HP Bc1500 - BladeSystem - Blade PC Manual

1
CCI Network Considerations Guide
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4
Anatomy of the CCI solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5
CCI, network speed, and the connection protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6
CCI network traffic patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Network reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Network bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8
Network latency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8
Connection methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
The HP BladeSystem PC Blade Switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
Internal Ethernet ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
External Ethernet ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11
Port speed, duplex, and flow control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Maximizing network cable reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Virtual Local Area Network (VLAN) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Management IP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13
Switch diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Switch management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Embedded Web System interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15
Link aggregation groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15
Spanning tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
PVST interoperability mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
Disabling the spanning tree PVST Interoperability mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17
SNMP and remote monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17
Configuring the switch to send SNMPv1traps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17
Configuring HP SIM to work with the switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18
HP Session Allocation Manager (SAM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
HP SAM minimum network requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19

2
The blade PC allocation process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20
Blade image management considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
The blade PC image PXE boot process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21
CCI solution component dependencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21
General notes regarding solution Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
CCI network topology reference designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
General notes regarding Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
Flat network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24
Redundant connectivity with IEEE 802.1D or 802.1w . . . . . . . . . . . . . . . . . . . . . . . . . . .25
Flat network with Per-VLAN Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26
Basic connectivity with per-VLAN Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
Redundant connectivity with per-VLAN Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . .28
Standards based redundancy with IEEE 802.1s (MSTP) . . . . . . . . . . . . . . . . . . . . . . . . . . 29
VLAN load balancing with IEEE 802.1s (MSTP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .30
For more information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .31
Appendix A: HP PC Blade Switch startup-config samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Default (canned startup-config) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .32
No STP (canned startup-config) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
STP/RSTP (canned startup-config) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
MSTP (canned startup-config) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Flat network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .35
Redundant connectivity with IEEE 802.1D or 802.1w . . . . . . . . . . . . . . . . . . . . . . . . . . .35
Flat Network with Per-VLAN Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Basic Connectivity with Per-VLAN Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Redundant Connectivity with Per-VLAN Spanning Tree . . . . . . . . . . . . . . . . . . . . . . . . . . .37
Standards Based Redundancy with IEEE 802.1s (MSTP) . . . . . . . . . . . . . . . . . . . . . . . . . . 37
VLAN Load Balancing with IEEE 802.1s (MSTP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Appendix B: Running the HP PC Blade Switch and the C-GbE Interconnect Switch
in a co-existing network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Before you deploy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Switch default configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
PVST interoperability - theory of operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Disabling the spanning tree PVST Interoperability mode . . . . . . . . . . . . . . . . . . . . . . . . . .42
Appendix C: Remote desktop considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42
Console display settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Streaming video . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42
Network printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Local printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3
This document provides an overview of the HP BladeSystem PC Blade Switch, as well as an explanation
of network terminology, configuration concepts, and examples relevant to the HP CCI solution. This docu-
ment is intended for those seeking technical networking knowledge of how CCI is applied to the enter-
prise network. To learn even more about the HP CCI Solution, refer to the CCI documentation or visit the
CCI Web site at www.hp.com/go/cci.
The HP PC Blade Switch is an industry-standard managed layer-2 Ethernet switch that is designed to dra-
matically reduce the number of Ethernet network cables required to provide redundant network connectiv-
ity to HP Blade PCs hosted by the HP PC Blade Enclosure. HP has tuned the default configuration of the
HP PC Blade Switch beyond a typical network edge switch to facilitate integrating the CCI solution into an
existing network, typically with minimal effort.
Existing networks that adequately support a traditional distributed desktop PC environment are fully capa-
ble of supporting a CCI implementation. However, because every network environment and CCI imple-
mentation will differ, important variables must be considered on a case-by-case basis. Even though CCI is
a proscribed solution, integrating it into an enterprise network takes skill, careful thought, and planning.
This document is designed to assist in that effort. Additionally, HP Services is available to assist with such
planning and design efforts.
Introduction
The HP CCI solution includes a rack mountable 3U (5.25 inch) HP PC Blade Enclosure that hosts up to 20
HP blade PCs. Each enclosure has a switch tray which includes an Integrated Administrator and redun-
dant, hot-pluggable power supplies and cooling fans. Each blade PC is populated with two Broadcom
5705F Fast Ethernet Embedded 10/100Megabits per second (Mbps) network controllers. A fully popu-
lated HP PC Blade Enclosure can require up to five Ethernet connections, four of which can be either giga-
bit copper or fiber, and the fifth fast Ethernet copper only. A typical 42U rack can house as many as 14
HP PC Blade Enclosures for a maximum of 280 blade PCs. Without the HP PC Blade Switch, the number
of Ethernet connections within this space can quickly become difficult to manage.

