The Virtual LAN The Virtual LAN

decisys
The
The Virtual
Virtual LAN
LAN
Technology
Technology Report
Report
The Virtual LAN Technology Report
Contents
Introduction
Defining VLANs
Membership by Port Group
Membership by MAC Address
Layer 3–Based VLANs
IP Multicast Groups as VLANs
Combination VLAN Definitions
Automation of VLAN Configuration
Communicating VLAN Membership Information
Standards and the Proprietary Nature of VLANs
VLAN Implementation Benefits
Reducing the Cost of Moves and Changes
Virtual Workgroups
Reduction of Routing for Broadcast Containment
Routing Between VLANs
VLANs Over the WAN
Security
VLANs and ATM
VLANs Transparent to ATM
Complexity Arising with ATM-Attached Devices
LAN Emulation
Routing Between Emulated LANs and/or VLANs
Edge Routing
The One-Armed Router
The Route Server
MPOA
VLANs and DHCP: Overlapping Solutions
DHCP Functionality
Best Use for Each
Overlap Between DHCP and VLANs
VLAN Architectures Going Forward
Infrastructural VLANs
Service-Based VLANs
VLAN Migration Strategies
Conclusion
2
3
3
3
4
5
5
5
6
6
7
7
8
9
10
11
11
11
11
11
11
13
14
14
14
15
15
16
16
16
17
17
18
19
20
Copyright©1996.
1
The Virtual LAN Technology Report
by David Passmore and John Freeman
Introduction
Virtual LANs (VLANs) have recently
developed into an integral feature of switched
LAN solutions from every major LAN
equipment vendor. Although end-user
enthusiasm for VLAN implementation has yet
to take off, most organizations have begun to
look for vendors that have a well-articulated
VLAN strategy, as well as VLAN functionality built into products today. One of the
reasons for the attention placed on VLAN
functionality now is the rapid deployment of
LAN switching that began in 1994/1995.
The shift toward LAN switching as a
replacement for local/departmental routers—
and now even shared media devices (hubs)—
will only accelerate in the future. With the
rapid decrease in Ethernet and Token Ring
switch prices on a per-port basis, many more
ambitious organizations are moving quickly
toward networks featuring private port (single
user/port) LAN switching architectures. Such a
desktop switching architecture is ideally suited
to VLAN implementation. To understand why
private port LAN switching is so well suited to
VLAN implementation, it is useful to review
the evolution of segmentation and broadcast
containment in the network over the past
several years.
In the early 1990s, organizations began to
replace two-port bridges with multiport, collapsed backbone routers in order to segment
their networks at layer 3 and thus also contain
broadcast traffic. In a network using only
routers for segmentation, segments and
broadcast domains correspond on a one-to-one
basis. Each segment typically contained
between 30 and 100 users.
With the introduction of switching, organizations were able to divide the network into
smaller, layer 2–defined segments, enabling
increased bandwidth per segment. Routers
could now focus on providing broadcast containment, and broadcast domains could now
span multiple switched segments, easily supporting 500 or more users per broadcast
domain. However, the continued deployment
David Passmore is president
and co-founder of Decisys, Inc.,
a Sterling, Virginia–based consulting firm specializing in
network design, architecture,
and management for end-user
organizations, and network
product marketing and strategic
planning for vendors. Before
founding Decisys, David was
vice president of the Gartner
Group and a partner in Ernst &
Young’s Center for Information
Technology and Strategy in
Boston, Massachusetts.
David received a B.S. in
computer science and engineering and an M.S. in electrical engineering and
computer science, both from
the Massachusetts Institute of
Technology.
2
of switches, dividing the network into more
and more segments (with fewer and fewer
users per segment) does not reduce the need
for broadcast containment. Using routers,
broadcast domains typically remain in the 100
to 500 user range.
VLANs represent an alternative solution
to routers for broadcast containment, since
VLANs allow switches to also contain
broadcast traffic. With the implementation of
switches in conjunction with VLANs, each
network segment can contain as few as one
user (approaching private port LAN switching), while broadcast domains can be as large
as 1,000 users or perhaps even more. In
addition, if implemented properly, VLANs can
track workstation movements to new locations
without requiring manual reconfiguration of IP
addresses.
Why haven’t more organizations deployed
VLANs? For the vast majority of end-user
organizations, switches have yet to be implemented on a large enough scale to necessitate
VLANs. That situation will soon change.
There are, however, other reasons for the
lukewarm reception that VLANs have received
from network users up to now:
• VLANs have been, and are still, proprietary,
single-vendor solutions. As the networking
industry has shown, proprietary solutions are
anathema to the multivendor/open systems
policies that have developed in the migration
to local area networks and the client server
model.
• Despite the frequently quoted numbers illuminating the hidden costs of networking,
such as administration and moves/adds/
changes, customers realize that VLANs have
their own administrative costs, both straightforward and hidden.
• Although many analysts have suggested that
VLANs enhance the ability to deploy centralized servers, customers may look at
enterprise-wide VLAN implementation and
see difficulties in enabling full, high-performance access to centralized servers.
This paper discusses these and other
issues in greater detail, and attempts to
determine the strategic implications that
VLANs, present and future, pose for enterprise
networks.
Defining VLANs
What is a VLAN? With the multitude of
vendor-specific VLAN solutions and implementation strategies, defining precisely what
VLANs are has become a contentious issue.
Nevertheless, most people would agree that a
VLAN can be roughly equated to a broadcast
domain. More specifically, VLANs can be
seen as analogous to a group of end-stations,
perhaps on multiple physical LAN segments,
that are not constrained by their physical
location and can communicate as if they were
on a common LAN.
However, at this point, issues such as the
extent to which end-stations are not constrained by physical location, the way VLAN
membership is defined, the relationship
between VLANs and routing, and the relationship between VLANs and ATM have been
left up to each vendor. To a certain extent these
are tactical issues, but how they are resolved
has important strategic implications.
Because there are several ways in which
VLAN membership can be defined, this paper
divides VLAN solutions into four general
types: port grouping, MAC-layer grouping,
network-layer grouping, and IP multicast
grouping. We will discuss the issue of manual
vs. automatic VLAN configuration, and
describe techniques by which VLANs may be
extended across multiple switches in the
network. Finally, the paper takes a look at the
present state of VLAN standards.
Membership by Port Group
Many initial VLAN implementations defined
VLAN membership by groups of switch
hes
switc
ltiple
u
m
g
nectin
h #2
e con
Switc
3
kplan
c
a
b
2
/
e
n
o
1
b
Back
8
7
6
5
1
#
h
4
c
Swit
3
2
1
4
ports (for example, ports 1, 2, 3, 7, and 8 on a
switch make up VLAN A, while ports 4, 5,
and 6 make up VLAN B). Furthermore, in
most initial implementations, VLANs could
only be supported on a single switch.
Second-generation implementations
support VLANs that span multiple switches
(for example, ports 1 and 2 of switch #1 and
ports 4, 5, 6, and 7 of switch #2 make up
VLAN A; while ports 3, 4, 5, 6, 7, and 8 of
switch #1 combined with ports 1, 2, 3, and 8
of switch #2 make up VLAN B). This
scenario is depicted in Figure 1.
Port grouping is still the most common
method of defining VLAN membership, and
configuration is fairly straightforward.
Defining VLANs purely by port group does
not allow multiple VLANs to include the
same physical segment (or switch port).
However, the primary limitation of defining
VLANs by port is that the network manager
must reconfigure VLAN membership when a
user moves from one port to another.
