2 PCs, 2 peripherals. I’m thinking of using a vlan to timeshare the peripherals under test, so that I can run tests on windows and on linux, but not have to duplicate on hardware variations/generations. anyone used a managed switch to configure vlans and make the patching happen by the black art of vlan configuration?
Today I have 4 peripherals in total, so (I lie I have 5) , and if I plug 2 PCs into that I can use a 8 port managed switch to drive , as long as I can work out what technology stack will be best suited for setting up vlans between the 2 PC’s and the hardware. At some point IU guess I can stack the managed switches up to my hearts content. anyone ever done something like tis before I go out and buy a switch and discover that it’s a pain to remote control it?
I’ve setup vlans on static networks before but never on demand… it’s an interesting idea!
I’d check your chosen switch allows configuration via SNMP or SSH/Telnet if you plan on automating the port-vlan id associations.
It’s been a long time since I did anything with vlans but I remember it being a faff to configure initially, but once I’d gotten over the initial setup/configuration/understanding hurdle for that particular switch vendor, it was straightforward.
One thing I would say is that switch specs and manuals didn’t really explain the more technical aspects of switch capabilities back then (missing info and blind spots when trying to determine if a switch does everything you want/need it to do).
Personally I’d aim for a mid range 24 port switch as you’ll probably get more capabilities.
Take this info with a pinch of salt as it’s based on what was true 15 years ago!
Edit: the more I think about this, the more I wonder if you would be better off doing this with dedicated subnets and firewall rule and/or NAT/routing in software. You could probably achieve the same thing with a Linux box…
The real reason I’m wanting to Vlan this, is about sharing hardware, and at the same time allowing for broadcast control and DHCP behaviour. Which perhaps is back to your point about subnets being more sensible, because I can then validate that broadcast nets are correct… which rolls back to DHCP server configuration. I guess I really have 2 areas to test, which pretty much means I have to either write a backlog Jira ticket now, or just update my test plan. I’m tempted to update my test plan, because Jira is starting to look more like a test-plan-ideas-list than a backlog.
You are right Mark, need/want a 24port switch, but for 2.5 Gig those are a bit pricey so I have to be realistic.
A bandwidth limiter device would be sweet, I need to see if I can use a linux box to do this for me later.
I’m not testing Windows/Linux networking at all, the product under test is designed to function on an isolated LAN, all devices support DHP only, no static addressing, only DHCP assignment.
UDP broadcasts are used for discovery (and DHCP of course) data is send at max possible rates and the tests need to validate speeds up to 2.5gbits. Because of the need for network performance tests, I don’t really want to use subnets.
Mostly each PC will have 2 NICs, the test NIC is going to always be a 2.5G adapter
Each device under test will have a 1GB or 2.5Gb NIC
If I can control and also run more than one DHCP server, I can get more power to stress test things - all managed switches have built in DHCP, which might come in handy. Usually each PC runs DHCP sever on the isolated NIC, but that’s not a product requirement to do it that way.
I want to be able to patch 4 or more devices under test into different 2 or more PCs. This removes the need to duplicate on hardware and leak subnets - subnets do still need some separate testing
I think I’ve just written an update to my test plan above. I love how a chat lets me bounce ideas off of people for discovery of more test cases.
I’m going to look into buying a Zyxel managed switch, I’m using an **un-**managed one like this one at the moment (Unmanaged).
