#---------------------------------------------------------------------------
@() $Name: cflowd-2-1-b1 $
#---------------------------------------------------------------------------
Please see the file named INSTALL in this directory for installation instructions.
Changes from 1-3b2 to 2.0
cflowd has been completely redesigned and reimplemented for the 2.0 release. The changes include:
- Added support for v1 flow-export.
- All tables are now per input interface.
- New tabular data: port matrix, interface matrix, nexthop table. The old port table has been replaced by the more granular port matrix.
- a new cflowdmux process with which clients may communicate to receive raw flow streams.
- A fully functional central collector is now included (cfdcollect). This allows you to archive time-series tabular data from multiple instances of cflowd.
- All counters are 64 bits.
- Local clients (cfdases, cfdnets, et. al.) will show the time interval for current data.
- Local clients can show pkts/sec and bits/sec in addition to packet and byte counters.
- Added manpages.
- mmap() is gone for the tabular data; local clients connect to a UNIX domain socket to view current data. This removed a lot of code complexity.
Things still missing:
- Need to store process IDs in pidfiles so I can easily manage the processes (cflowdmux, cflowd and cfdcollect).
Notes on source code
It has become my habit to document function calls in the header files and not in the implementation. I do things this way because the documentation of functions, classes and members is then still available in a binary installation, assuming the header files and libraries are also installed. It also lets me use a modified version of cocoon to generate class reference documentation in HTML. I don't put function documentation in the implementation because it becomes painful to duplicate documentation in multiple places (and like anything else that's duplicated, it has a tendency to get out of sync).