchCache and
recommendations for implementation strategies.
Appendices
The appendices contain details of the test environment and methodology, as well as the
test result details.
The Remote Office Dilemma
The Dilemma
It was just a few years ago that organizations recognized the need to address the issues
of server sprawl – the proliferation of file and application servers across their corporate
landscape. Among the first steps taken were physical consolidation of servers from
distributed computer rooms and closets to more and more centralized data centers. This
change has resulted in significant cost reductions, especially with the emergence of
server virtualization as the ultimate consolidation tool.
Consolidation success has exposed a new concern: as servers that used to be “just down
the hall” are moved to “across the continent,” the speed of file access to which users
have been accustomed has been adversely affected by network latency and the lower
speed of Wide Area Networks (WANs).
In addition, the nature of server access and the files users are downloading have
changed as well. Where small Microsoft Word documents and Microsoft Excel
spreadsheets of less than 1 MB in size once predominated, today mixed media
documents containing images, movies, audio, and video are becoming more and more
common. These much larger sized files are also likely to be delivered via Microsoft
Office SharePoint Server 2007 or via other corporate websites.
The lower speed and higher latency of WANs and the much larger file sizes being
transferred have had consequences that can negate some of the advantages of
consolidation. This possibility can require organizations to increase the bandwidth of
their WAN connections (often at a very high cost), bring back local servers (thus
reviving server sprawl), or invest in Wide Area Application Acceleration Appliances.
While the last of these options, acceleration appliances, can be beneficial where business-
critical applications need acceleration, these devices can add significant cost and
complexity to the management of remote offices.
The BranchCache Solution
Microsoft BranchCache is designed to provide the file transfer performance

organizations are looking for without the need for costly WAN upgrades or additional
hardware to manage. BranchCache improves performance when copying files between
central and remote or branch offices using the SMB, HTTP, and HTTPS protocols. SMB
is the protocol used for transferring files between shared folders on Windows networks.
HTTP and HTTPS are the protocols used by web browsers and many other applications.
The concept behind BranchCache is straightforward: when BranchCache is enabled, a

copy of data accessed from an Intranet web, SharePoint, or file server is cached locally in
the branch office. When another client on the same network in the branch office requests
the same file, the file is transferred from the local cache rather than from across the
WAN.
BranchCache operates in two modes:

 Distributed Cache. In this mode, Windows 7 client computers utilize a peer-to-peer
architecture. The first client to download a file caches a copy of the file and sends it
directly to other clients as they request it.
 Hosted Cache. In this mode, content is cached to a computer on the branch network
running Windows Server 2008 R2 as the Hosted Cache. Other clients needing the
same content retrieve it directly from the Hosted Cache. Computers running the
Hosted Cache can run the Server Core installation option of Windows 2008 R2 and
can host other applications.
Figure 3 illustrates these modes:
Figure 3 - BranchCache modes (from Microsoft)
The BranchCache caches are passive: data is only retrieved when a client requests it.
This means that using BranchCache does not increase WAN utilization. Since only read
requests are cached, write operations are not affected. BranchCache operations also
respect security policies and technologies, such as SSL and IPSec.
BranchCache requires the use of Windows 7 Ultimate or Enterprise editions on
workstations and Windows Server 2008 R2 on servers. For more complete list of servers
that support BranchCache, please visit: http://technet.microsoft.com/en-
us/library/ee307962(WS.10).aspx.
Testing BranchCache Performance
Edison established a lab environment for testing BranchCache performance claims that
use of cached content will result in better user productivity and higher return on
investment. The lab was configured to test performance over a range of use cases and
test conditions. These use cases are simulations of typical branch office user interactions
with central office servers. The test conditions utilize a WAN emulator to demonstrate
the effect of different WAN networks. Tests compared the effect of Windows 7 with
BranchCache on network bandwidth and end user download times in comparison to the
older Microsoft operating systems Windows XP Professional and Windows Vista
Business.
These tests are a simulation designed to illustrate the benefits of using BranchCache.
Real-world implementations of BranchCache will experience different results that are
dependent upon many factors, including the configuration of the PCs in the branch
offices, the design of the branch office LAN, the configuration and provisioning of the
WAN, and many other factors. In addition, these tests were run using late beta or release
candidate versions of the operating systems and some of the server products tested. In
keeping with past experience with pre-released software, performance with the released
versions may be better than the results Edison has documented.
Test Methodology
As has been previously described, BranchCache accelerates performance of the SMB,

HTTP, and HTTPS protocols, offering two branch office configuration modes —
Distributed Cache and Hosted Cache. In order to provide use cases that illustrate these
protocols, Edison defined several scenarios and devised a testing methodology that
would compare the network bandwidth used and end user download times with
BranchCache in Windows 7 in both modes and in comparison to the older Microsoft
operating systems Windows XP Professional and Windows Vista Business.
The files transferred averaged 20 MB in size. This size was chosen to require measurable
and differential transfer times and bits transferred readings on the WAN emulator.
Transfer time was measured with a stop watch or the transferring application, where
applicable.
Test Conditions
Each set of tests is performed with the WAN emulator configured to emulate three
different WAN conditions. The table below (Table 1 - WAN Conditions and Use Cases,
Page 10) lists and describes the three WAN conditions with real-world use cases for each
condition.
Table 1 - WAN Conditions and Use Cases

WAN Conditions Real-World Use Case
1.5 Mbps from HQ to remote office, 384
Regional (up to 1,500 mile) VPN over
Kbps from remote office to HQ, 40 ms
DSL line at remote office
round trip latency .01% packet loss
New York to Los Angeles MPLS VPN or
1.5 Mbps, 95 ms latency, .02% packet loss
leased line
5 Mbps, 195 ms latency, .05% packet loss Transpacific MPLS link
These line conditions were chosen based on AT&T MPLS service level agreements.
Test Scenarios
The following are the test scenarios used for this project. Each of the four scenarios, run
in Hosted Cache and Distributed Cache modes, was tested in a lab environment
consisting of five desktop PCs. For each scenario, a brief description of the scenario and
the use case context is followed by a description of the specific test methodology used.
References made to specific hosts, drives, and so forth are explained in the Lab
Configuration section which follows.
Scenario 1 – CIFS/SMB File Access

When users download a file from a folder on a file server to their client PC, the Server
Message Block (SMB) protocol is used. Over the years, Microsoft has added features to
SMB, usually under the Common Internet File System (CIFS) name. For the purposes of
this white paper, SMB and CIFS are used interchangeably. This test scenario measured
the time and WAN bandwidth required for this file transfer process.
A file share on the server (named Content Server) was configured to generate
BranchCache hashes as per the Microsoft deployment guide. 1 Each workstation was
mapped to this share as Drive N. Microsoft Office PowerPoint 2007 was opened on each
1Microsoft BranchCache Deployment Guide can be accessed on the Microsoft TechNet site at:
http://technet.microsoft.com/en-us/library/ee649232%28WS.10%29.aspx
workstation and the file opened. A stopwatch was used to measure the time from when
the open button was clicked until the image of the first slide appeared on the screen.
This process was repeated on each workstation for multiple files; the workstations were
then rebooted into the next operating system, and the process repeated.
Scenario 2 – HTTP File Accesses

