* Take rate input as bits/s
* Update README.md
* Fix throttling to be as precise as possible.
* Minor doc fix.
* Update README.md
* Add support for title, improve handshake code.
* Update README.md
* Intermediate checkin
* Last check-in before deleting ACK message in Ethr.
* Support for client port, throttling and tos almost working.
* Most functionality working as expected with code all cleaned up.
* Linux/OSX Fixes.
* Fix handshake mechanism.
* Minor cleanup.
* More improvements for external mode.
* Improve admin-mode, root user permission checking.
* Improve detection of IP version for ICMP.
* Update README.md
* Simpler single port version of Ethr working now.
* More simplification to use single port
* Minor changes
* More fixes to enable single port and no control channel sync.
* Further improve code, merge client & xclient, add conn latency for both, pretty print.
* Intermediate change
* Initial traceRoute is working.
* Cleanup MTR functionality.
* Organize help better and add validations.
* Update README.md
* Update README.md
* Make ICMP ping and traceRoute working.
* Update README.md
Co-authored-by: Pankaj Garg <pankaj@intellectualheaven.com>
* Remove hardcoded buffer of 1 from latency test on server and client(#1)
* Remove hardcoded buffer of 1 from latency test on server and client
* Default latency test (#2)
* Default latency test so we dont introduce regression
* Fixed /proc/net/snmp parsing bug
* Fixed a bug when parsing /proc/net/snmp for Linux machines
* Removed unused, commented code
* Changed error messages to Go best practices (wrapping-friendly)
* * Upgrade Dockerfile
* Upgrade to Go 1.13
Initial version of -r for TCP bandwidth tests + few other enhancements:
- Add timestamp in log messages.
- Handle the case where timer tick is delayed on client side. Server side still
needs work.
- Use high value of GOMAXPROCS to fix starvation of some goroutines.
- Support option to disable per connection statistics. If large number
of sessions are used, then per connection stats were making it hard
to read the results on console.
- Synchronize stats timer between client/server. This makes server and
client print similar test results. Earlier, under varying TCP throuhgput,
results in server and client side were different due to different time
periods for calculating throughput.
Initial version of HTTPS support. This uses a dynamically generated self-signed certificate. The main use case is to measure performance of middle boxes like firewalls and proxies.
Initial version of TCP ping.
This needs enhancements listed below, but is a workable solution (for now):
- Better goroutine synchronization for exit tracking
- Code cleanup for handling of warmup run
- Support for custom number of warmup connections
- Support for custom sleep between successive connections
- Option to select -4 or -6 to force IP version
- Logging the results in json format
* Intermediate changes for Connected UDP.
* Add more changes to support connected UDP.
There is an issue with this approach, as multiple sockets are not able
to listen on connected UDP. We need to add SO_REUSEPORT to make this a
viable solution. This would be done in a later commit but before this
change is merged into master.
* Minor code cleanup.
Ethr is a network performance measurement tool. It allows measurements
of bandwidth, latency, packets/s, connections/s for multiple protocols
such as TCP, UDP, HTTP, HTTPS and ICMP.
