A multi-core CPU is, in a way, a network; a network on a chip. Just like homes and businesses may be ‘hosts’ or nodes on a Service Providers network, or PCs and phones on an enterprise or home network, the hosts on netElastic’s network are CPU cores.
netElastic’s virtualized network includes a small army of cores, with each core performing specific tasks. For example, one core may be assigned multiple tasks, such as managing Access Control Lists (ACL) and QoS policies, while another core may be given only one task – solely supporting a 10G port and forwarding packets in and out as fast as possible.
We can even assign a small “brigade” of 10 cores to service a 100G NIC card and forward ports at near line-rate from it.
In the past, you needed an ASIC chip to do this well. Not today. Advances in Intel architecture have made forwarding packets on general-compute architectures much more cost-efficient. These advances include technologies like DPDK, DDP, and super-fast on-chip memory.
To be fair, because they are application-specific, ASICs will theoretically forward packets faster, but this will only provide visible improvements where terabits of speed are important and far greater costs are acceptable. And this is generally only in the datacenter – not with the service provider.
Just imagine a virtualized, dual-socket, single rack unit (RU), Xeon server. One socket can host netElastic virtual BNG, and the other can run a RADIUS Server, billing system, Netflix cache, and other network services.
Deploy two of these servers in a geo-redundant pair and you have the ultimate Service Provider network backbone. These can be easily fed by MikroTik routers at tower bases, netElastic vRouters in MDU’s, or Calix switches in GPON clusters.
So let the netElastic network, with our vast army of CPU cores, help your network. We’ll speed things up, simplify your support, logistics, and NOC operations, and, yes, even save you CapEx.