The HTTP protocol is the protocol of the World Wide Web and is used by web browsers
such as Internet Explorer 8 and many other applications to transfer content over the
Internet. In the business world, many corporate Intranets host web servers from which
files, such as presentations, forms, and so forth, are regularly transferred using the
HTTP protocol. In this scenario, files to be transferred are stored on the HTTP or web
server and accessed through a link on a web page.
Microsoft Office SharePoint Server 2007 (MOSS) also uses the HTTP protocol. In our test
scenarios, we are treating MOSS as a separate scenario. This was done for two reasons:
First, since MOSS uses a database to store the files, the time required for file transfer
could be affected by latency inherent in the SharePoint architecture. Second, the secure
version of HTTP, HTTPS, needed to be represented, as well. Since many MOSS
installations utilize HTTPS, it was felt that tests for the HTTPS protocol using MOSS was
a reasonably realistic scenario.
For this scenario, the Content Server was prepared as a web server for BranchCache as
described in the BranchCache Early Adopter’s Guide 2 and the BranchCache
Deployment Guide. An index.htm file of links to the TIFF files described above was
placed on the default web site of the server.
The workstations were then booted into Windows XP. On each workstation a TIFF file
was selected in Internet Explorer 8 and Save Target As was selected from the context
menu. The file was saved to the C:\TEMP folder. The transfer times reported by Internet
Explorer were used for our results.
This process was repeated for each of the workstations in turn waiting for the transfer
on each station to complete before starting the next transfer. The process was repeated
with two additional files to identify anomalous results.
2The Microsoft BranchCache Early Adopters Guide is available at

http://technet.microsoft.com/en-us/library/dd637762%28WS.10%29.aspx (at the time of writing).
The process was repeated with the workstations running Windows Vista and Windows
7 using the Distributed Cache mode. Finally the BranchCache Group Policy Objects were
changed to enable the Hosted Cache mode, and additional files were downloaded.
The amount of data transferred in the download was recorded by resetting the counters
on the Apposite Linktropy 4500 as the Save button was clicked on the workstation and
noting the value at the end of the download. As there is handshaking traffic between the
two networks at all times, variations in data transfer values that represent data rates
below 20 Kbps are considered insignificant.
Scenario 3 – SCCM
System Center Configuration Manager (SCCM), a member of Microsoft’s System Center
family of Information Technology management solutions, is a tool for the
comprehensive assessment, deployment, and updating of servers, client computers, and
devices across physical, virtual distributed, and mobile environments. SCCM provides a
range of services to an organization. Among these is the distribution of software
application packages to client computers for installation.
When transferring these packages, SCCM uses a combination of the SMB protocol and
the Background Intelligent Transfer Service (BITS) protocol. The BITS protocol is used to
facilitate prioritized, throttled, and asynchronous transfer of files using idle network
bandwidth. The BITS protocol can make use of BranchCache to accelerate file transfers,
but the protocol itself is not affected by doing so. The prioritization, throttling, and other
BITS features are still in effect, so the degree of acceleration experienced by BITS
transfers is not as great as for the other scenarios. At the time of testing, SCCM with
BITS was not supported in Hosted Cache mode, so no testing was performed for this
configuration. The release version of BranchCache and the updated SCCM now support
Hosted Cache mode. Edison expects that the performance gains experienced in our other
tests would carry over to SCCM in Hosted Cache Mode.
For this scenario, self-extracting ZIP files were each packaged as an Advertised
Application via SCCM. Using the Advertised Applications Control Panel applet, each
ZIP file was downloaded to the workstation and the time to download recorded.
Scenario 4 – MOSS 2007

Microsoft Office SharePoint Server 2007 is Microsoft’s fastest growing product ever.
Companies who are running Microsoft Windows Servers, Microsoft Windows clients,
and Microsoft Office productivity applications are implementing SharePoint servers at
an incredible pace. If a company is not yet using a full-blown MOSS 2007 server farm,
chances are they are using the free Windows SharePoint Services which provides many
of the same features without the infrastructure and management overhead associated
with MOSS.
As tested, getting a file from a SharePoint server is the same as getting one from any
web server. For this scenario a document repository of PowerPoint .PPTX files was
created, and they were downloaded as in the HTTP test. As previously described, the
HTTPS protocol was used to simulate the typical use of that secure protocol. Otherwise,
the test methodology was exactly the same as for the HTTP scenario.
Test Data
A series of 14 scanned photos in TIFF format, each 21-22.5 Mb in size, were saved on the
Content server. For SMB testing, each photo was also inserted into a PowerPoint 2007
.PPTX file. For SCCM, each TIFF file was converted into a self-extracting ZIP file using
WinZip. The file names and sizes appear in the appendix.
Laboratory Configuration
The following is the laboratory configuration for this project (see Figure 4, page 14).
Whether noted explicitly below or not, all operating systems, server, and productivity
applications had the latest service packs and patches applied.
At Emulated Headquarters
1 Physical server (Dell 2950, dual 1.86 GHz quad core Xeon processors, 16 GB
memory)
 Windows Server 2008 Enterprise Edition R2 w/Hyper-V with the following virtual
servers
 VM1 – Content Server – Windows Server 2008 R2
 IIS, Valid HTTPS certificate
 Domain controller for TestDomain.local
 VM2 – SQL08A
 System Center Configuration Manager 2007 SP2 Virtual Machine (SCCM)
(Downloadable from Microsoft)
 VM3 –
 Windows Server 2008 R2
 Microsoft Office SharePoint Server 2007 (MOSS 2007)
Gateway® DX4200-09 with AMD Phenom™ X4 9100e (1.80 GHz, quad core) processor
and 4 GB memory
 Windows Server 2008 R2
 SCCM server
PowerEdge 2950
2 5
0 3 6
1 4 7
HQ File, web and SCCM

