Hacker News Reader

To be clear, 80% of kernel-time in that 1KB write is spent in fsync(), _not_ in the network stack. Network overhead is roughly similar between read & write requests. What Arrakis seems to be able to do is avoid the overhead of write & sync, presumably because it doesn't go through the VFS + filesystem + block IO code paths.

View on HN

>To be clear, 80% of kernel-time in that 1KB write is spent in fsync(), _not_ in the network stack

Are you sure?

Of the total 3.36 μs (see Table 1) spent processing each packet in Linux, nearly 70% is spent in the network stack

Table 1 is looking specifically at getting a chunk of data off the wire and into the users code. Look at Table 2 for a comparison of redis read/write. Average time for a write is 163 μs - of which 137 μs is spent in fsync(2)

Right but you can't compare table 2 data to the network stack, table 2 data is only timing the redis operations. Which as stated take up significantly less time than the time spent in the network stack.

Reply on news.ycombinator.com