John Freeman is a senior consultant at Decisys, Inc., where
he specializes in the development of technology marketing and vendor strategies.
John also works with end-user
clients to help them understand
and evaluate emerging technologies and vendor strategies.
Before joining Decisys, John
worked as a consultant in
Japan in the areas of networking and systems integration. He is fluent in
Japanese and is an expert in
the Japanese networking
market.
John holds a B.A. in East Asian
Studies from Harvard
University.
Membership by MAC Address
VLAN membership based on MAC-layer
address has a different set of advantages and
disadvantages. Since MAC-layer addresses
are hard-wired into the workstation’s network interface card (NIC), VLANs based on
MAC addresses enable network managers to
move a workstation to a different physical
location on the network and have that workstation automatically retain its VLAN membership. In this way, a VLAN defined by
MAC address can be thought of as a userbased VLAN.
5
6
7
8
Hub
Hub
Hub
A
VLAN
B
VLAN
Figure 1. VLANs Defined by Port Group
3
One of the drawbacks of MAC
address–based VLAN solutions is the requirement that all users must initially be configured
to be in at least one VLAN. After that initial
manual configuration, automatic tracking of
users is possible, depending on the specific
vendor solution. However, the disadvantage of
having to initially configure VLANs becomes
clear in very large networks where thousands of
users must each be explicitly assigned to a particular VLAN. Some vendors have mitigated
the onerous task of initially configuring MACbased VLANs by using tools that create
VLANs based on the current state of the
network—that is, a MAC address–based
VLAN is created for each subnet.
MAC address–based VLANs that are
implemented in shared media environments
will run into serious performance degradation
as members of different VLANs coexist on a
single switch port. In addition, the primary
method of communicating VLAN membership
information between switches in a MAC
address–defined VLAN also runs into performance degradation with larger-scale implementations. This is explained in “Communicating VLAN Membership Information,” later
in this paper.
Another, but minor, drawback to VLANs
based only on MAC-layer addresses emerges
in environments that use significant numbers
of notebook PCs with some docking stations.
The problem is that the docking station and
integrated network adapter (with its hard-wired
MAC-layer address) usually remain on the
desktop, while the notebook travels with the
user. When the user moves to a new desk and
docking station, the MAC-layer address
changes, making VLAN membership
impossible to track. In such an environment,
VLAN membership must be updated constantly as users move around and use different
docking stations. While this problem may not
be particularly common, it does illustrate some
of the limitations of MAC address–based
VLANs.
Acronyms and
Abbreviations
AAL5
ATM Adaptation Layer Type 5
ASIC
Application-specific integrated
circuit
ATM
Asynchronous Transfer Mode
DHCP
Dynamic Host Configuration
Protocol
ELAN
Emulated LAN
FDDI
Fiber Distributed Data Interface
IPX
Internet Packet Exchange
LANE
LAN Emulation
LEC
LAN Emulation client
LES
LAN Emulation server
Layer 3–Based VLANs
VLANs based on layer 3 information take into
account protocol type (if multiple protocols
4
are supported) or network-layer address (for
example, subnet address for TCP/IP networks)
in determining VLAN membership. Although
these VLANs are based on layer 3 information, this does not constitute a “routing”
function and should not be confused with
network-layer routing.
Even though a switch inspects a packet’s
IP address to determine VLAN membership,
no route calculation is undertaken, RIP or
OSPF protocols are not employed, and frames
traversing the switch are usually bridged
according to implementation of the Spanning
Tree Algorithm. Therefore, from the point of
view of a switch employing layer 3–based
VLANs, connectivity within any given VLAN
is still seen as a flat, bridged topology.
Having made the distinction between
VLANs based on layer 3 information and
routing, it should be noted that some vendors
are incorporating varying amounts of layer 3
intelligence into their switches, enabling
functions normally associated with routing.
Furthermore, “layer 3 aware” or “multi-layer”
switches often have the packet-forwarding
function of routing built into ASIC chip sets,
greatly improving performance over CPUbased routers. Nevertheless, a key point
remains: no matter where it is located in a
VLAN solution, routing is necessary to
provide connectivity between distinct VLANs.
There are several advantages to defining
VLANs at layer 3. First, it enables partitioning
by protocol type. This may be an attractive
option for network managers who are dedicated to a service- or application-based VLAN
strategy. Second, users can physically move
their workstations without having to reconfigure each workstation’s network address—a
benefit primarily for TCP/IP users. Third,
defining VLANs at layer 3 can eliminate the
need for frame tagging in order to communicate VLAN membership between switches,
reducing transport overhead.
One of the disadvantages of defining
VLANs at layer 3 (vs. MAC- or port-based
VLANs) can be performance. Inspecting
layer 3 addresses in packets is more time consuming than looking at MAC addresses in
frames. For this reason, switches that use
layer 3 information for VLAN definition are
generally slower than those that use layer 2
information. It should be noted that this performance difference is true for most, but not
all, vendor implementations.
VLANs defined at layer 3 are particularly
effective in dealing with TCP/IP, but less
effective with protocols such as IPX™,
DECnet®, or AppleTalk®, which do not
involve manual configuration at the desktop.
Furthermore, layer 3–defined VLANs have
particular difficulty in dealing with “unroutable” protocols such as NetBIOS. Endstations running unroutable protocols cannot
be differentiated and thus cannot be defined
as part of a network-layer VLAN.
flexible definition of VLAN membership
enables network managers to configure their
VLANs to best suit their particular network
environment. For example, by using a combination of methods, an organization that utilizes
both IP and NetBIOS protocols could define IP
VLANs corresponding to preexisting IP
subnets (convenient for smooth migration),
and then define VLANs for NetBIOS endstations by dividing them by groups of MAClayer addresses.
Acronyms and
Abbreviations (Cont.)
MAC
Media access control
MPOA
Multiprotocol over ATM
Automation of VLAN Configuration
NIC
Network interface card
Another issue central to VLAN deployment is
the degree to which VLAN configuration is
automated. To a certain extent, this degree of
automation is correlated to how VLANs are
defined; but in the end, the specific vendor
IP Multicast Groups as VLANs
solution will determine this level of autoIP multicast groups represent a somewhat difmation. There are three primary levels of
ferent approach to VLAN definition, although
automation in VLAN configuration:
the fundamental concept of VLANs as
• Manual. With purely manual VLAN configbroadcast domains still applies. When an IP
uration, both the initial setup and all subpacket is sent via multicast, it is sent to an
sequent moves and changes are controlled
address that is a proxy for an explicitly defined
by the network adminisgroup of IP addresses that is
trator. Of course, purely
established dynamically.
manual configuration
Each workstation is given
The dynamic
enables a high degree of
the opportunity to join a
nature of VLANs
control. However, in
particular IP multicast group
defined by IP
larger enterprise
by responding affirmatively
multicast groups
networks, manual configto a broadcast notification,
enables a very
high degree of
uration is often not
which signals that group’s
flexibility
and
practical. Furthermore, it
existence. All workstations
application
defeats one of the
that join an IP multicast
sensitivity.
primary benefits of
group can be seen as
VLANs: elimination of
members of the same virtual
the time it takes to
LAN. However, they are
administer moves and changes—although
only members of a particular multicast group
moving users manually with VLANs may
for a certain period of time. Therefore, the
actually be easier than moving users across
dynamic nature of VLANs defined by IP mulrouter subnets, depending on the specific
ticast groups enables a very high degree of
vendor’s VLAN management interface.
flexibility and application sensitivity. In
• Semiautomated. Semiautomated configuaddition, VLANs defined by IP multicast
ration refers to the option to automate either
groups would inherently be able to span
initial configuration, subsequent reconfigurouters and thus WAN connections.
rations (moves/changes), or both. Initial conCombination VLAN Definitions
figuration automation is normally accompDue to the trade-offs between various types of
lished with a set of tools that map VLANs to
VLANs, many vendors are planning to include
existing subnets or other criteria. Semimultiple methods of VLAN definition. Such a
automated configuration could also refer to
OSPF
Open Shortest Path First
PVC
Permanent virtual circuit
RIP
Routing Information Protocol
SVC
Switched virtual circuit
TCP/IP
Transmission Control
Protocol/Internet Protocol
TDM
Time-division multiplexing
5
situations where VLANs are initially configured manually, with all subsequent
moves being tracked automatically. Combining both initial and subsequent configuration automation would still imply semiautomated configuration, because the
network administrator always has the option
of manual configuration.
