3 ms·

It reminds me of the ReVirt [1] paper read in an advanced OS class (actually mentioned in the slides). I didn't watch the complete talk. Wondering how much is i

by lnyng 8y ago

It reminds me of the ReVirt [1] paper read in an advanced OS class (actually mentioned in the slides). I didn't watch the complete talk. Wondering how much is it different from ReVirt and other record debugging tools. [1] https://www.usenix.org/legacy/events/osdi02/tech/full_papers/dunlap/dunlap.pdf https://www.usenix.org/legacy/events/osdi02/tech/full_papers...

Mic92 8y ago
It is probably more mature then a research prototype as it has received more real-world testing.
DSingularity 8y ago
ReVirt logs non-deterministic inputs at the hypervisor level. RR will be recording these inputs at the Kernel level. So this makes the inability to replay the OS execution the biggest difference.
- MarkUndo 8y ago
  Interestingly, hypervisor-level logging of this stuff is both harder (because it has the constraints of kernel-level code) and simpler (because the non-deterministic behaviours are fewer and better-documented at the hardware level than the Linux API level!) I think it's very likely that, overall, recording a single process is substantially more complex to implement than recording a whole VM. (With significant caveats - recording a whole VM with good performance is going to be hard and making it really useful probably is a whole load of extra code)
roca 8y ago
The basic idea is the same as ReVirt, but there are a lot of different details because rr has to run as a pure user-space Linux application whereas ReVirt was baked into the hypervisor. For rr to run efficiently we have to use Linux APIs in creative ways. https://arxiv.org/abs/1705.05937 https://arxiv.org/abs/1705.05937 has more details.
MarkUndo 8y ago
VMware used to support record / replay of VMs in their commercial product, which was quite cool: https://pubs.vmware.com/ws71_ace27/wwhelp/wwhimpl/js/html/wwhelp.htm?context=ws_user&file=record_replay.14.2.html https://pubs.vmware.com/ws71_ace27/wwhelp/wwhimpl/js/html/ww...

by lnyng 8y ago

Mic92 8y ago

It is probably more mature then a research prototype as it has received more real-world testing.

DSingularity 8y ago

ReVirt logs non-deterministic inputs at the hypervisor level. RR will be recording these inputs at the Kernel level. So this makes the inability to replay the OS execution the biggest difference.

MarkUndo 8y ago
Interestingly, hypervisor-level logging of this stuff is both harder (because it has the constraints of kernel-level code) and simpler (because the non-deterministic behaviours are fewer and better-documented at the hardware level than the Linux API level!) I think it's very likely that, overall, recording a single process is substantially more complex to implement than recording a whole VM. (With significant caveats - recording a whole VM with good performance is going to be hard and making it really useful probably is a whole load of extra code)

roca 8y ago

The basic idea is the same as ReVirt, but there are a lot of different details because rr has to run as a pure user-space Linux application whereas ReVirt was baked into the hypervisor. For rr to run efficiently we have to use Linux APIs in creative ways. https://arxiv.org/abs/1705.05937 https://arxiv.org/abs/1705.05937 has more details.

MarkUndo 8y ago

VMware used to support record / replay of VMs in their commercial product, which was quite cool: https://pubs.vmware.com/ws71_ace27/wwhelp/wwhimpl/js/html/wwhelp.htm?context=ws_user&file=record_replay.14.2.html https://pubs.vmware.com/ws71_ace27/wwhelp/wwhimpl/js/html/ww...