Wrong. You’re like an order of magnitude wrong - rofl there is no fucking way the stock Linux kernel will even do 40Mpps at 64 byte packets. It chokes way before that. This is partly why things like DPDK exist.
16 iperf threads...sending at what packet size? Do you understand the notion of line rate? 85Gbps at 1500B is only 7MPPS, which is half of 10Gbps at line rate.
You're both right. It's an older term from the early 90's when a router's selling point was being able to hit "line rate" with the smallest possible packet size. Example, how many tiny datagrams can you forward to fill that link. Back then, people were still doing lots of routing on general purpose machines and Cisco/Juniper were just starting to get into the high performance game.
These days line-rate just means sending enough traffic to fill the link at whatever rate you want. That's generally good enough for server folk since they just want to get you the cat pics ASAP.
That's not good enough for people running transit networks, since they care more about packets per second performance. Sending huge amounts of data is easy for them; what they really care about is PPS.
Aside, the next generations of router NPU's are trash in terms of PPS performance. I take that back, they're not trash. They're the trash in the dumpsterfire. That's how bad they are. We're fairly screwed there.
My guess is GoDaddy was looking at increased PPS performance either for DNS or maybe building their own DDoS mitigation framework (Arbor gear is pricey).
I've never heard this weird qualification for the definition of "line rate" that it somehow requires minimum packet size, so I looked it up. The first three sources for a quoted big-g search all imply or directly state that it's the same as bandwidth:
Line rate does imply pps at the smallest sized frames in the context of networking equipment performance. Vendors use it extensively in their docs.
64B is the minimum frame size in Ethernet, including interframe gap and preamble its 84B on the wire. It is the same with Ethernet, Gigabit Ethernet and even 100Gbit Ethernet, that source is not correct.
No line rate does not "imply pps at the smallest sized frames."
Network hardware always quote PPS using the smallest sizes. And this makes sense for things like route and switch processors. Perhaps you are confusing that.
You should reread your link a little more carefully. From your link:
">However it is also important to make sure that the device has the capacity or the ability to switch/route as many packets as required to achieve wire rate performance."
The key phrase there is "as required." Almost nobody needs to sustain forwarding Ethernet frames with empty TCP segments or empty UDP datagrams in them. In fact many vendors will spec for an average size. Since packet size x PPS will give you your throughput, if the average packet size is larger you need much less PPS to achieve line rate.
Line rate doesn't imply small packets. But most userspace benchmarking uses 64B packets. That being said, the "imix" packet size, which is supposed to represent internet traffic, is around 500B.
There are numerous imix definitions floating around now, it really depends on who you ask. The more realistic ones define a pattern of different sized packets. And performance varies greatly depending on which imix flavour you use, which probably goes back to the earlier poster's 'dumpster fire' comment (although I don't know exactly what NPU generation they are referring to).
No, PPS and bandwidth are two distinct metrics. Although there can be a linear relationship between them line rate does not always imply a payload equal or close to the interface MTU. You see this with network vendors and some of the higher end gear. Network vendors always give specs for their gear by quoting both metrics. And there is some gear that that is capable of doing line rates even with small packets, example the Cisco ASR 1000:
"For example, because one of the newest Cisco routers, the Cisco ASR 1000 Series Router, is capable of forwarding packets at up to 16 Mp/s with services enabled, it can support the processing of the equivalent of 10 Gb/s of traffic at line rate, with services, even for small packets."[1]
The point of discussion here is that the Linux kernel struggles to do line rate 10Gbps. This was misinterpreted as "the Linux kernel struggles to do 10Gbps".