• Fully Automatic. A system that fully
automates VLAN configuration implies that
workstations automatically and dynamically
join VLANs depending on application, user
ID, or other criteria or policies that are preset
by the administrator. This type of VLAN
configuration is discussed in greater detail
toward the end of this paper.
Communicating VLAN Membership Information
Switches must have a way of understanding
VLAN membership (that is, which stations
belong to which VLAN) when network traffic
arrives from other switches; otherwise,
VLANs would be limited to a single switch. In
general, layer 2–based VLANs (defined by
port or MAC address) must communicate
VLAN membership explicitly, while VLAN
membership in IP-based VLANs is implicitly
communicated by the IP address. Depending
on the particular vendor’s solution, communication of VLAN membership may also be
implicit in the case of layer 3–based VLANs in
a multiprotocol environment.
To date, outside of implementing an ATM
backbone, three methods have been implemented for interswitch communication of
VLAN information across a backbone: table
maintenance via signaling, frame tagging, and
time-division multiplexing (TDM).
• Table Maintenance via Signaling. This
method operates as follows: When an endstation broadcasts its first frame, the switch
resolves the end-station’s MAC address or
attached port with its VLAN membership in
cached address tables. This information is
then broadcast continuously to all other
switches. As VLAN membership changes,
these address tables are manually updated by
a system administrator at a management
console. As the network expands and
switches are added, the constant signaling
6
necessary to update the cached address
tables of each switch can cause substantial
congestion of the backbone. For this reason,
this method does not scale particularly well.
• Frame Tagging. In the frame-tagging
approach, a header is typically inserted into
each frame on interswitch trunks to
uniquely identify which VLAN a particular
MAC-layer frame belongs to. Vendors
differ in the way they solve the problem of
occasionally exceeding the maximum
length of MAC-layer frames as these
headers are inserted. These headers also add
overhead to network traffic.
• TDM. The third, and least utilized method, is
time-division multiplexing. TDM works the
same way on the interswitch backbone to
support VLANs as it does in the WAN environment to support multiple traffic types—
here, channels are reserved for each VLAN.
This approach cuts out some of the overhead
problems inherent in signaling and frame
tagging, but it also wastes bandwidth,
because a time slot dedicated to one VLAN
cannot be used by another VLAN, even if
that channel is not carrying traffic.
Deploying an ATM backbone also enables
the communication of VLAN information
between switches, but it introduces a new set
of issues with regard to LAN Emulation
(LANE). ATM is discussed in detail in a
separate section of this paper. However, for the
time being, it should be remembered that with
port group–defined VLANs, the LANE
standard provides for a nonproprietary method
of communicating VLAN membership across
a backbone.
Standards and the Proprietary Nature of VLANs
Given the variety of types of VLAN definitions and the variety of ways that switches can
communicate VLAN information, it should not
be surprising that each vendor has developed
its own unique and proprietary VLAN
solutions and products. The fact that switches
from one vendor will not interoperate entirely
with VLANs from other vendors may force
customers to buy from a single vendor for
VLAN deployment across the enterprise. An
exception to this rule arises when VLANs are
format for frame tagging, in particular,
implemented in conjunction with an ATM
known as 802.1Q, represents a major
backbone and LANE. This is discussed further
milestone in enabling VLANs to be implein “VLANs and ATM,” later in this paper.
mented using equipment from several
The fact that single-vendor VLAN
vendors, and will be key in encouraging
solutions in the LAN backbone will be the rule
more rapid deployment of VLANs.
for the foreseeable future contributes to the
Furthermore, establishment of a frame
recommendation that VLANs should not be
format specification will allow vendors to
deployed indiscriminately throughout the
immediately begin incorporating this
enterprise. It also implies that purchase
standard into their switches. All major
decisions should be more highly centralized or
switch vendors, including 3Com, Alantec/
coordinated than they may traditionally have
FORE, Bay Networks, Cisco, and IBM
been. Thus, from both a procurement and a
voted in favor of this proposal.
technological perspective, VLANs should be
However, due to the lag time necessary for
considered as elements of a strategic approach.
some vendors to incorporate
The following two
the frame format specifiVLAN standards have been
cation and the desire on the
proposed:
The standardized
format
for
part of most organizations to
• 802.10 “VLAN
frame
tagging,
have a unified VLAN manStandard.” In 1995,
known as
agement platform, VLANs
Cisco Systems proposed
802.1Q, repwill, in practice, continue to
the use of IEEE 802.10,
resents a major
retain characteristics of a
which was originally
milestone in
enabling VLANs
single-vendor solution for
established to address
to be implesome time. This has sigLAN security for
mented using
nificant ramifications for
VLANs. Cisco attempted
equipment
deployment and proto take the optional
from several
curement of VLANs.
802.10 frame header
vendors.
Department-level proformat and “reuse” it to
curement for LAN
convey VLAN frame
equipment, particularly in
tagging instead of
the backbone, is not practical for organizations
security information. Although this can be
deploying VLANs. Purchasing decisions and
made to work technically, most members of
standardization on a particular vendor’s
the 802 committee have been strongly
solution throughout the enterprise will become
opposed to using one standard for two
the norm, and price-based product competition
discrete purposes. In addition, this solution
will decrease. The structure of the industry
would be based on variable-length fields,
itself may also shift in favor of the larger netwhich make implementation of ASIC-based
working vendors that can furnish a complete
frame processing more difficult and thus
solution across a wide range of components.
slower and/or more expensive.
• 802.1 Internetworking Subcommittee. In
VLAN Implementation Benefits
March, 1996, the IEEE 802.1 Internetworking Subcommittee completed the initial Why are vendors paying so much attention to
VLAN implementation? Will VLANs solve
phase of investigation for developing a
VLAN standard, and passed resolutions con- all of the network manager’s problems with
respect to moves, changes, broadcasts, and
cerning three issues: the architectural
performance?
approach to VLANs; a standardized format
for frame tagging to communicate VLAN
Reducing the Cost of Moves and Changes
membership information across multiple,
multivendor devices; and the future direction The reason most often given for VLAN implementation is a reduction in the cost of handling
of VLAN standardization. The standardized
7
user moves and changes. Since these costs are
quite substantial, this argument for VLAN
implementation can be compelling.
Many venders are promising that VLAN
implementation will result in a vastly increased
ability to manage dynamic networks and
realize substantial cost savings. This value
proposition is most valid for IP networks.
Normally, when a user moves to a different
subnet, IP addresses must be manually updated
in the workstation. This updating process can
consume a substantial amount of time that
could be used for more productive endeavors
such as developing new network services.