Server
Remote Remote Remote
PC 3 PC 2 PC 1
Shunra WAN
Emulator
Remote
PC 4 10/100 Network
10/100 Network
Switch Switch
Remote
PC 5 PowerEdge 1950
0
1
2
Remote Office Server
Remote Office Headquarters
Figure 4 - Test Network Configuration Diagram
Remote Office
1 Physical Server
 Acer Veriton 460G PCs. (Intel Core 2 Duo E7500 Processor, 3 GB memory, 160 GB
hard drive, and Marvell Gigabit Ethernet.)
Windows Server 2008 R2 Enterprise Edition providing Domain Controller, DHCP, and
hosted BranchCache services to the five workstations.
5 Workstations
 Acer Veriton 460G PCs. (Intel Core 2 Duo E7500 Processor, 3 GB memory, 160 GB
hard drive, and Marvell Gigabit Ethernet.)
 Each workstation has three partitions, multiple boot configuration
 All configurations saved to disk images w/Ghost for easy reversion
 Partition 1
 Windows XP SP3 fully updated via Windows Update
 Microsoft Office 2007 SP2
 Partition 2
 Windows 7
 Partition 3
 Windows Vista Business SP2
 A second hard drive holds an image of the three working partitions created with
Acronis True Image, allowing systems to be reverted to a pristine state after each test
run.
Network
 The PCs are connected to an unmanaged Dell 10/100 Ethernet switch.
 Subnet is 172.17.128.0/24 (an RFC 1918 reserved private network).
 Apposite Linktropy 4500 WAN Emulator which routes traffic to the headquarters
network (10.16.252.0/24 also an RFC 1918 reserved net).
 Belkin IP-enabled KVM switch.
Test Results
Summary of Results
Edison believes that the best way to view the results of our testing is from the
perspective of how much less time and how much less WAN bandwidth in MB is
required to transfer files to a branch office using BranchCache compared to a previous
version of the operating system (Windows Vista or Windows XP).
The following charts illustrate the average BranchCache savings on file transfer time and
bandwidth utilization for the four test scenarios in all test conditions. As can be seen in
Figure 5 the download times for Windows 7 with BranchCache show consistent
download times and are significantly faster than for the earlier operating systems.
BranchCache Savings
File Download Time
250
Average Seconds to Download File
200
150 Windows 7
Windows XP
100 Windows Vista
50
0
HTTP MOSS SCCM SMB
Figure 5 - BranchCache Savings on File Download Time

(Smaller is better.)
The bandwidth savings delivered by Windows 7 with BranchCache enabled are, as seen
in Figure 6, are also significant, though less consistent due to the differing nature of the
file transfer protocols being accelerated. 3
BranchCache Savings
35
30
Average MB Transferred
25
20 Windows 7
15 Windows XP
Windows Vista
10
5
0
HTTP MOSS SCCM SMB
Figure 6 - BranchCache Savings for Bandwidth Utilization

(Smaller is better)
For HTTP, file download times with Windows 7 with BranchCache enabled was over 60
percent faster than Windows XP or Windows Vista. Windows 7 with BranchCache
enabled required almost one half the bandwidth than either Windows XP or Windows
Vista.
When file downloads were performed using MOSS 2007, the need for MOSS to access
SQL Server slowed down the file downloads somewhat, but the file download using
Windows 7 with BranchCache enabled still required approximately two-thirds less time
and two-thirds less bandwidth than Windows XP or Windows Vista.
3 See the scenario descriptions above or the results details below for more information on how the
transfer protocols in the four scenarios can affect file transfer time and bandwidth utilization.
Downloads through SCCM exhibited a 76 to 79 percent download time reduction for
Windows 7 with BranchCache enabled and over a 53 to 58 percent bandwidth savings
over Windows XP or Windows Vista.
CIFS/SMB file download times were from 55 to 69 percent faster for Windows 7 with
BranchCache enabled and utilized from 45 to 59 percent less bandwidth than Windows
XP or Windows Vista.
The following table (Table 2) shows the performance in time and bandwidth
summarized above.
Table 2 Average Results for Scenarios

HTTP MOSS SCCM SMB
Time
Windows 7 11.50 34.20 45.86 48.52
Windows XP 107.51 108.06 189.33 158.44
Windows Vista 108.06 98.50 214.17 108.44
Windows 7 Advantage over
Windows XP 63% 68% 76% 69%
Windows Vista 63% 65% 79% 55%
Bandwidth
Windows 7 11.50 7.72 13.29 9.46
Windows XP 22.53 22.80 31.71 22.97
Windows Vista 22.45 22.91 28.50 17.14
Windows XP 49% 66% 58% 59%
Windows Vista 49% 66% 53% 45%
As can be seen in Table 2, the effect of BranchCache on file download time is significant
enough in productivity savings, but the effect on bandwidth can have the greatest effect
on operational cost savings.4
Scenarios
The sections below provide summaries of the effect of Windows 7 with BranchCache
enabled for each of the four test scenarios. The data upon which these summaries are
based is presented in the appendices. All of the tests in these scenarios were performed
using the set of test conditions described in the lab configuration section. These tests
included both asymmetric and symmetric WAN emulation. Asymmetric networks are
typical of those provided by DSL and cable networks. In these networks, the download
speed is usually greater than the upload speeds.
Since the results of BranchCache acceleration for both Hosted and Distributed cache
modes were very similar and the performance of Windows XP and Windows Vista were
also similar the charts only show some of the results of the testing. This choice was made
so that the charts would be easier to read. Detailed results tables can be found in the
appendices.
Test Scenario 1 – CIFS/SMB File Accesses

The CIFS/SMB File Access Test consists of file downloads initiated from within
Microsoft PowerPoint 2007 for files on a network share (mounted as Drive N). The times
were based upon the time required to open the file.
Results
The following charts show the average performance comparison between Windows 7
with BranchCache enabled versus the performance of Windows XP and Windows Vista.
The Average Bandwidth Used chart (Figure 8, Page 20) and the Average Download
Time chart (Figure 8, Page20) show that Windows 7 with BranchCache enabled requires
much less time and bandwidth than either Windows Vista or Windows XP. It should be
noted that Windows Vista had improvements in CIFS/SMB protocol processing, which is
shown by the better performance of Vista over XP.
4It is important to note that when a cached file is transferred, it is the performance of the caching
computer and Local Area Network that is affecting transfer time. Almost no Wide Area Network
bandwidth is being used for the transfer other than that required for determining the existence
and state of a previously transferred and cached file and checking credentials and access controls.
Figure 8 - Average Bandwidth Used: Figure 8 - Average Download Time:
CIFS/SMB Protocol (Smaller is better) CIFS/SMB Protocol (Smaller is better)
The two charts on the following pages show the results of acceleration on CIFS/SMB file
downloads. The charts show that downloads to XP or Vista clients take consistent
amounts of time 5 and use the same bandwidth for each client transfer, and that
Windows 7 with BranchCache greatly accelerates the download.
The different test conditions – network speed and latency – have a definite effect on the
speed of all the transfers: slower speeds require longer transfer times (Figure 9 - SMB
Test Results: Transfer Time, Page 21). The test conditions have no effect on Windows 7
bandwidth utilization, as can be seen by the overlap for all Windows 7 results (Figure 10
- SMB Results: Bandwidth Utilization, Page 22). Also clearly seen is the BranchCache
effect on speed and bandwidth, with markedly sharp falloffs in both metrics.
The download time for Windows 7 Distributed Mode for the Symmetrical 5 Mbps WAN
condition does not show as great a savings, as the faster network speed shortened the
time required for the initial file download, while the shortest and most accelerated
download time approached the limits of the available bandwidth for all three WAN
conditions.
5The slightly shorter time illustrated for the third client is an anomaly of chart preparation. For
clarity purposes, only the first of two test runs for the first three clients was selected for the chart.
For the SMB scenario, the performance of the third client was slightly faster than for the previous
two clients. The fourth and fifth clients and second test run were more consistent than this
particular download instance.
Figure 9 - SMB Test Results: Transfer Time
Figure 10 - SMB Results: Bandwidth Utilization
Test Scenario 2 – HTTP File Downloads
The HTTP File Access Test consists of file transfers initiated from a page hosted on an
HTPP or web server.
Results
The chart, Average Download Time (Figure 11) illustrates the 63 percent time savings
delivered by Windows 7 with BranchCache over the nearly identical download times
(see appendices for details) required by Windows Vista and Windows XP.
Figure 11 - Average Download HTTP Protocol

