My company has a couple of access switches in the core network and dismiss my claims that we need a proper core network. I'm prepping a cost/benefit analysis report and would like to hear /networking 's input.

Hi /networking

The Situation:

I work for a company that is scaling up from being a startup to a proper enterprise. On my first day with the company, I took a pen and paper and started making hand drawn network diagrams in order to familiarize myself with their network. I was a bit surprised in the way that they had designed and setup the network in that they have a completely linear setup.

There are two virtualization hosts running as a cluster with both connecting to a single 24 port HP Procurve 1810. That Procurve has a LACP channel connecting to a single HPE OfficeConnect 1920s which is functioning as the core switch. Most of our server infrastructure runs on this cluster as well as the controller for our Ubiquity wireless system.

There are two of these HPE OfficeConnect 1920 switches in the "core", but they aren't stackable nor are they directly connected to each other. We have most of the APs running into these HPE OfficeConnect switches.

All of the VLANs terminate on a Ubiquity Edge Router that serves as the firewall and demarcation point of our network before hitting the ISP.

At this time, I have identified 4 single points of failure: HP Procurve Switch, 2 HPE OfficeConnect switches, a single SFP with a fiber connection to the CPE device from the ISP. If any of these devices fail, connections to the virtualization cluster also fails. If we lose the SFP or the main HPE OfficeConnect, no one has internet access and for a company that really leans into the use of cloud applications, no internet access means no one is working.

My Plans to Fix:
Over the past weekend, I grabbed a couple of my guys for us to put our heads together to see how we can make our network more resilient so that should we have a hardware failure, we won't be dead in the water.

This proved to be futile since I can't stack the HPE OfficeConnect switches, so that means no LACP channels. I briefly considered interconnecting all of the switches to form a more mesh topology and then implement MSTP. But then I said to myself.... why TF am I setting up MSTP? This 2019 dammit.

So I explained this to management and I've been told that for the needs of the company, our current setup is sufficient. I am crawling up the walls about management not taking this risk seriously. I'm starting a cost/benefit analysis report to show them that continuing to operate like this is reckless for a company of 90 people.

I am preparing for an uphill battle with management over this and before I embark on this journey, I need to make sure I am looking at this situation from all angles. Would any of you with any experience in doing this before offer some advice or suggestions so that I will increase my chances in convincing management that we need to do something?

Thanks!!!

​