VLANs eliminate that hassle, because VLAN
membership is not tied to a workstation’s
location in the network, allowing moved workstations to retain their original IP addresses and
subnet membership.
It is certainly true that the phenomenon of
increasingly dynamic networks absorbs a substantial portion of the budgets of most IS
departments. However, not just any VLAN
implementation will reduce these costs.
VLANs themselves add another layer of
virtual connectivity that must be managed in
conjunction with physical connectivity. This is
not to say that VLANs cannot reduce the costs
of moves, and changes—if properly implemented, they will. However, organizations
must be careful not to simply throw VLANs at
the network, and they must make sure that the
solution does not generate more network
administration than it saves.
Virtual Workgroups
One of the more ambitious VLAN objectives
is the establishment of the virtual workgroup
model. The concept is that, with full VLAN
implementation across the campus network
environment, members of the same department
or section can all appear to share the same
“LAN,” with most of the network traffic
staying within the same VLAN broadcast
domain. Someone moving to a new physical
location but remaining in the same department
could move without having workstations
reconfigured. Conversely, a user would not
have to change his or her physical location
when changing departments—the network
8
manager would simply change the user’s
VLAN membership.
This functionality promises to enable a
more dynamic organizational environment,
enhancing the recent trend toward cross-functional teams. The logic of the virtual workgroup model goes like this: teams formed on a
temporary, project basis could be virtually connected to the same LAN without requiring
people to physically move in order to minimize
traffic across a collapsed backbone. Additionally, these workgroups would be dynamic:
VLANs corresponding to these cross-functional project teams could be set up for the
duration of the project and torn down when the
project was completed, all the while allowing
users to remain in the same physical locations.
Although this scenario seems attractive,
the reality is that VLANs alone cannot pave
the way for full utilization of the virtual
workgroup model. There are several managerial and architectural issues that, at this
point, pose problems for the virtual
workgroup model:
• Managing Virtual Workgroups. From a
network management perspective, the transitory nature of these virtual workgroups
may grow to the point where updating
VLAN membership becomes as onerous as
updating routing tables to keep up with adds,
moves, and changes today (although it may
save on the time and effort involved in physically moving the user’s workstation).
Moreover, there are still cultural hurdles to
overcome in the virtual workgroup model:
people usually move to be physically close
to those with whom they work, rather than to
reduce traffic across a collapsed backbone.
• Maintaining the 80/20 Rule. Virtual LAN
support for virtual workgroups is often tied
to support of the “80/20 rule,” that is, 80
percent of the traffic is “local” to the
workgroup while 20 percent is remote or
outside of the workgroup. In theory, by
properly configuring VLANs to match
workgroups, only the 20 percent of the
traffic that is nonlocal will need to pass
through a router and out of the workgroup,
improving performance for the 80 percent of
the traffic that is within the workgroup.
is able to route inter-VLAN packets at wire
However, many believe that the applicability
speed, there is no performance advantage for
of the 80/20 rule is waning due to the
overlapping VLANs over routing between
deployment of servers and/or network appliVLANs to allow universal access to a cencations such as e-mail and Lotus Notes® that
tralized server. Remember, only interusers throughout the enterprise access on an
VLAN packets would need to be routed—
equal basis.
not all packets. Several vendors support
• Access to Local Network Resources. The
integrated routing as an alternative to overvirtual workgroup concept may run into the
lapping VLANs.
simple problem that users must sometimes be
While workgroup VLANs may be
physically close to certain resources such as
extended to centralized server farms (for
printers. For example, a user is in the
example, including a particular file server in
Accounting VLAN, but is physically located
a particular workgroup’s VLAN), this is not
in an area populated by members of the Sales
always possible. In some networks, the MIS
VLAN. The local network printer is also in
people who control the servers may want to
the Sales VLAN. Every time this Accounting
place routers between the server farms and
VLAN member prints to the local printer, his
the rest of the network in order to create a
print file must traverse a router connecting
separate administrative domain or to
the two VLANs. This problem can be
enhance network security via router access
avoided by making that printer a member of
control lists. Depending on the vendor
both VLANs. This clearly favors VLAN
implementation, most switching products
solutions that enable overlapping VLANs,
will not support VLANs that extend across
discussed later. If overlapping VLANs are
routers (the exception to this would be
not possible, this scenario would require that
“VLANs” that equate to IP multicast
routing functionality be built into the
groups). It should be kept in mind that corbackbone switch. Then, the example print
doning off servers with external routers confile would be routed by the switch rather than
flicts with one of the reasons for utilizing
having to go through an external router.
switches and VLANs in the first place—to
• Centralized Server Farms. Server farms
avoid the delay introduced by routers.
refer to the placement of departmental
servers in a data center, where they can be
provided with consolidated backup, uninter- Reduction of Routing for Broadcast Containment
rupted power supply, and a proper operating Even the most router-centric networking
vendors have come to embrace the philosophy
environment. The trend toward server farm
of “switch when you can,
architecture has accelroute when you must.”
erated recently and is
LAN switches
Although switches certainly
expected to continue in
supporting
provide substantial perorder to ease adminisVLANs can be
formance enhancements
trative costs.
used to effecover layer 3 packet forCentralized server
tively control
warding (routing), as users
broadcast
farms raise problems for
traffic,
reducing
learned years ago with
the virtual workgroup
the need for
bridges, switches normally
model when vendor
routing.
do not filter LAN broadcast
solutions do not provide
traffic; in general, they
the ability for a server to
replicate it on all ports. This
belong to more than one VLAN simultanot only can cause large switched LAN envineously. If overlapping VLANs are not
possible, traffic between a centralized server ronments to become flooded with broadcasts, it
is also wasteful of precious wide area network
and clients not belonging to that server’s
bandwidth. As a result, users have traditionally
VLAN must traverse a router. However, if
been forced to partition their networks with
the switch incorporates built-in routing and
9
routers that act as broadcast “firewalls.”
Hence, simple switches alone do not allow
users to phase out routers completely.
One of the primary benefits of VLANs is
that LAN switches supporting VLANs can be
used to effectively control broadcast traffic,
reducing the need for routing. Broadcast traffic
from servers and end-stations in a particular
VLAN is replicated only on those switch ports
connected to end-stations belonging to that
VLAN. Broadcast traffic is blocked from ports
with no end-stations belonging to that VLAN, in
effect creating the same type of broadcast
firewall that a router provides. Only packets that
are destined for addresses outside the VLAN
need to proceed to a router for forwarding.
There are multiple reasons for utilizing
VLANs to reduce the need for routing in the
network:
• Higher Performance and Reduced Latency.
As the network expands, more and more
routers are required to divide the network
into broadcast domains. As the number of
routers increase, latency begins to degrade
network performance. A high degree of
latency in the network is a problem now for
many legacy applications, but it is particularly troublesome for newer applications
that feature delay-sensitive multimedia and
interactivity. Switches that employ VLANs
can accomplish the same division of the
network into broadcast domains, but can do
so at latencies much lower than those of
routers. In addition, performance, measured
in packets per second, is usually much higher
for switches than for traditional routers.
However, it should be noted that there are
some switches supporting network
layer–defined VLANs that may not perform
substantially faster than routers. Additionally,
latency is also highly correlated to the
number of hops a packet must traverse, no
matter what internetworking device (switch
or router) is located at each hop.
• Ease of Administration. Routers require
much more complex configuration than
switches; they are “administratively rich.”
Reducing the number of routers in the
network saves time spent on network management.
