Burning Up the Wire
The affordable Cisco 3500 switch offers a GUI interface to IOS, Gigabit Ethernet support, and clustering capabilities. What more could a network engineer want?
What is it about the switch that makes it so much more desirable
than the hub? Dividing a network into multiple collision domains doesn’t hurt
and is the primary reason for purchasing switches. The fact that most switches
are manageable means that you can baseline and track the impact of changes.
Then there’s just something undeniably attractive about anything that has the
capability to move data at a speed rated in gigabits.
The Cisco Systems Catalyst 3500 series of switches builds
on the success of the older 3000 series of switches, but where the 3000’s biggest
feature was the stack functionality—where you could control many switches through
a single one—the 3500 adds another appealing feature, Gigabit Ethernet. Not
only does the 3500 support Gigabit, but it also supports Full Duplex Gigabit
Ethernet… over copper!
What Sets the 3500 Apart
What makes this switch different from any of Cisco’s other
smaller switches, the 1900s and 2900s? Port densities don’t change; you can
order up to 48 ports on an individual switch. Cisco has announced an 80-port
model of the 2900 series, and while they haven’t done the same for the 3500s,
I wouldn’t be surprised if they do by mid-summer.
If you’re familiar with Cisco’s other more economical switches,
then you know that from a physical standpoint, the 2900 looks like a 1900 with
some slots added to the top. The 3500 though is about as thick as a 1900 while
offering the same number of ports as the 2900. The 2900 has an advantage here
in that the A and B slots may be used for ATM and FDDI in addition to Gigabit
Ethernet, while the extra slots on the 3500 may only be used for Gigabit.
Another major difference is the backplane capacity of the
switches. At 3.2 gigabits per second, the 2900 series is almost as fast as a
Catalyst 5500; but the 3500 series, at 10 gigabits per second, is three times
faster. This higher capacity is very useful because it’s possible to configure
a Gigabit EtherChannel circuit of four gigabits per second between two devices.
You would do this by grouping both Gigabit ports together and running them to
the same device on the other end.
As a Cisco instructor, one of the biggest complaints I hear
from beginning students is that Cisco’s IOS (Internetwork Operating System)
is tough to learn; why doesn’t the company offer a GUI with their devices? Well,
your prayers have been heard. While it doesn’t have capabilities on par with
CiscoWorks 2000 (what do you expect for free?), the 3500 series switches do
include a Web-based management graphical interface that’s surprisingly robust.
Let’s face it: Cisco’s IOS is extremely cryptic for those
who aren’t used to it. Previously, the GUI Cisco had you use was a version of
CiscoWorks, which is still available. Cisco now provides a Web-based GUI on
the 3500 series switch that works with both Internet Explorer and Netscape Navigator.
This interface can make completing basic tasks much easier, but it doesn’t eliminate
the need to touch IOS. In order to be able to use the GUI, you need to configure
the switch with an IP address; knowledge of how to complete this task in IOS
is required, but you can learn it easily. Many of the more advanced tasks require
configuration from the command line, and there are a few instances where the
task may be supported in the GUI but only for certain versions of software.
A Look Inside
What is it about Gigabit Ethernet that has everyone wanting
it? Oh, yeah, speed. The 3512, 3524, and 3548 come with two built-in gigabit
slots in addition to their Fast Ethernet ports. The 3508 doesn’t have any Fast
Ethernet ports; instead, it has eight gigabit slots. It’s not uncommon to see
a closet of 3500s in a rack where you have seven 3524s running Fast Ethernet
to the desktop, and then they’re connected together with the 3508, which also
has Gigabit Ethernet to the data center or server room. You need to be aware
of how Gigabit Ethernet works on these guys though in order to get the best
First, you need to decide how you want to connect two devices
via Gigabit Ethernet. I’ll try to whittle your choice down by one. If the distance
is more than one meter, then you can’t use copper. Note that the IEEE standard
does provide for copper lengths of more than one meter, but Cisco requires proprietary
cables. Now, you only have to choose between single-mode fiber and multi-mode
fiber. The deciding factors here are how far you need to run cable and how much
you’re willing to pay. The copper connectors are much cheaper than the single-mode
fiber but fiber goes a lot farther than 1m. Make sure you purchase the correct
GBIC (Gigabit Interface Converter) for your needs. If you plan on using the
GBIC in a 2900, make sure you purchase a Gigabit Ethernet module as well, since
the 2900XL doesn’t come with one.
You can connect two cables to each GBIC within the restriction
for looping that the installation guide lists. This means that a 3512 may be
connected via Gigabit Ethernet to four other switches, but when you do this,
you only get half duplex connectivity. If you want full duplex, connect a single
cable to the GBIC. The upshot here is that you have two gigabits of bandwidth
to play with per Gigabit Ethernet port. You can split it between two switches
or send it all to one, your choice. You may also channel the two ports by creating
an EtherChannel port group. This would allow for four gigabits of bandwidth
between two devices. I just want to know where the NIC is that will handle this!