(Smaller is Better)
The bandwidth savings for Windows 7 with BranchCache enabled were showed savings
of 49 percent as compared to the bandwidth required by Windows Vista or Windows
XP. (Figure 12)
Figure 12 - Average Bandwidth Used: HTTP
Protocol (Smaller is better)
The following two charts show the results of acceleration on HTTP file transfers (Figure
13 - HTTP Test Results: Transfer Time, Page 25 and Figure 14 - HTTP Test Results:
Bandwidth Utilization, Page 26). The charts show that file downloads to Windows XP or
Windows Vista clients take consistent amounts of time and use the same bandwidth for
each client transfer, and that Windows 7 with BranchCache greatly accelerates the file
download.
speed of all the file downloads: slower speeds require longer transfer times. This is
evident for the 5 Mbps downloads for Windows Vista where the download time is less
than half for the slower networks. It also is apparent with the faster initial client
performance for Window 7 with BranchCache enabled. The acceleration provided by
BranchCache is not as profound over the faster network, at least for a file size that is
small relative to the connection speed.
The test conditions have no effect on BranchCache bandwidth utilization, as can be seen
in the near total overlap of the BranchCache results. BranchCache is equally efficient
across the full range of test conditions utilized in this study.
HTTP Test Results
Transfer Time
140
120 Asymmetrical
1.5Mbps/384Kbps
Windows Vista
100 Symmetrical
1.5Mbps Windows
Vista
Time in Seconds
80 Symmetrical 5Mbps
Windows Vista
60 Asymmetrical
1.5Mbps/384Kbps
Windows 7
Distributed Cache
40 Symmetrical
1.5Mbps Windows 7
Distributed Cache
Windows 7
Distributed Cache
0
1 2 3
Clients
Figure 13 - HTTP Test Results: Transfer Time
HTTP Test Results
0.025
Asymmetrical
0.02 1.5Mbps/384Kbps
Windows Vista
Symmetrical
1.5Mbps Windows
0.015 Vista
Symmetrical 5Mbps
MB Transferred
Windows Vista
0.01
Asymmetrical
1.5Mbps/384Kbps
Windows 7
Distributed Cache
0.005 Symmetrical
1.5Mbps Windows 7
Distributed Cache
Symmetrical 5Mbps
0 Windows 7
Distributed Cache
-0.005
Clients
Figure 14 - HTTP Test Results: Bandwidth Utilization
Test Scenario 3 – SCCM Application Package Download
Microsoft System Center Configuration Manager (SCCM) enables the automated
transfer of application installation and patch files to end-user computers. SCCM utilizes
the BITS throttling mechanism to manage the transfers, while the files themselves are
transferred using the SMB protocol. The BranchCache Acceleration Affect only operates
on the actual SMB file transfer aspect of an SCCM transaction, so the results are not as
marked as for the two previous (HTTP and MOSS) tests.
At the time of testing, SCCM with BITS was not supported in Hosted Cache mode, so no
testing was performed for this configuration. 6 The release version of BranchCache and
an updated SCCM now support Hosted Cache mode. Edison expects that the
performance gains experienced in our other tests will carry over to SCCM in Hosted
Cache Mode.
Results
SCCM download times for Windows 7 with BranchCache enabled were 76 percent faster
than for Windows XP and 79 percent faster than for Windows Vista, as can be seen in
the chart (Figure 15 – Average Download Time: SCCM, Page 27). Application and
operating system installations and patching can be very disruptive of productivity.
Savings of this order of magnitude can be a major productivity boost.
Figure 15 - Average Download Time: SCCM
6The inability to test SCCM in hosted mode is what drove the decision to show only Windows 7
with BranchCache enabled in distributed mode in these charts.
Bandwidth utilization was also much lower with Windows 7 with BranchCache enabled
than with the older operating systems, though perhaps not so spectacularly. (Figure 16 –
Average Bandwidth Used: SCCM, Page 28) Nevertheless, Windows 7 used 53 percent
less bandwidth than Windows Vista and 58 percent less bandwidth than required by
Windows XP.
Figure 16- Average Bandwidth Used: SCCM
The following two charts show the results of acceleration on downloads of SCCM
Application Packages (Figure 17 - SCCM Test Results: Transfer Time, Page 30 and Figure
18 - SCCM Test Results: Bandwidth Utilization, Page 31). As in the other scenarios, the
charts show that transfers to XP or Vista clients take consistent amounts of time and use
the same bandwidth for each client transfer, and that Windows 7 with BranchCache
greatly accelerates the data transfer.
speed of all the transfers: slower speeds require longer transfer times. The effect of the
higher network speeds is less pronounced with SCCM. Edison attributes this difference
to the differences in the BITS protocol as compared to the other protocols tested.
The download for Windows Vista under Asymmetrical conditions, as would be

expected, utilized more bandwidth than downloads for Windows Vista under the
symmetrical conditions. This was due to the longer time required for the client to
communicate back to the SCCM server under asymmetrical conditions, thus slowing the
entire process. This example shows a clear advantage for Windows 7 clients using
BranchCache for OS and application downloads or patching from an organization’s
SCCM servers when there is an asymmetric connection. The savings in bandwidth from
over 30 MBs of data transferred to nearly zero data transferred can have a considerable
positive effect on branch office WAN utilization.
SCCM Test Results
Transfer Time
350
300 Asymmetrical
1.5Mbps/384Kbps
Windows Vista
250 Symmetrical 1.5Mbps