10
• Cost. Router ports are more expensive than
switch ports. Also, by utilizing cheaper
switch ports, switching and VLANs allow
networks to be segmented at a lower cost
than would be the case if routers alone were
used for segmentation.
In comparing VLANs with routing,
VLANs have their disadvantages as well. The
most significant weakness is that VLANs have
been, to date, single-vendor solutions and
therefore may lead to switch vendor lock-in.
The primary benefits of VLANs over routing
are the creation of broadcast domains without
the disadvantages of routing and a reduction in
the cost of moves and changes in the network.
Therefore, if neither of these is a problem,
then the user organization may want to forgo
VLANs and continue deploying a multivendor
network backbone, segmented by a mix of a
few routers and a relatively large number of
simple switches.
Assuming a major implementation of
VLANs, what is the role of routers in a
network? Routers have two remaining responsibilities: to provide connectivity between
VLANs, and to provide broadcast filtering
capabilities for WAN links, where VLANs are
generally not appropriate.
Routing Between VLANs. VLANs can be
used to establish broadcast domains within the
network as routers do, but they cannot forward
traffic from one VLAN to another. Routing is
still required for inter-VLAN traffic. Optimal
VLAN deployment is predicated on keeping as
much traffic from traversing the router as
possible. Minimizing this traffic reduces the
chance of the router developing into a bottleneck. As a result, the corollary to “switch
when you can, route when you must” in a
VLAN environment becomes “routing is used
only to connect VLANs.”
Having said this, however, keep in mind
that in some cases routing may not prove to be
much of a bottleneck. As mentioned earlier,
integrating routing functionality into the
backbone switch eliminates this bottleneck if
this routing is accomplished at high speed for
inter-VLAN packets.
VLANs Over the WAN. Theoretically, VLANs
can be extended across the WAN. However,
this is generally not advised, since VLANs
defined over the WAN will permit LAN
broadcast traffic to consume expensive WAN
bandwidth. Because routers filter broadcast
traffic, they neatly solve this problem.
However, if WAN bandwidth is free for a particular organization (for example, an electric
utility with dark fiber installed in its right of
way), then extending VLANs over a WAN can
be considered. Finally, depending on how the
they are constructed, IP multicast groups
(functioning as “VLANs”) can be effectively
extended across the WAN, as well as the
routers providing the WAN connections,
without wasting WAN bandwidth.
Security
The ability of VLANs to create firewalls can
also satisfy more stringent security requirements and thus replace much of the functionality of routers in this area. This is primarily true when VLANs are implemented in
conjunction with private port switching. The
only broadcast traffic on a single-user segment
would be from that user’s VLAN (that is,
traffic intended for that user). Conversely, it
would be impossible to “listen” to broadcast or
unicast traffic not intended for that user (even
by putting the workstation’s network adapter
in promiscuous mode), because such traffic
does not physically traverse that segment.
VLANs and ATM
While the concept of VLANs originated with
LAN switches, their use may need to be
extended to environments where ATM
networks and ATM-attached devices are also
present. Combining VLANs with ATM
networks creates a new set of issues for
network managers, such as relating VLANs to
ATM emulated LANs (ELANs), and determining where to place the routing function.
VLANs Transparent to ATM
In a LAN backbone with VLANs spanning
more than one LAN switch, switches
determine where frames have originated by the
techniques discussed earlier in “Communi-
cating VLAN Membership Information”
(VLAN tables, frame tagging, and TDM). In
an environment where ATM exists only in the
backbone (that is, there are no ATM-connected
end-stations), ATM permanent virtual circuits
(PVCs) may be set up in a logical mesh to
carry intra-VLAN traffic between these
multiple LAN switches.
In this environment, any proprietary
technique the vendor has employed is transparent to the ATM backbone. ATM switches
do not have to be VLAN “aware.” This means
that ATM backbone switches could be from a
different vendor than the LAN switches; ATM
backbone switches could be selected without
regard for VLAN functionality, allowing
network managers to focus more on performance-related issues. As convenient as this
situation sounds, it does not reflect reality for
many network environments.
Complexity Arising with ATM-Attached Devices
Usually, organizations that implement ATM
backbones would also like to connect workstations or, more likely, servers directly to
those backbones. As soon as any logical endstation is connected via ATM, a new level of
complexity arises. LAN Emulation must be
introduced into the network to enable ATMconnected end-stations and non-ATM-connected end-stations to communicate.
LAN Emulation
With the introduction of ATM-connected
end-stations, the network becomes a truly
“mixed” environment, with two types of
networks operating under fundamentally different technologies: connectionless LANs
(Ethernet, Token Ring, FDDI, etc.) and connection-oriented ATM. This environment puts
the responsibility on the ATM side of the
network to “emulate” the characteristics of
broadcast LANs and provide MAC-to-ATM
address resolution.
The LAN Emulation (LANE) specification, standardized in 1995 by the ATM
Forum, specifies how this emulation is accomplished in a multivendor environment. LANE
specifies a LAN Emulation server (LES),
which can be incorporated into one or more
11
net
Ether
net
Ether
LAN
5
ATM k
or
netw
1
net
h
Ether
switc
LEC
SVC
LEC
SVC
ATM
h
switc
File
r
serve
net
Ether
4
LES 2
SVC
net
Ether
3
LEC
LAN
net
Ether
h
switc
Figure 2. LAN Emulation
switches or a separate workstation to provide
the MAC-to-ATM address resolution in conjunction with LAN Emulation clients (LECs),
which are incorporated into ATM edge
switches and ATM NICs.
Figure 2 briefly illustrates how LANE
operates:
1. The LAN switch receives a frame from an
Ethernet-connected end-station. This frame
is destined for another Ethernet end-station
across the ATM backbone. The LEC
(which in this situation resides in the LAN
switch) sends a MAC-to-ATM address resolution request to the LES (which in this
case resides in an ATM switch).
2. The LES sends a multicast to all other
LECs in the network.
3. Only the LEC that has the destination
(MAC) address in its tables responds to the
LES.
4. The LES then broadcasts this response to
all other LECs.
5. The original LEC recognizes this response,
learns the ATM address of the destination
switch, and sets up a switched virtual
circuit (SVC) to transport the frame via
ATM cells as per AAL5, which governs
segmentation and reassembly.
In looking at the path of traffic between
an Ethernet-attached client and an ATMattached server, the section that is governed by
LANE extends from the LEC in the ATM
interface of the LAN switch to the LEC
12
residing in the server’s ATM NIC. From the
standpoint of either MAC driver, frames pass
directly between them just as if they were connected by a non-ATM backbone, with each
LEC acting as a proxy MAC address. VLANs
defined by port group would treat the ATM
interface on the LAN switch as just another
Ethernet port, and all ATM-attached devices
would then be members of that VLAN. In this
way, VLANs could be deployed without
regard to whether the ATM switches in the
backbone are from the same vendor (so long
as they support LANE).
However, from an administrative point of
view, many organizations may not want to
employ separate management software for the
ATM backbone and may prefer to source both
edge devices (LAN switches) and backbone
devices (ATM switches) from the same
vendor.