The statistics on each interface will show if one or both sides of the GBIC
are in use. (See Figure 1.)
|Figure 1. Cisco's 3500 switch gigabit interface. (Click
on image to see larger version.)
One of the tasks supported from the GUI is troubleshooting.
Assuming you can get to the switch via the GUI, then you have the option of
selecting troubleshooting tools. “Show tech-support” is a command you’ll often
use when talking to a TAC engineer (Figure 2). You also have the option to see
interface information and log contents in addition to accessing the Visual Switch
|Figure 2. The GUI and the "show tech-support" link.
(Click on image to see larger version.)
Once in the Visual Switch Manager, the first thing you’ll
notice is how clean everything is. The next thing is that you have no clue where
Click on the labels at the top of the page (port, system,
security, and the like) to get a pull-down menu. (See Figure 3.)
|Figure 3. A pull-down menu in the Visual Switch Manager.
(Click on image to see larger version.)
From here, you can do many configuration tasks that would
have you fumbling around in IOS. For example, say you want to enable SNMP on
the switch. Click System | SNMP Configuration, and you can now configure SNMP.
Click the “Enable SNMP” check box and click Apply. You’re then placed in a screen
where you can configure SNMP Community Strings. (See Figure 4.)
|Figure 4. Setting SNMP. (Click on image to see larger
You can also see that the switch will automatically enable
certain defaults such as Public for a Read Community String. Since Community
Strings are pretty much the same as passwords when it comes to SNMP, this value
should be changed when using SNMP.
The 3500 is a direct descendent of the 3000 series switches
but has undergone some amazing transitions in the last few years. One of the
main features of the 3000 series switches was the ability to stack them into
a single manageable device. The 3500s have retained this ability, and stacking,
now called clustering, has also migrated to the 2900s and 1900s. This means
you can have multiple devices in a rack and manage them from a single connection.
In order to activate the clustering, you have to designate
a command switch. This is the device that will be in charge of the stack and
that will replicate many global properties, like IP addressing, down to other
switches in the cluster. Turning on clustering is a command line task; while
you can configure the switch from either the console port or the browser window,
you have to type the commands. (See Figure 5.)
|Figure 5. Turning on a cluster from the IOS. (Click
on image to see larger version.)
Once you have clustering turned on, you can get into Cluster
Builder. Opening this window shows what devices are connected to the one you’re
connected to (see Figure 6) but also opens a window called the “Suggested Candidate
Window” (Figure 7). This is a list of devices that Cluster Builder has decided
may join the stack if you wish. This makes it easy to add a switch to a rack
then connect it to the cluster. Cisco even provides for a check box so you can
prevent the Suggested Candidate Window from opening every time.
|Figure 6. Cluster Builder. (Click on image to see larger
|Figure 7. The suggested candidate window. (Click on
image to see larger version.)
Once you’ve created a cluster and added a second device to
it, you can see that there’s only one IP address on the two devices (Figure
8), and that one is 10.0.0.3 on device 3524. Device 3512 doesn’t have its own
IP address anymore. Once the cluster is enabled, you may find it easier to manage
the cluster from the Visual Switch Manager. Setting individual ports up is as
easy as right-clicking on the port you wish to modify and then selecting the
settings you want. Note that there’s a separate section for Virtual LAN modification.
|Figure 8. A cluster of switches with a single IP address.
(Click on image to see larger version.)
You can put a total of 16 devices into a cluster. Since this
could be 15 80-port 2900 series switches and one 3548, you could have a theoretical
maximum of 1,248 ports, all being managed from a single IP address. I haven’t
found any information saying that there’s a maximum number of ports that a single
IP address may manage.
One question I’m positive will appear in my mailbox is: “If
you manage the switch cluster via one IP address and if I telnet to that IP
address, I end up at the prime switch for the cluster; how do I configure a
remote switch from the command line?” Worry not, young admin, Cisco has considered
your plight and provided a solution. If you telnet to the management switch
for a cluster and want to configure a port on another switch in that cluster,
you need to know the cluster number of the switch you wish to configure. You
can find this out by entering, “show cluster members”, on the main switch. Once
you’ve found the number for the switch you want, enter the command, “rcommand
#”, where # is the switch number.
Sweet and Sturdy
Overall, this is a sweet set of boxes. I liked the 2900 when
it came out. The 3500 built on the success of the 2900 while still retaining
the main capability that set the 3000s apart from the other switches, the clustering.
If you think about it, 1,248 ports is a ton of ports—more than any organization
I’ve ever seen needed in one closet. The 3500 has a robust backplane to provide
for a massive amount of throughput at an economical cost. Costs will vary by
vendor and purchase levels but a quick check on the Internet showed the 3512
available for $1,800. On the down side, this is still a complex device and will
take time and effort to learn how to configure best to fit your organization.
The graphical interface makes things a bit easier for the novice and a lot easier
for someone already familiar with the workings of the device.