Windows Vista

Time in Seconds
Windows Vista
150 Asymmetrical
1.5Mbps/384Kbps
Windows 7
Distributed Cache
100 Symmetrical 1.5Mbps
Windows 7
Distributed Cache
Windows 7
Distributed Cache
0
1 2 3
Clients
Figure 17 - SCCM Test Results: Transfer Time
SCCM Test Results
Bandwidth Utiliztion
45
40
Asymmetrical
1.5Mbps/384Kbps
35 Windows Vista
Symmetrical 1.5Mbps
30 Windows Vista
MB Transferred
Windows Vista
20
Asymmetrical
1.5Mbps/384Kbps
Windows 7
15 Distributed Cache
Symmetrical 1.5Mbps
Windows 7
10 Distributed Cache
Symmetrical 5Mbps
5 Windows 7
Distributed Cache
0
1 2 3
Clients
Figure 18 - SCCM Test Results: Bandwidth Utilization
Test Scenario 4 – MOSS 2007 File Access
Microsoft Office SharePoint Server 2007 enables users to access files from a database-
supported and application-driven portal. In this test scenario, files were downloaded
using the same technique as was used for the HTTP tests in Scenario 2. The only
differences were that MOSS access was over the secure HTTPS protocol, and the files
themselves were stored in the MOSS database.
Results
The BranchCache savings in time and bandwidth were very symmetrical. As illustrated
in the chart (Figure 20 – Average Bandwidth Used: MOSS 2007) bandwidth savings were
the same 66 per cent for Windows 7 as compared to Windows XP or Windows Vista.
Time savings for Windows 7 with BranchCache enabled over Windows XP were 68
percent. The savings over Windows Vista were 65 percent (Figure 20 – Average
Download Time: MOSS 2007).
Figure 20 - Average Bandwidth Used: MOSS Figure 20 - Average Download Time: MOSS
2007 (Smaller is better.) 2007 (Smaller is better.)
The chart (Figure 21 - MOSS 2007 Test Results: Transfer Time, Page 34) shows the results
of acceleration on MOSS 2007 HTTPS file downloads. The chart shows that file
downloads to Windows Vista clients take virtually identical times for the slower
asymmetrical and symmetrical network conditions without noticeable variation during
the tests. This showed that there was little to no communication to the MOSS server once
the download commenced. The performance for Windows Vista at 5 Mbps shows some
speed variation. As in a previous test, this variation was an anomaly of the chart making
procedure: the download to second client for the first test run was slightly slower than
for the other clients in either test run. Since the chart only shows three runs, this
difference is more obvious than it would be if all the test runs were shown.
Bandwidth utilization for file downloads with MOSS 2007 (Figure 22 - MOSS 2007 Test
Results: Bandwidth Utilization, Page 35) was very consistent for both operating systems
illustrated. The chart line overlaps reveal that the two 1.5 Mbps downloads to Windows
Vista utilized virtually identical bandwidth, while the 5 Mbps Windows Vista download
utilized slightly less bandwidth, but the results under all three conditions were very
similar in utilization. Once again, the chart line overlap shows that Windows 7 with
BranchCache results were nearly identical for all three test conditions, quickly
approaching zero WAN bandwidth utilization.
Organizations with MOSS 2007 servers hosting remote offices connected over a range of
network conditions will see a significant productivity boost due to faster file download
speeds. Bandwidth costs and contention will also be positively affected by running
Windows 7 clients with BranchCache enabled.
Figure 21 - MOSS 2007 Test Results: Transfer Time
Figure 22 - MOSS 2007 Test Results: Bandwidth Utilization
Conclusions and Recommendations
Edison analysts created and ran a series of tests designed to provide quantified data that
would illustrate the effects of the file transfer acceleration provided by using Windows 7
and Windows Server 2008 R2 with BranchCache enabled. The result of Edison’s testing
shows that BranchCache delivers on its promise. File downloads were, on average, 69
percent faster with Windows 7 and BranchCache than for Windows XP and 66 percent
faster than for Windows Vista. Bandwidth utilization averaged 58 percent less for
Windows 7 versus Windows XP and 53 percent less than Windows Vista.
With the vast range of possible configurations and tariffs, Edison has made no attempt
to calculate the potential financial savings attributable to lower bandwidth utilization,
but it should be obvious that the significant bandwidth savings possible with
BranchCache could have an ameliorating affect on operating system upgrade costs,
possibly sufficient to justify branch office upgrades.
Edison believes that the results support two recommendations for organizations
contemplating upgrading their end-user operating systems to Windows 7. The first of
these is to stage upgrades to central servers and branch offices. Upgrade remotely
accessible servers to Windows Server 2008 R2 and remote branch computers to use
Windows 7, adding branch servers using Windows Server 2008 R2 where necessary.
Distributed Cache mode allows IT professionals to take advantage of BranchCache with
minimal hardware deployments in the branch office. However, if the branch has
deployed other infrastructure (for example, servers running other workloads) using
Hosted Cache mode may be beneficial for increased cache availability and caching for
the entire branch office. (Microsoft suggests that Distributed Cache mode can be
adequate for offices of fewer than 50 users on a subnet)
The second recommendation is to consider the replacement of shared folder network

designs with SharePoint. Whether an organization utilizes Windows SharePoint Services
(WSS, available at no charge) or the more sophisticated and centrally manageable
Microsoft Office SharePoint Server (MOSS), the productivity benefits of higher transfer
speeds and the lower bandwidth costs can offset many SharePoint adoption costs while
providing the recognized business value of the SharePoint portal paradigm.
Appendices
Test Environment
This section consists of a detailed listing of the equipment and software utilized in
testing, including applicable configuration settings details.
Lab Configuration
Remote Office Workstations (5)
Model Acer Verton 460G

CPU Intel Core 2 Duo E7500 (2.93 MHz)
Memory 4 GB
Windows XP Professional
Operating Systems Windows Vista Business
Windows 7 Enterprise
Hard Drive 160 GB Western Digital Caviar Blue SATA II
Remote Office Switch

Dell PowerConnect 2216 unmanaged 10/100 16 ports
Remote Office Server
Model Acer Veriton 460G

CPU Intel Core 2 Duo E7500 (2.93 MHz)
Memory 4 GB
Operating Systems Windows Server 2008 R2 Enterprise
Hard Drive 160 GB Western Digital Caviar Blue SATA II
Active Directory Domain Controller
IIS web server
Roles and Features
BranchCache Hosted Cache server
File Server
WAN Emulator
Apposite Linktropy 4500 WAN Emulator configured as router
Virtual Server Host
Model Dell PowerEdge 2950

CPU Intel Xeon E5320 (Quad core 1.83 GHz)
Memory 16 GB
Operating Systems Windows Server 2008 Datacenter with Hyper-V
Hard Drive 4x73 GB SAS on PERC 5i RAID controller
HQ2008R2 – A File and Web server used as the data source
Virtual Machines for SMB and HTTP testing
SQL08A – A Windows Server 2003R2 server running SQL
Server 2008 hosting the SCCM and SharePoint databases.
SCCM and MOSS hosts
Model Gateway DX4200-09