LANE can also allow for multiple ELANs
by establishing more than one LEC in the
ATM interfaces of participating devices (as
well as a separate LES for each ELAN). Each
LEC in the ATM interface of the LAN switch
is treated as a separate logical Ethernet port,
and each LEC in a single ATM-attached
device is seen as a separate Ethernet-attached
end-station. Therefore, multiple LECs in a
single ATM-attached device can be members
of different VLANs, allowing these VLANs to
overlap at ATM-attached devices. Since
LANE supports only ATM-attached devices,
VLAN
VLAN
LAN switch with ATM interface
and VLANs defined by port group
#2
#1
MAC er
driv
MAC er
LEC2
driv
LEC1
ATM
1
#
ELAN
LEC1
Represents a single ATM interface
with two LECs, each emulating an
Ethernet port assigned to different VLANs
ELAN
#2
LEC2
MAC r
drive
MAC r
drive
n
catio
Appli
Represents a single ATM interface
with two LECs, each a member of
different ELANs and VLANs*
ATM-attached server running
applications accessible by both VLANS
* Note: Each LEC on a single ATM interface must be on separate ELANs. They are shown here on separate VLANs
only because their corresponding LECs on the ATM switch have been explicitly assigned to different VLANs.
Figure 3. VLANs as Supersets of ELANs
while VLANs are defined for both ATM and
non-ATM network devices, VLANs can be
seen as supersets of ELANs (Figure 3).
With this structure, an ATM backbone
can enable all end-stations from multiple
VLANs to access a centralized server or
servers without passing through a router by
establishing a separate ELAN for each VLAN.
Since most traffic in a network is between
client and server, establishing VLANs that
overlap at ATM-attached servers greatly
reduces the number of packets that must be
routed between VLANs. Of course, there is
still likely to be a small amount of inter-
VLAN traffic remaining. Therefore, a router is
still required for traffic to pass from one
VLAN to another (and, therefore, from one
ELAN to another). Figure 4 depicts this type
of structure.
Routing Between Emulated LANs and/or VLANs
Since routing remains necessary in any mixed
ATM/shared media environment to forward
inter-VLAN traffic, network designers are
faced with the question of where to locate the
router functionality. The following are four
architectural solutions to the problem of where
to locate the routing functionality: edge
net
Ether
VLAN #1
LAN
net
Ether
h
switc
net
Ether
This router connects both
VLANs and thus both ELANs
r
Serve
r
Route
ELAN #1
File
r
serve
ATM
network
net
Ether
net
Ether
r
net
Serve
itch
N sw
Ether
LA
ELAN #2
VLAN #2
Figure 4. Router Connecting Overlapping VLANs/ELANs
13
routing, the “one-armed” router, the route
server, and MPOA.
Edge Routing. Basically, edge routing dictates
that the routing function across the ATM
backbone be incorporated into each LAN
switch at the “edge” of the ATM backbone.
Traffic within VLANs can be switched across
the ATM backbone with minimal delay, while
inter-VLAN packets are processed by the
routing function built into the switch. In this
way, an inter-VLAN packet does not have to
make a special trip to an external router, eliminating a time-consuming extra hop.
There are three other major advantages to
this architecture. First, unlike solutions that
have centralized routing, there is no single
point of failure with edge routing architectures.
Second, several solutions featuring edge
routing are available today. Third, edge routing
will function in multivendor environments if
each vendor’s equipment supports LAN
Emulation.
The primary disadvantage of edge routing
is the difficulty of managing multiple physical
devices relative to having centralized management of a consolidated router/routing
function. Additionally, edge routing solutions
may be more expensive than centralized routing
solutions made up of a centralized router and
multiple, less-expensive edge switches.
The One-Armed Router. The concept of the
so-called “one-armed router” has become particularly attractive because it removes the more
processing-intensive, higher-latency routing
function from the primary data path. A one-
LAN
One-armed
router
armed router sits off the side of an ATM
backbone switch with a single ATM link,
allowing packets that do not need to traverse
the router to pass through the ATM backbone
unimpeded. Another advantage of the onearmed router is that, relative to other configurations, it is less complex to configure and
administer.
The key to the one-armed router
structure, shown in Figure 5, is to keep as
much traffic as possible out of the one-armed
router. By structuring VLANs to support the
80/20 rule (so that 80 percent of the traffic
remains within each VLAN), the router is not
required to handle most traffic. For this to
work well, optimal configuration of VLANs
to minimize inter-VLAN traffic (traffic
passing through the one-armed router) is
critical. There are several vendors presently
shipping one-armed router solutions.
One of the disadvantages of the onearmed router is that it represents a single
point of failure in the network. For this
reason, two or more redundant one-armed
routers are generally preferred. However,
perhaps the most significant drawback of the
one-armed router is that its one arm can
develop into a bottleneck if VLAN traffic
does not support the 80/20 rule. This can
occur particularly in networks with large
amounts of peer-to-peer traffic.
The Route Server. The route server model (see
Figure 6) is physically similar to the one-armed
router model, but logically very different in
that it breaks up the routing function into distributed parts. In a one-armed router configuration, a packet from VLAN A heading to
h
switc
h
switc
ATM
ATM network
Traffic within the same VLAN
Traffic between VLANs
ATMitch
sw
ATMitch
sw
LAN
Figure 5. One-Armed Router
14
h
switc
LAN
h
switc
Traffic within the same VLAN
Traffic between VLANs
Bidirectional signaling required
for address resolution
ATMitch
sw
Route
server
h
switc
ATM
ATM network
ATMitch
sw
LAN
h
switc
Figure 6. Route Server
VLAN B is sent to the one-armed router,
where it waits for address resolution, path calculation, establishment of a connection across
the ATM backbone, and, finally, transmission.
In a route server scheme, the same packet
waits in the cache of the LAN switch at the
edge of the ATM backbone before transmission. In this process, the packet itself never
traverses a router. The only traffic to and from
the route server is the signaling required to set
up a connection between LAN switches across
the ATM backbone. The advantage is that less
routed traffic must be diverted to the route
server, often reducing the number of hops
required through the backbone. Also, overall
traffic across the route server’s one arm is
reduced.
There are, of course, disadvantages to the
route server approach as well. First, initial
vendor implementations are strictly proprietary
and do not support standard routing protocols.
Secondly, at this point available route servers
only support IP. Of course, the route server
shares one of the one-armed router’s
drawbacks in that it can be a single point of
failure, but, as with the one-armed router, this
problem can be mitigated through redundancy.
Finally, because a route server architecture
requires LAN switches to have a certain level
of routing functionality, route server solutions
tend to be more expensive and more complex
to configure than the relatively simple LAN
switches deployed in the one-armed router
architecture.
MPOA. There is at least one development that
may eventually standardize the route server
approach. The Multiprotocol over ATM
(MPOA) standards working group of the
ATM Forum is currently working out the
details of an implementation model for MPOA
service. While a variety of models have been
proposed, MPOA is expected to provide direct
virtual circuit connectivity between ATMnetwork-attached devices that may belong to
different routing subnets. In other words,
MPOA can let logical end-stations that are
part of different ELANs communicate directly
across an ATM network without requiring an
intervening router.
Since ELANs are subsets of VLANs,
MPOA holds the promise of enabling an
ATM backbone to connect VLANs without
the need for an external router. MPOA can be
considered an enhancement beyond LANE
that integrates routing functionality into the
LAN-ATM edge switch. All inter-VLAN
traffic would be able to leverage this capability, and network latency would be reduced.
An MPOA standard is not expected to be
finalized until at least 1997, and the initial
implementation will most likely support only
TCP/IP. It should be noted that some of the
disadvantages of the route server approach,
such as cost and management complexity,
would remain in MPOA solutions.
VLANs and DHCP: Overlapping Solutions
With Microsoft’s recent introduction of the
Dynamic Host Configuration Protocol
(DHCP), users now have another alternative
for reducing the workload associated with
administration of workstation IP address.
