Internet exchanges are a key internet infrastructure that boost network connectivity and reduce internet latency and bandwidth costs. OpenSwitch IX is India’s first virtual internet exchange. Internet exchanges help ISPs reduce bandwidth costs while customers experience lower latency while accessing content. Content providers get more eyeball traffic to their networks. It is a win-win situation for both the ISP and their customers and content providers.
The main purpose of a virtual internet exchange is learning for hobbyists and peering for personal ASNs. They may at time help in reducing latency but that may not be the case all the time. However, it makes it easier to maintain peering sessions as only a single route server bgp session can get you peered with a lot of networks. The virtual internet exchange culture is very popular in Europe and US just like the number of physical internet exchanges. Due to wonderful routing by TCL which happens to be my upstream’s upstream, latency to European networks is generally 180-230ms. Also a virtual internet exchange in particular which I won’t be naming has this tantrum of asking everyone to buy their overpriced VMs to access their internet exchange as most of the time their tunnels just don’t work. So my friend Vineeth Penugonda and I had had enough of them and decided to open this virtual internet exchange as a hobby project which we would maintain. It would be a virtual IXP for hobbyist networks however we got two ISPs from Brazil to peer with us.
After we decided to start the virtual internet exchange we then had to design on the topology design. We got a VM from my friend Dipan Roy’s production proxmox server at his home. Being connected with the same ISP helped us get nearly 1-2ms of delay between our nodes. Next we decided to implement some sort of fault tolerance for the network. Both our core routers are connected over gre tunnel. His proxmox VM hosts our Route Server 1 and zero-tier tunnels terminate on it. The linux bridge on Router Server1 and Route Server 2 are connected over a vxlan tunnel. GRE and VxLan tunnels terminate on Route Server 2 at my end. So if any of the Nodes are down, not all peers are affected. For those without static ip we offer zero-tier to peer. However, VxLAN is our preferred method of peering. EoIP will also be implemented soon. This project uses VyOS which has turned out to be a fabulous choice as it does everything from recursive routing on ipv6 to vxlan, everything that RouterOS cannot.
Currently, we have our only PoP in Kolkata. However, a physical PoP is planned at Fremont, US. Asia Pacific networks shall experience lower latency while connecting to Openswitch IX as compared to European virtual internet exchanges. If you have a public ASN you can join us here: openswitchix. If you are wondering how to get started with BGP with as little as $25 read my other article here. Route collector server is on our to do list. Also BGP community based filtering will also follow soon.