CPU AMD Phenom™ X4 9100e (1.80 GHz, quad core)
Memory 4 GB
Operating Systems Windows Server 2008 r2 Enterprise
Hard Drive 160 GB Western Digital Caviar Blue
HQ2008R2 – A Windows Server 2008R2 Enterprise system
providing file and web services used as the data source for
Virtual Machines
SMB and HTTP testing
SQL08A – A Windows Server 2003R2 server running SQL
Server 2008 hosting the SCCM and SharePoint databases.
Test Results
The following pages present the results of Edison’s testing. The results are presented for
each test scenario in the form of a summary table that lists each test condition and
operating system, and in detail tables showing the results of each test run.
In the summary tables, the results for each platform, for each WAN condition are shown
in one table. A second table shows the percent advantage of Windows 7 with
BranchCache enabled versus the legacy operating systems; Windows Vista and
Windows XP.
In the details tables, file transfer duration is presented in minutes and seconds (MM:SS).
Data transferred is presented in MB or KB as measured during transferred through the
WAN emulator.
CIFS/SMB Test Results
Results Summary
The test results show that Windows 7 with BranchCache enabled is 58 percent faster and
requires 45 percent less bandwidth than Windows Vista. The results also show that
Windows 7 with BranchCache enabled is 69 percent faster while requiring 59 percent
less bandwidth than Windows XP.
The performance advantages are not as great when compared to the results from some
other scenarios. Edison believes the performance difference is due to the way the
application (PowerPoint 2007) interacts with the network while retrieving the file. While
the files were being transferred, it appeared that the first data blocks were being sent
from the remote server while BranchCache was checking whether the file existed in a
local cache. When the cached file was identified, network transfers stopped and the
remainder of the file was transferred from the branch cache. Presuming our observation
matches BranchCache design, it is probable that transferring larger files would
demonstrate a greater difference in speed and bandwidth utilization.
CIFS/SMB Scenario Test Results
Bandwidth
File Download Utilization
WAN Condition/OS Time in Seconds MB Transferred
Windows 7
Asymmetrical 1.5 Mbps/384 Kbps
Windows 7 Distributed Cache 56.75 9.56
Windows 7 Hosted Cache 54.13 9.09
Symmetrical 1.5 Mbps Windows 7
Distributed Cache 58.63 9.21
Symmetrical 1.5 Mbps Windows 7 Hosted
Cache 52.13 9.11
Symmetrical 5 Mbps Windows 7
Symmetrical 5 Mbps Windows 7 Hosted
Cache 33.63 10.14
Windows XP
Windows XP 154.00 23.05
Symmetrical 1.5 Mbps Windows XP 160.67 22.93
Symmetrical 5 Mbps Windows XP 160.67 22.93
Windows Vista
Asymmetrical 1.5 Mbps/384Kbps
Windows Vista 128.50 23.50
Symmetrical 1.5 Mbps Windows Vista 127.33 5.06
Symmetrical 5 Mbps Windows Vista 69.50 22.87
Average bandwidth and download times by platform

Platform Average Bandwidth Used Average Download Time
W7 w. BC Enabled 9.46 48.52
Windows XP 22.97 158.44
Comparison of Windows 7 with Windows Vista and Windows XP
Windows 7 Advantage Average Bandwidth Used Average Download Time
Versus Vista 45% 58%
Versus XP 59% 69%
Details
The following tables show the results for all the test runs for each WAN bandwidth
condition. The files on the remote server were pre-cached, so only the first Windows 7
workstation to download the file utilized WAN bandwidth. The other Windows 7
workstations downloaded the file from the branch workstation’s cache. Test runs ceased
after three successful transfers occurred without significant deviation in performance
(indicated as NTR in the table).
Asymmetrical 1.5 Mbps/384 Kbps Windows XP
1.TIF Transfer 1.TIF Data 2.TIF Transfer 2.TIF Data

Workstation Time Transferred Time Transferred
1 2:35 22.6 MB 2:33 23.5 MB
2 2:34 22.7 MB 2:34 23.4 MB
3 2:35 22.6 MB 2:34 23.5 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR
Asymmetrical 1.5 Mbps/384 Kbps Windows Vista

1 2:09 23.6 MB 2:08 23.4 MB
2 2:09 23.6 MB 2:08 23.4 MB
3 2:08 23.6 MB 2:09 23.4 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR
Asymmetrical 1.5 Mbps/384 Kbps Windows 7 Distributed Cache

1 2:11 22.6 MB 2:14 23.4 MB
2 0:18 3 MB 0:29 4.5 MB
3 0:32 5.7 MB 0:41 6 MB
4 0:30 4.5 MB 0:39 6.8 MB
5 NTR NTR NTR NTR
Asymmetrical 1.5 Mbps/384 Kbps Windows 7 Hosted Cache

1 2:14 23.9 MB 2:07 22.0 MB
2 0:24 3.4 MB 0:29 4.5 MB
3 0:25 3.4 MB 0:35 5.6 MB
4 0:24 3.4 MB 0:35 5.6 MB
5 NTR NTR NTR NTR
Symmetrical 1.5 Mbps Windows XP

1 3:01 23.0 MB 3:06 23.6 MB
2 3:00 23.2 MB 3:06 23.6 MB
3 3:01 23.1 MB 3:06 23.6 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR
Symmetrical 1.5 Mbps Windows Vista

1 2:08 23 MB 2:10 23.6 MB
2 2:07 23.1 MB 2:10 23.6 MB
3 2:07 23.1 MB 2:11 23.6 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR
Symmetrical 1.5 Mbps Windows 7 Distributed Cache

1 2:12 23.1MB 2:14 22.7 MB
2 0:25 3.4 MB 0:25 3.4 MB
3 0:38 5.6 MB 0:49 7.2 MB
4 0:25 3.4 MB 0:31 4.2 MB
5 NTR NTR NTR NTR
Symmetrical 1.5 Mbps Windows 7 Hosted Cache

1 1:05 23.2 MB 1:11 22.5 MB
2 0:27 4.5 MB 0:24 5.6 MB
3 0:29 5.6 MB 0:30 8.9 MB
4 0:35 6.7 MB 0:06 35 KB
5 NTR NTR NTR NTR
Symmetrical 5 Mbps Windows XP

1 2:44 22.9 MB 2:33 23.0 MB
2 2:43 22.9 MB 2:39 23.1 MB
3 2:42 22.8 MB 2:43 23.0 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR
Symmetrical 5 Mbps Windows Vista

1 1:33 22.9 MB 1:26 22.9 MB
2 1:34 22.8 MB 1:00 22.9 MB
3 1:07 22.8 MB 0:49 22.9 MB
4 1:04 22.8 MB 0:56 23 MB
5 1:06 22.8 MB 1:00 22.9 MB
Symmetrical 5 Mbps Windows 7 Distributed Cache

1 1:05 23.2 MB 1:11 24.3 MB
2 0:27 4.5 MB 0:24 4.0 MB
3 0:29 5.6 MB 0:30 5.8 MB
4 0:35 6.7 MB 0:30 5.7 MB
5 NTR NTR NTR NTR
Symmetrical 5 Mbps Windows 7 Hosted Cache

1 1:07 23 MB 0:57 23.6 MB
2 0:20 5.6 MB 0:28 7.8 MB
3 0:24 5.6 MB 0:16 3.4 MB
4 0:23 6.6 MB 0:34 5.5 MB
5 NTR NTR NTR NTR
HTTP Test Results
Results Summary
The test results show that Windows 7 with BranchCache enabled is 63 percent faster
than either Windows Vista or XP. Bandwidth savings for Windows 7 versus Windows
Vista or Windows XP was 49 percent.
HTTP Scenario Test Results