Unfortunately, DHCP can actually conflict
15
with VLAN implementation, especially with
layer-3, IP-based VLANs.
DHCP Functionality
When considering the ability of VLANs to
deal with ever-changing networks, it should be
remembered that most of the difficulty in supporting adds, moves, and changes occurs in IP
networks. In order to deal with the problem of
reconfiguring IP addresses, Microsoft has
developed DHCP, a TCP/IP-based solution
incorporated into the Windows NT™ server
and most Windows® clients.
Rather than establishing location-independent broadcast domains as VLANs do,
DHCP dynamically allocates IP addresses to
logical end-stations for fixed periods of time.
When the DHCP server detects a workstation
whose physical location no longer corresponds to its allocated IP address, it simply
allocates that end-station a new address. By
doing so, DHCP enables workstations to be
moved from subnet to subnet without the
network administrator having to manually
configure the workstation’s IP address or
update host table information.
The element of DHCP that equates most
closely to VLAN functionality is the network
administrator’s ability to specify a range of IP
addresses available for a particular logical
workgroup. These logical groups are termed
“scopes” in the Microsoft lexicon. However,
scopes should not be equated with VLANs,
because members of a single scope are still
bound by their physical subnet, although there
can be multiple scopes residing in each subnet.
Consequently, DHCP implementation may
reduce the labor-intensive administration of
TCP/IP networks, but DHCP alone does not
control network broadcasts in the same way
that VLANs do.
Best Use for Each
In what types of network environments should
VLANs be implemented, and in what types of
network environments does DHCP make the
most sense? Since DHCP is solely an IP-based
solution, it has little appeal in environments
where IP users are a minority, since all nonTCP/IP clients would be excluded from scope
membership. In particular, network envi-
16
ronments where non-TCP/IP protocols are
required for mission-critical applications may
benefit more from VLAN implementation,
since VLANs can be used to contain multiprotocol broadcast traffic.
However, for smaller, purely TCP/IP
network environments (under 500 nodes),
DHCP alone may suffice. By simply having
fewer total network nodes and fewer physical
subnets, the need to establish fully locationindependent logical groups is greatly
reduced. Additionally, for medium-sized
organizations that, for whatever reason, do
not support location-independent workgroups, VLANs lose much of their appeal
when compared to DHCP.
There is one area in which VLANs and
DHCP do not compete: reducing the necessity
for routing in the network. Although DHCP
servers dynamically maintain address tables,
they lack routing functionality and cannot
create broadcast domains. Therefore, DHCP
has no impact on an organization’s need for
routing in the network. In environments where
the containment of broadcast traffic without
having to resort to routers is a major
requirement, VLANs are a better solution.
Overlap Between DHCP and VLANs
It what ways can DHCP and VLANs work
together, and in what situations do they represent competitive solutions?
DHCP and layer-3, IP-based VLANs
clearly represent competitive solutions
because of addressing problems that stem
from implementing layer 3–based VLANs in
conjunction with DHCP. If a client workstation physically moves to a new subnet, the
DHCP server will allocate a new IP address
for that workstation. Yet, this workstation’s
VLAN membership is based on the old IP
address. Therefore, the network administrator
would have to manually update the client’s IP
address in the switch’s VLAN tables. This
would eliminate the primary benefit of DHCP
and one of the primary benefits of IP-based
VLANs. In summary, these two solutions represent an either/or proposition for most
network environments.
Implementing VLANs defined by MAClayer address in conjunction with DHCP is a
need to be made available to users regardless
somewhat more plausible solution. However,
of their VLAN membership. Ideally, this
DHCP together with MAC-based VLANs
would create a two-tiered, redundant matrix of access should be provided without most user
traffic having to traverse a router.
logical groups (MAC address–based VLANs
Organizations that implement VLANs recand DHCP scopes). Having two tiers of logical
ognize
the need for certain logical end-stations
groups would make otherwise easy-to-manage,
(for example, centralized servers) to commu“drag-and-drop” moves, adds, and changes
nicate with multiple VLANs on a regular basis,
unnecessarily difficult and might entail more
either through overlapping VLANs (in which
labor-intensive network administration than if
network-attached end-stations simultaneously
neither solution was implemented.
belong to more than one VLAN) or via intePort group–based VLANs and DHCP can
grated routing that can process inter-VLAN
coexist, and their joint implementation can
packets at wire speed. From a strategic
even be complementary. As stated earlier,
standpoint, these organizations have two ways
when users in VLANs based purely on port
to deploy VLANs: an “infrastructural” VLAN
groups move from one port group to another,
implementation or a “service-based” VLAN
their VLAN membership changes. In a nonimplementation. The choice of approach will
DHCP environment where IP subnets corhave a substantial impact on the overall
respond one-to-one with VLANs, users who
network architecture, and may even affect the
move from one port group to another would
management structure and business model of
still need to have their workstation reconthe organization.
figured to reflect their new IP subnet.
Implementing DHCP would make this reconInfrastructural VLANs
figuration automatic. The port group–based
An infrastructural approach to VLANs is based
VLANs, of course, provide the broadcast conon the functional groups (that is, the
tainment that DHCP implementation alone
departments, workgroups,
does not. In this way, DHCP
sections, etc.) that make up
and port-group-based
the organization. Each
VLANs can work together
The choice of
functional group, such as
approach will
to accomplish both
have a subaccounting, sales, and engibroadcast containment and
stantial impact
neering, is assigned to its
automation of moves and
on the overall
own uniquely defined
changes.
network archiVLAN. Based on the 80/20
Port group–based
tecture, and
rule, the majority of
may even affect
VLANs and DHCP, in connetwork traffic is assumed
the
management
junction with deployment of
structure and
to be within these funcarchitectures that reduce the
business model
tional groups, and thus
need for external routing of
of the organiwithin each VLAN. In this
inter-VLAN traffic (such as
zation.
model, VLAN overlap
multiple VLAN memberhip
occurs at network resources
or integrating routing into
that must be shared by
the switch), represent a
multiple workgroups. These resources are
fairly complete short- to medium-term
solution, which will alleviate the most pressing normally servers, but could also include
printers, routers providing WAN access, workproblems faced in many network envistations functioning as gateways, and so forth.
ronments.
The amount of VLAN overlap in the
infrastructural
model is minimal, involving
VLAN Architectures Going Forward
only
servers
rather
than user workstations—
Due to the trends toward server centralization,
making
VLAN
administration
relatively
enterprise-wide e-mail, and collaborative
straightforward. In general, this approach fits
applications, various network resources will
17
®
®
il
E-ma r
serve
g
untin
Acco base
data ver
ser
Sales se
a
databver
ser
are
NetW rver
file se
UNIX er
rv
file se
g
eerin
EnginLAN
V
Sales
VLAN
g
untin
Acco AN
VL
Figure 7. Infrastructural VLANs
well in those organizations that maintain clean,
discrete organizational boundaries. The infrastructural model is also the approach most
easily enabled by presently available solutions
and fits more easily with networks deployed
today. Moreover, this approach does not
require network administrators to alter how
they view the network, and entails a lower cost
of deployment. For these reasons, most organizations should begin with an infrastructural
approach to VLAN implementation.
As can be seen in the example in Figure 7,
the e-mail server is a member of all of the
departments’ VLANs, while the accounting
database server is only a member of the
accounting VLAN.
Service-Based VLANs
A service-based approach to VLAN implementation looks, not at organizational or functional groups, but at individual user access to
servers and applications—that is, network
resources. In this model, each VLAN corresponds to a server or service on the network.
