setup-network.txt 7.43 KB
Newer Older
##### Configuration suggestions for Cisco switches
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83

The purpose of this document is to aid in designing and setting up the control
and experimental networks on other emulabs.

First a bit of background: On the the control network will be your control
nodes (boss, ops, any tipservers, etc.), control hardware (SNMP-controllable
devices such as power controllers and switch IP interfaces), your connection to
the outside world, and the control interfaces of your nodes. So far, we haven't
tried to distribute the control net across multiple switches, but this should
be theorically possible. You'll need to set up the VLANs, etc. on the control
net by hand.

The experimental network will consist of one or more (this, we have tested)
switches connected with trunk (802.1q, or proprietary, like Cisco ISL) lines.
Most configuration of these switches will be taken care of by our software.
(See setup-cisco.txt for some configruation options you many want to apply to
these switches if they are Ciscos.)

##### Splitting up the control net

We basically have 4 VLANs on the control network:
'external' contains our connection to the outside world
'private' contains the boss node, and all IP-controllable devices
        (namely, power controllers and switch IP interfaces)
'public' contains our ops node
'control' contains the control net interfaces of all experimental nodes

This is done for security - we route (using a module in our control-net switch)
between these VLANs, and do some firewalling between each of them. The main
goals are:
1) Protect both control and experimental nodes from the outside world (and
   vice-versa - we don't want people attacking the outside world from our nodes)
2) Protect the control nodes from the experimental nodes
3) Protect the control hardware (power controllers, etc.) from nodes
   and the outside world
4) Protect the boss node (which is _not_ publically accessible) from the ops
   node (which all experimenters have shells on.)

Now, it's entirely possible to combine these VLANs into one big one - this is
what we've done on our mini-testbed here. But, there are some serious security
implications with doing it this way - namely, that the nodes can theoretically
impersonate each other, power cycle each other, and all kinds of nasty things.

At the very least, you should have a firewall between your testbed and the
outside world, to satisfy #1.

It is also a good idea to separate the nodes' control net into a separate VLAN,
which satisfies #2 and #3. After all, you are giving people root access to the
experimental nodes. In situations where you are only giving acess to a small
number of trusted people, this is probably not too big a deal, but once access
gets outside the small circle of your friends, or if you are allowing students
access, then taking these precautions are a very good idea.

If possible, putting boss and ops on separate networks is also a good idea
(#4), but is probably the least important part to worry about.

Basically, it's up to you to decide how much security you want to worry about.
I'd recommend at least going after #1, #2, and #3. We've put a lot of thought
into this, so if you're wondering how your choice here will affect other
aspects of security, ask Utah, and we can probably help you.

A (somewhat outdated) overview of the firewall rules we have in place can be
found in vlans.txt.

##### Connecting the contol net to the experimental net

In order to be able to control the experimental switches (ie. create new VLANs,
etc.) you need a way to talk to them. So, you'll need one line that goes from
your control net to the experimental net. However, you will need to be _very_
careful that no experimental traffic can leak across this line, and that it
cannot be used as a 'back door' into the private net by nodes.

On the control switch, it should be in the same VLAN as your boss node.

On the experimental switch, it should be in the same VLAN as the switch's IP
interface. The simplest way to accomplish this is just to leave it in VLAN 1,
the default one. However, you want to make sure that no ports on the nodes can
talk to it, so you'll want to disable them all. (Our software will re-enable
them, and disable them as needed.)                           

84 85 86 87 88
You will also need to give the experimental switch(es) IP addresses. With
CatOS, you do it like this:
  set interface sc0 <IP> <netmask>
  set interface sc0 up

89 90 91 92 93 94 95 96 97 98 99 100 101 102
##### DHCP through the router

If your boss node is on a seperate VLAN from the node control net, you'll need
to make sure that DHCP traffic can get from the control net to your boss node,
since normally, DHCP is not forwarded through routers. On Cisco routers, this
is done with the 'ip helper-address'. For example, here, the name of the
router's interface in the node control net is 'Vlan3'.  So, I'd log  into the
router, and run the following:                                 
  configure terminal                                                      
  interface Vlan3                                                         
  ip helper-address                                   

Of course, replace 'Vlan3' with the name of your router's node
control-net interface, and replace the IP address with that of your boss node.                                                          
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
##### IGMP snooping on the control net

In order for multicast to work correctly, you need to make sure that IGMP
snooping is enabled on the control switch. This is needed for frisbee, our
disk-loading system. It's up to you wheter you want to enable this on the
experimental switches. On CatOS, the command is:
  set igmp enable

#### VTP domains

If you have only one Cisco experimental-net switch, you don't need to have it
using VTP, which is a mechanism for keeping VLANs in sync across multiple
switches. You do this (in CatOS) with:
  set vtp mode transparent

If you have multiple experimental switches connected by trunk lines, you should
use FTP. Pick a domain name (we call ours simply 'Testbed',) and run the
following on all of your switches:
  set vtp domain <domainname>

Pick one switch to be the master - it doesn't really matter which one. (See the
switch setup insturctions in setup-db.txt, and make sure you name the stack
after the master switch.) On this one, run
  set vtp mode server
On all the others, run
  set vtp mode client

130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
##### Some variations on the above configuration

There are some variations on the above configuration that we haven't tried, but
may work for you.

First, you may try putting all IP-controllable control hardware on a VLAN (or
another switch/hub) hanging directly off another interface on the boss node. In
the example above, the control hardware is given 'real' IP addresses, but
prevented from talking to the outside worly by firewall rules. An alternative
to this would be to give them 'private' (ie. 10.0.0.X) addresses, and put
another interface in the boss node that talks only to these devices.

As an added measure of security on the experimental network, you could move
your switch IP interfaces (and the wire that runs from them to the control net)
out of VLAN 1. When ports are not in use, we put them in VLAN 1 and disable
them. But, if a port accidentally ends up in VLAN 1 still enabled, it could
146 147 148 149
theoretically talk to the switch and change configurations. Furthermore, we've
had circumstances in which delay nodes accidentally ended up in tis VLAN, still
forwarding packets, which created a very nasty forwarding loop. Using some
VLAN other than #1 prevents such problems from impacting the boss node.