Bandwidth
File Download Utilization MB
WAN Condition/OS Time in Seconds Transferred
Windows 7
Asymmetrical 1.5 Mbps/384 Kbps Windows 7
51.40 9.00
Distributed Cache
Asymmetrical 1.5 Mbps/384 Kbps Windows 7
53.80 9.13
Hosted Cache
Symmetrical 1.5 Mbps Windows 7 Distributed
50.60 8.99
Cache
53.80 9.13
Cache
Symmetrical 5 Mbps Windows 7 Distributed
18.50 9.14
Cache
Symmetrical 5 Mbps Windows 7 Hosted Cache 11.50 23.63
HTTP Scenario Test Results
Bandwidth
Windows Vista
Asymmetrical 1.5 Mbps/384 Kbps Windows XP 123.70 22.58
Windows XP
Asymmetrical 1.5 Mbps/384 Kbps Windows
121.50 22.57
Vista

W7 w. BC Enabled 11.50 39.93
Windows XP 22.53 107.51

Versus XP 49% 63%
Details

1 2:07 22.6 MB 2:00 22.5 MB
2 2:06 22.8 MB 2:01 22.5 MB
3 2:07 22.6 MB 2:01 22.4 MB
4 2:08 22.7 MB 2:00 22.6 MB
5 2:06 22.6 MB 2:01 22.5 MB
Asymmetrical 1.5Mbps/384Kbps Windows Vista

1 1:59 22.3 MB 2:04 22.9 MB
2 2:00 22.3 MB 2:02 22.8 MB
3 2:00 22.4 MB 2:03 22.8 MB
4 2:01 22.3 MB 2:02 22.8 MB
5 2:00 22.3 MB 2:04 22.8 MB

1 2:06 22.4 MB 2:02 22.5 MB
2 2:04 22.4 MB 2:01 22.5 MB
3 0:04 15 KB 0:03 18 KB
4 0:04 22 KB 0:03 27 KB
5 0:04 67 KB 0:03 21 KB

1 2:05 23.0 MB 2:11 22.6 MB
2 2:03 22.9 MB 2:08 22.7 MB
3 0:06 9 KB 0:05 14 KB
4 0:05 11 KB 0:05 33 KB
5 0:05 5 KB 0:05 41 KB

1 1:52 22 MB 2:01 23 MB
2 2:02 22 MB 2:02 22 MB
3 2:01 23 MB 2:01 22. MB
4 2:01 22 MB 2:02 23 MB
5 2:02 22 MB 2:01 22 MB

1 2:07 22 MB 2:02 22. MB
2 2:07 22 MB 2:03 22 MB
3 2:08 22 MB 2:03 22 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 2:03 22 MB 2:02 22.9 MB
2 2:02 22 MB 2:02 22.8 MB
3 0:02 7 KB 0:03 12 KB
4 0:03 41 KB 0:03 54 KB
5 0:03 19 KB 0:03 34 KB

1 2:05 23.0 MB 2:11 22.6 MB
2 2:03 22.9 MB 2:08 22.7 MB
3 0:06 9 KB 0:05 14 KB
4 0:05 11 KB 0:05 33 KB
5 0:05 5 KB 0:05 41 KB

1 1:26 22.8 1:21 22.6
2 1:16 22.8 1:19 22.7
3 1:11 22.7 1:17 22.6
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 2:07 22 MB 2:02 22 MB
2 2:07 22 MB 2:03 22 MB
3 2:08 22 MB 2:03 22 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 0:38 23.15 0:47 22.4
2 0:41 23.7 0:41 22.5
3 0:03 21 K 0:03 34 K
4 0:03 47 K 0:03 43 K
5 0:03 23 K 0:03 48 K

1 0:41 23.1 MB 0:42 22.8 MB
2 0:48 23.2 MB 0:46 22.9 MB
3 0:03 .012 KB 0:03 0
4 0:03 8 KB 0:03 12 KB
5 NTR NTR NTR NTR
The larger variations in transfer times on this test are caused by the higher packet loss
rate. Since TCP/IP drops its net utilization when a packet is lost, packets lost close
together or at the end of a file transfer have smaller effects than those spread out or at
the end of the file transfer. Vista and Windows 7 improved TCP/IP stacks recover faster
and handle latency better on long fat pipes.
MOSS2007 Test Results
Results Summary
than Windows Vista, and 68 percent faster than Windows XP. Bandwidth savings for
Windows 7 versus Windows Vista or Windows XP was 66 percent.
MOSS 2007 Test Results

Bandwidth
Windows 7
Windows 7 Hosted Cache 32.75 5.71
Cache 32.75 5.72
Symmetrical 5 Mbps Windows 7 Hosted
Cache 16.22 7.53
Windows XP
Windows XP 124.00 23.10
Windows Vista
W7 w. BC Enabled 7.72 34.20
Windows XP 22.80 108.06

Versus XP 66% 68%
Details

1 2:07 23.7 MB 2:01 22.5 MB
2 2:07 23.7 MB 2:01 22.5 MB
3 2:07 23.7 MB 2:01 22.5 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 2:07 23.6 MB 2:01 22.6 MB
2 2:07 23.7 MB 2:01 22.6 MB
3 2:07 23.6 MB 2:01 22.6 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 2:07 23.6 MB 2:01 22.7 MB
2 2:06 23.5 MB 2:00 22.4 MB
3 0:04 6 kb 0:03 4 KB
4 0:03 12 kb 0:03 22 KB
5 0:03 27 kb 0:03 19 KB

1 2:02 23 MB 2:02 22.6 MB
2 0:03 22 KB 0:03 11 KB
3 0:03 4 KB 0:03 5 KB
4 0:03 13 KB 0:03 16 KB
5 NTR NTR NTR NTR

1 2:03 22.7 MB 2:04 22.8 MB
2 2:03 22.7 MB 2:04 22.8 MB
3 2:04 22.8 MB 2:03 22.8 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 2:07 23.6 MB 2:01 22.6 MB
2 2:07 23.6 MB 2:01 22.6 MB
3 2:07 23.7 MB 2:02 22.6 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 2:01 22.8 MB 2:02 22.8 MB
2 2:01 22.6 MB 2:01 22.6 MB
3 0:03 2.4 KB 0:03 6 KB
4 0:03 11 KB 0:03 8 KB
5 0:03 108 KB 0:03 16 KB