Servers do not belong to multiple VLANs—
groups of users do. In a typical organization,
all users would belong to the e-mail server’s
VLAN, while only a specified group such as
the accounting department plus top-level executives would be members of the accounting
database server’s VLAN.
By its nature, the service-based approach
creates a much more complex set of VLAN
membership relationships to be managed.
Given the level of most VLAN visualization
18
tools presently available, a large number of
overlapping VLANs using the service-based
approach could generate incomprehensible
multilevel network diagrams at a management
console. Therefore, to be practical, servicebased VLAN solutions must include a high
level of automatic configuration features.
However, in response to the types of applications organizations want to deploy in the
future, as well as the shift away from traditional, more rigid organizational structures, the
trend in VLAN implementation will be toward
the service-based approach. Figure 8 depicts
the service-based VLAN model.
As bandwidth to the desktop increases and
as vendor solutions become available to better
manage greater VLAN overlap, the size of the
groups that belong to a particular set of
VLANs may become smaller and smaller. At
the same time, the number of these groups
becomes larger and larger, to the point where
each individual could have a customized mix
of services delivered to his or her workstation.
Taking that concept a step further, control over
what services are delivered at a given time
could be left up to each individual user. At that
point, the network structure begins to take on
the multiple-channel characteristics of a cable
TV (CATV) network. In fact, at this stage, this
model finds the greatest degree of similarity in
VLANs defined by IP multicast group—each
workstation has the choice of which IP multicast or “channel” it wants to belong to.
In such a future environment, VLANs lose
the characteristics of static or semistatic
broadcast domains defined by the network
manager, and become channels to which users
subscribe. Users simply sign up for the applications they need delivered to them at a particular time. Application use could be
accounted for, enabling precise and automated
chargeback for network services. Network
managers could also retain control in order to
block access to specific channels by certain
users for security purposes.
VLAN Migration Strategies
As this paper has demonstrated, there are
many factors to be considered in VLAN
implementation: technological, architectural,
and organizational. Given the effects of
VLANs on network architecture, organizational structure, and even the business model
of some organizations, it is difficult to deploy
VLAN technology solely as a tactical solution,
only where and when it is needed. However,
this does not imply an all-or-nothing strategy
in which the network architecture is transformed overnight from one based on physical
subnets and router-based segmentation to one
of service-based VLANs.
What steps are necessary before applying
VLANs to an enterprise network? Initially,
VLANs should be seen as a solution to at least
one of two problems:
• Containment of broadcast traffic to
minimize dependence on routers
• Reduction in the cost of network moves and
changes
il
E-ma r
serve
g
untin
Acco base
a
t
a
d
r
serve
LAN
LAN
rver V
ile se e server V er VLAN
f
X
I
VLAN
il
v
f
UN
e
e
s
r
a
e r
erver
NetW s databas atabase s
gd
Sale
untin ver VLAN
Acco
il ser
E-ma
An organization where broadcast traffic is
not yet a problem, or where the cost of
network moves and changes is tolerable, may
want to forgo implementing VLANs for the
time being. However, the majority of large
enterprise networks are now experiencing one
or both of these problems.
In organizations that are rapidly replacing
routers with switches and may soon face
broadcast traffic containment issues, another
element of the network architecture should be
considered: the degree to which the network
has evolved toward a single user/port switched
LAN architecture. If the majority of users are
still on shared LAN segments, the ability of
VLANs to contain broadcasts is greatly
reduced. If multiple users belonged to different
VLANs on the same shared LAN segment,
that segment would receive broadcasts from
each VLAN—defeating the goal of broadcast
containment.
Having determined that VLANs need to
be a part of network planning in the immediate
future, server access, server location, and
application utilization must all be thoroughly
analyzed to determine the nature of traffic flow
in the network. This analysis should answer
the remaining questions about where VLAN
broadcast domains should be deployed, what
role ATM needs to play, and where the routing
function should to be placed.
Because of the limitations of present
VLAN technology, initial VLANs are likely
to employ an infrastructural approach.
Sales se
a
databver
ser
are
NetW rver
file se
UNIX er
rv
file se
g
eerin
Engin
Sales
g
untin
Acco
Figure 8. Service-Based VLANs
19
However, as vendor solutions develop, many
organizations will want to consider migration
toward a more service-based model, which
will more easily let users subscribe to various
network services.
This concept of user-controlled subscribership, as opposed to administrator-controlled membership, is augmented by NICs
with built-in VLAN functionality operating in
environments with a single user per switch
port. In this scheme, the NIC driver dynamically tells the switch which multicast groups
or VLANs it wants to belong to. Certainly, this
type of distributed VLAN control leverages the
increasing processing power of the desktop
and enables a higher degree of other, related
functionality such as automatic VLAN configuration and traffic monitoring. In addition,
agents residing in each NIC will enable the
workstation to collect and report information
on specific application usage (rather than just
simple layer 2 traffic statistics in the case of
RMON1). This capability facilitates the
automated chargeback for network services
described earlier for service-based VLANs.
If individual users control VLAN membership, what about security? Clearly, users
cannot be allowed to simply subscribe to any
network service they wish. The network
administrator must be able to establish policies
that define which users have access to what
resources and what class of service each user is
entitled to. One solution to the security
problem may come in the form of an authentication server. These servers may well develop
into the primary method by which the VLANs
of the future are defined. Authentication
servers define VLAN membership by user ID
(password or other authentication device)
rather than by MAC address or IP address.
Defining VLANs in this way greatly increases
flexibility and also implies a certain level of
integration of VLANs with the network
operating system, which typically asks the user
for a password anyway to allow or deny access
to network resources. One of the primary
advantages of authentication servers is that
20
they allow the user to take his or her VLAN
anywhere, without regard to which workstation
or protocol is being used.
The analysis of network traffic, applications usage, server access, and so on that is
necessary in the VLAN migration process, and
which will be greatly furthered by the implementation of RMON2, may simply produce
VLANs that correspond to functional teams or
departments. On the other hand, if migration is
undertaken with a holistic view of the capabilities of VLAN technology, and the network
designers ask the question, “Who should talk
to whom?” rather than “Who is talking to
whom?,” it may become apparent that fundamental process and organizational changes are
needed. Many organizations are making such
changes: trends such as flatter hierarchies,
revamped workflows, and innovative business
models are helping to fully leverage the possibilities of emerging applications.
Conclusion
The concept of service-based VLAN technology holds the potential for harmonizing
many of today’s organizational and managerial
changes with the structural and technological
developments in the network. Despite the
promise of this vision, VLAN implementation
must solve real-world problems in order to be
financially justified. Organizations that have
deployed or are planning to deploy large
numbers of switch ports, dividing the network
into smaller segments to increase bandwidth
per user, can make a very strong case for
VLAN implementation in order to contain
broadcasts. However, any organization that
expends substantial resources dealing with
moves and changes in the network may also be
able to justify VLAN implementation. This is
simply because VLANs, if implemented as
part of a strategic solution, may be able to substantially reduce the cost of dealing with
moves and changes. For these organizations,
the switching infrastructure upon which most
VLAN solutions are based can be seen as an
added, and quite valuable, benefit.
AppleTalk is a trademark of Apple Computer. DECnet is a trademark of Digital Equipment Corporation. Lotus Notes is a trademark of Lotus Development Corporation. Windows and Windows NT are trademarks of Microsoft.
IPX and NetWare are trademarks of Novell. UNIX is a trademark of UNIX Laboratories.
Printed in U.S.A.
200374-001 5/96
Download PDF