4
The HP PC Blade Switch is a managed Layer 2+ Ethernet switch that provides up to a 41-to-1 reduction in
Ethernet network cables and provides network connection redundancy for the HP PC Blade Enclosure.
Cable reduction can help to reduce the time required to deploy, manage, and service the HP CCI solu-
tion.
Anatomy of the CCI solution
As shown in Figure 1, the CCI topology is as follows:
•In the Access Tier, a user accesses an available blade PC from a remote access device.
•In the Compute Tier, for a dynamic environment implementation, the Session Allocation Manager
allocates the user direct access to either a new or an existing blade PC session.
A new session consists of a user without either an active or disconnected remote desktop session
running on a blade PC within the CCI solution.
An existing session consists of a blade PC resource that is either:
• Actively hosting the user’s session from a different access device.
or
• A session that the user has disconnected from (either the same or different access device)
without first logging off.
•In the Resource Tier, users authenticate with the directory server each time they attempt to log onto
a blade PC. After successful authentication, users are logged onto the blade PC with access to the
resources in the Resource Tier allowed by their security level.
•The Management Tier is reserved for the tools necessary to manage the CCI solution, such as HP
Systems Insight Manager and the HP Rapid Deployment Pack.

5
Figure 1 CCI topology
CCI, network speed, and the connection protocol
Network speed is really all about “how much longer than instantaneous does it take a signal to travel a
given route?” Anything more than instantaneous is “latency.” So, the question is how much latency is in a
network supporting CCI, and at what point does this latency impact user experience?
One of the advantages of CCI over competitive offerings and traditional computer systems is that the
remote host does not download files to the local access device. Instead, the remote host transmits only a
bitmap representation of the blade PC screen to the access device. Even at a dial-up network speed of 48
Kilobits per second (Kbps), you can open large files utilizing the blade PC almost instantly while viewing
them remotely. For example, if a user on a traditional system wants to open a 10-MB email attachment,
the user must launch an email application on the remote computer to gain access to email, and then liter-
ally wait for hours for the email attachment to download over a dial-up connection. This download of the
file to the remote computer introduces a security risk. Alternatively, the CCI solution allows the remote user

6
to access a blade PC located in the corporate data center to run an email application and manipulate the
10-MB file without download the file to the remote access device. In this case, access to read and edit
such a file is achievable within a reasonably expected amount of time.
An understanding of the transport software used by CCI is imperative before trying to understand how
network latency can impact end users experience. The transport software used to move signals back and
forth between the access device and the blade PC in the data center is called the Remote Desktop Con-
nection (RDC). This software uses Microsoft Remote Desktop Protocol (RDP) or any other remote desktop
connection client compatible with Microsoft Terminal Services.
RDP is a multi-channel protocol that allows you to connect to a computer running Microsoft Terminal Ser-
vices. RDP software exists for most versions of Windows, as well as other operating systems such as Linux.
The first version of RDC (version 4.0), introduced with Terminal Services in Windows NT 4.0 Server, was
based on the ITU T.share protocol (T.128). Terminal Server Edition Version 5.0, introduced with Windows
2000 Server, included support for a number of features including localized printer support and was
intended to improve utilization of network bandwidth. Version 5.1 was introduced with Windows XP and
includes features such as support for 24-bit color and sound.
Microsoft’s Remote Desktop Protocol software is available as a free download for other Microsoft operat-
ing systems at: http://www.microsoft.com/downloads/details.aspx?FamilyID=33ad53d8-9abc-4e15-
a78f-eb2aabad74b5&DisplayLang=en).
You can find additional information about RDP at:
•Remote Desktop Protocol - from the Microsoft Developer Network
(http://msdn.microsoft.com/library/en-us/termserv/termserv/remote_desktop_protocol.asp)
•Understanding the Remote Desktop Protocol - from support.microsoft.com
(http://support.microsoft.com/default.aspx?scid=kb;EN-US;q186607)
•Remote Desktop Connection for Windows Server 2003 - latest version of Microsoft’s free client for
Windows 95 and upwards
(http://www.microsoft.com/downloads/details.aspx?displaylang=en&familyid=a8255ffc-
4b4a-40e7-a706-cde7e9b57e79)
•Remote Desktop Connection Client for Mac - Microsoft’s free client for Mac OS X
(http://www.microsoft.com/mac/otherproducts/otherproducts.aspx?pid=remotedesktopclient)
•rdesktop - free open source client for Unix platforms
(http://www.rdesktop.org/)
You can optimize RDP for a variety of connection speeds. For instance, You can define less rich color and
screen resolution to reduce network traffic. For more information about optimizing your RDP settings, refer-
ence Microsoft RDP settings for HP’s Consolidated Client Infrastructure Environment.