1 2:03 22.9 MB 2:01 22.6 MB
2 0:03 226 KB 0:03 0
3 0:03 0.59 KB 0:03 22 KB
4 0:03 25 KB 0:03 14 KB
5 NTR NTR NTR NTR

1 1:17 22.8 MB 1:20 22.2 MB
2 1:16 22.8 MB 1:16 22.3 MB
3 1:17 22.8 MB 1:23 22.3 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 0:42 22.8 MB 0:46 22.1 MB
2 0:53 22.9 MB 0:48 22.3 MB
3 0:46 22.8 MB 0:49 22 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR
SCCM Test Results
Results Summary
than Windows Vista and 79 percent faster than Windows XP. Bandwidth savings for
Windows 7 versus Windows Vista was 58 percent. Windows 7 required 53 percent less
bandwidth than Windows XP.
SCCM Test Results
Bandwidth
Windows 7
Windows XP
Windows XP 190.67 29.28
Windows Vista
Asymmetrical 1.5 Mbps/384Kbps

W7 w. BC Enabled 13.29 45.86
Windows XP 28.50 214.17

Versus XP 53% 79%
Details
The following tables show the results for all the test runs for each WAN bandwidth
condition. The files on the remote server were pre-cached, so only the first Windows 7
workstation to download the file utilized WAN bandwidth. The other Windows 7
workstations downloaded the file from the branch workstation’s cache. Test runs ceased
after three successful transfers occurred without significant deviation in performance
(indicated as NTR in the table).
At the time of testing SCCM was not supported in Hosted mode. These results are
shown as N/A.
Asymmetrical 1.5Mbps/384Kbps Windows XP

1 2:39 24.9 MB 3:31 32.6
2 2:47 25.1 MB 3:31 32.6
3 3:05 27.9 MB 3:31 32.6
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 3:27 32.6 MB 2:33 23.9 MB
2 3:26 32.6 MB 2:34 23.9 MB
3 3:27 32.6 MB 2:32 23.9 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 4:01 32.2 MB 3:31 32.6
2 0:11 166 KB 0:28 0.041
3 0:37 39 KB 0:21 0.085
4 0:28 12 KB 0:46 0.009
5 0:26 153 KB NTR NTR

1 N/A N/A N/A N/A
2 N/A N/A N/A N/A
3 N/A N/A N/A N/A
4 N/A N/A N/A N/A
5 N/A N/A N/A N/A

1 4:25 39.8 MB 2:59 26.6 MB
2 4:22 39.7 MB 3:00 26.5 MB
3 4:25 39.7 MB 3:00 26.6 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 4:22 31.8 MB 2:42 16.6 MB
2 4:22 31.9MB 2:40 16.5MB
3 4:25 31.9 MB 2:37 16.5 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 5:03 39.6 MB 4:50 25.6 MB
2 0:21 1.7 MB 0:22 42 KB
3 0:56 65 KB 0:34 49 KB
4 0:27 114 KB 0:27 36 KB
5 NTR NTR NTR NTR

1 N/A N/A N/A N/A
2 N/A N/A N/A N/A
3 N/A N/A N/A N/A
4 N/A N/A N/A N/A
5 N/A N/A N/A N/A

1 2:38 33.5 MB 2:26 31.9 MB
2 2:42 33.5 MB 2:34 31.9 MB
3 2:45 35.0 MB 2:28 31.8 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 4:30 31.8 MB 2:42 16.6 MB
2 4:21 31.9 MB 2:40 16.5 MB
3 4:30 31.9 MB 2:37 16.5 MB
4 NTR NTR NTR NTR
5 NTR NTR NTR NTR

1 1:37 16.4 MB 1:41 16.7 MB
2 0:21 1.1 MB 0:23 1.2 MB
3 0:22 42 KB 0:22 42 KB
4 0:14 32 KB 0:15 34 KB
5 NTR NTR NTR NTR

1 N/A N/A N/A N/A
2 N/A N/A N/A N/A
3 N/A N/A N/A N/A
4 N/A N/A N/A N/A
5 N/A N/A N/A N/A

2010 Samples Branch Cache

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

2010 Samples Branch Cache

Încărcat de

Drepturi de autor:

Formate disponibile

89 Fifth Avenue, 7th Floor

New York, NY 10003

Enhanced Branch Office Productivity

All products are trademarks of their respective owners.

Executive Summary ..................................................................................................................... 1

Organizations seeking to lower operating expenses are investigating technology

With the release of Windows 7, Microsoft has delivered BranchCache™, a solution to

The concept behind BranchCache is straightforward: when BranchCache is enabled, a

Microsoft commissioned Edison Group to examine WAN bandwidth savings with

Figure 1 – BranchCache Savings for Bandwidth Utilization.

For businesses looking to improve the productivity of workers in remote offices,

Average Seconds to Download File 250

Figure 2 - BranchCache Savings - File Download Time

Contents of this Report

This white paper contains the following sections:

The Remote Office Dilemma

Conclusions and Recommendations

The BranchCache Solution

Microsoft BranchCache is designed to provide the file transfer performance

The concept behind BranchCache is straightforward: when BranchCache is enabled, a

BranchCache operates in two modes:

Figure 3 illustrates these modes:

Figure 3 - BranchCache modes (from Microsoft)

As has been previously described, BranchCache accelerates performance of the SMB,

Table 1 - WAN Conditions and Use Cases

Scenario 1 – CIFS/SMB File Access

Scenario 2 – HTTP File Accesses

2The Microsoft BranchCache Early Adopters Guide is available at

Scenario 4 – MOSS 2007

HQ File, web and SCCM

Remote Office Server

Remote Office Headquarters

Figure 4 - Test Network Configuration Diagram

Figure 5 - BranchCache Savings on File Download Time

Figure 6 - BranchCache Savings for Bandwidth Utilization

Table 2 Average Results for Scenarios

Test Scenario 1 – CIFS/SMB File Accesses

Figure 11 - Average Download HTTP Protocol

Figure 13 - HTTP Test Results: Transfer Time

Figure 14 - HTTP Test Results: Bandwidth Utilization

Figure 15 - Average Download Time: SCCM

Figure 16- Average Bandwidth Used: SCCM

The download for Windows Vista under Asymmetrical conditions, as would be

250 Symmetrical 1.5Mbps

200 Symmetrical 5Mbps

Figure 17 - SCCM Test Results: Transfer Time

Figure 18 - SCCM Test Results: Bandwidth Utilization

The second recommendation is to consider the replacement of shared folder network

Remote Office Workstations (5)

Model Acer Verton 460G

Remote Office Switch

Remote Office Server

Model Acer Veriton 460G

Virtual Server Host

Model Dell PowerEdge 2950

SCCM and MOSS hosts

Model Gateway DX4200-09

CIFS/SMB Test Results

Average bandwidth and download times by platform

Asymmetrical 1.5 Mbps/384 Kbps Windows XP

1.TIF Transfer 1.TIF Data 2.TIF Transfer 2.TIF Data

Asymmetrical 1.5 Mbps/384 Kbps Windows Vista

1.TIF Transfer 1.TIF Data 2.TIF Transfer 2.TIF Data

1.TIF Transfer 1.TIF Data 2.TIF Transfer 2.TIF Data