7
CCI network traffic patterns
Traditional distributed PCs have historically created peak network traffic at specific times. Traffic tradition-
ally peaks within the first 30 minutes of the workday as users log in and large amounts of data move
between the datacenter and end users. During this time, MS Exchange data, system hot fixes, etc.,
traverse the network. Afterward, network demand tends to drop substantially.
In contrast, because CCI computers (blade PCs) are powered on continuously in the datacenter under
control of the datacenter manager, there is a very different pattern of network traffic for the following rea-
sons:
•All traditional network traffic from datacenter to distributed PC is internal to the datacenter.
•Data center managers deploy image updates including application updates, operating system ser-
vice packs, and hot fixes. Managers can schedule these updates at times when network traffic is low.
•Most datacenters have a higher capacity internal network infrastructure compared to the infrastruc-
ture between the datacenter and users.
•The typical CCI network traffic pattern from the datacenter to the user is very steady and much lower
than the peaks seen with traditional PCs. Instead of sending files to the user, CCI uses RDC to render
what the blade is doing on the user’s screen.
Based on CCI network observations, HP recommends the following guidelines for the number of active
users per network segment while providing a substantial buffer for extraordinary peaks in demand.
•10,000 simultaneous active users per 1 Gigabit per second network segment.
•1,000 simultaneous active users per 100 Megabits per second network segment.
•100 simultaneous active users per 10 Megabits per second network segment.
While these are general guidelines that work for most infrastructures, differing circumstances and local
variables can require analysis on a case-by-case basis.

8
Network reliability
Businesses require reliable networks, which are imperative to support solutions that successfully address
business needs. In CCI, network failure affects user access to virtualized desktops. Therefore, HP recom-
mends that you use redundant design practices to provide multiple network paths to all centralized com-
puting resources. Carefully set expectations of how minor to catastrophic network outages will affect the
end user environment. Consider customer business rules and job functions when establishing these expec-
tations.
Network bandwidth
The term “bandwidth” in network design refers to the transfer data rate supported by a network connec-
tion or interface and is commonly expressed in terms of bits per second (bps). The term comes from the
field of electrical engineering, where bandwidth represents the total distance between the highest peak
and lowest valley of a communication signal (band). Network bandwidth is one of the only factors that
determines the perceived speed of a network.
Network latency
Network latency describes the time (in milliseconds) that it takes a packet to traverse a network. An easy
way to measure network latency is with the MS DOS “PING” command. This command tests any network
device that has a valid IP address. A successful ping yields a message displaying response time. Because
CCI end users access their blade PC sessions across some distance, maintaining low network latency has
a direct affect on user experience. A few factors that can increase network latency that you should con-
sider when evaluating a CCI network include:
•poor quality connections
•network congestion
•network saturation
Consider conducting a network traffic study to determine the level of saturation a connection may encoun-
ter, and take steps (well documented outside the scope of this paper) to alleviate the underlying condi-
tions.

9
Connection methods
CCI end users can connect to their blade PC using a network in a variety of ways: traditional dial-up
modems, DSL, cable modems, LANs, etc. General performance with RDC (including CCI and any net-
work architecture that uses MS-RDC) is based on response times as characterized in the following table:
Based on testing by the HP CCI team, the implications of users working off the corporate LAN and access-
ing the CCI from a remote location should yield the following experiences:
•Broadband users: The experience when connecting to a corporate WAN/LAN should be similar to
being connected directly to the corporate LAN with no noticeable latency. However, every network
and network connection is different, so other considerations such as congestion or poor connection
quality can impact the end user experience. All testing to date indicates that CCI performance across
broadband connections is very similar to being on a Local Area Network.
•Analog dial-up users: For users who connect to their corporate network using a analog dial-up
modem, round-trip latency of 100ms to 200ms is common, and latency higher than 200ms is not
unheard of. In this situation, CCI works well for reading email, working in spreadsheets, creating text
documents, etc. However, latency above 200ms can become an inhibitor when performing complex
tasks, such as creating PowerPoint documents or working with graphically intense applications. In
case-by-case scenarios where the remote user only has access to an analog connection, a higher
quality modem and/or custom modem initialization string might bring the latency down to a more
acceptable level. You should involve someone technically skilled at troubleshooting analog network
connections.
Response Time
(Round Trip Ping) Typical Network Type End User Experience
<1 ms - 10 ms LANs or WANs No discernible latency; same experience as if the
user is using a traditional desktop PC (all other things
equal)
10 ms - 20 ms Broadband modems/bridges
10 ms - 100 ms Busy LANs or WANs Some latency, but rarely discernible (might be detect-
able if rapidly scrolling Power Point in full screen
mode, for example)
100 ms - 200 ms Dial-up modems Discernible latency (for example, when rapidly scroll-
ing PowerPoint or typing quickly, screen updates
might not appear for a second or two). Reducing
screen resolution and color depth can improve end
user experience.
> 200 ms Congested Networks Substantial latency (for example, when rapidly scroll-
ing PowerPoint or typing quickly, letters might not
appear on the screen for several seconds or more,
scrolling text documents is “jumpy,” etc.). Using RDC
with latency exceeding 200ms is generally not a sat-
isfactory experience.

10
•Additional considerations on dial-up modems: Analog modems vary in quality. Higher quality
56kbps modems may produce lower latency than lower quality modems. Because analog modems
are designed to retrain their speed depending on line condition, you may need to fine tune modem
configuration or seek a more robust method to access the network, such as an xDLS broadband con-
nection or dedicated leased-line.
The HP BladeSystem PC Blade Switch
The HP PC Blade Switch is in the interconnect tray for the HP PC Blade Enclosure that uses a non-blocking,
managed Layer 2+ Ethernet switch (see Figure 2). The HP PC Blade Switch consolidates 40 Ethernet NICs
from blade PCs and provides four external dual personality (Copper or Fiber) Gigabit “uplink” Ethernet
ports. You can configure the HP PC Blade Switch to combine all Ethernet signals into one physical uplink
(see “Maximizing network cable reduction” on page 12) or up to four physical uplinks configured as a
single logical uplink using IEEE 802.3ad link aggregation (see “Link aggregation groups” on page 15).
Figure 2 HP PC Blade Switch tray external panel
Item Description
1, 4 RJ-45 Auto MDIX, 10/100/1000T Gigabit Ethernet Uplink Ports e43 and e44
(Assigned to VLAN 2 by default)
2, 3 Small Form-factor Pluggable (SFP) Port for Ethernet Uplink GBIC Connector to Ports e43 and e44
(Assigned to VLAN 2 by default)
5 RJ-45 Auto MDIX, 10/100T Fast Ethernet Uplink Port e42
(Assigned to VLAN 1 by default)
6 DB-9 RS-232 Serial Port for Integrated Administrator console access
7 Integrated Administrator Reset Button
8, 11 RJ-45 Auto MDIX, 10/100/1000T Gigabit Ethernet Uplink Ports e45 and e46
(Assigned to VLAN 1 by default)
9, 10 Small Form-factor Pluggable (SFP) Port for Ethernet Uplink GBIC Connector to Ports e45 and e46
(Assigned to VLAN 1 by default)
12 3 4 5 6
7
8 9 10 11