====== Project Title ======
**//Ringmap// Capturing Stack for High Performance Packet Capturing in %%FreeBSD%%**

\\ 

====== Project Description ======
Packet capturing in high-speed networks is not an easy task due to the system
limitations such as RAM and system bus throughput. Unfortunately, it is often
impossible to capture every packet arriving at the network interface.  Hardware
may not be the only bottleneck, however, as inefficient software is also to
blame for poor resources usage and the resulting packet loss during capture.

The new //[[http://code.google.com/p/ringmap/|ringmap]]// capturing stack for %%FreeBSD%% already allows for efficient capturing at 1GBit/sec on commodity
hardware with very low packet loss and low system load. Under similar conditions, with Intel 1GbE cards
and 64 bytes packets, the standard capturing software of %%FreeBSD%%-7.x during capturing generates a system load of up to 100% and results in close to 100% packet loss.

Similar to "//zero-copy BPF//" implementation, our idea is to eliminate packet copy operations by using shared memory buffers. However, unlike the "//zero-copy BPF//" model, ringmap eliminates **all** packet copies during capturing: the network adapter's DMA buffer is mapped directly into the user-space. //Ringmap// also adapts //libpcap// accordingly to provide user-space applications with access to the captured packets without any additional overhead.

\\ 

====== Tasks for GSoC-2010 ======

  * Refactoring //ringmap// code in order to make it portable to other network adapters
  * Porting //ringmap// to 10GbE Intel network adapters
  * Porting the //ringmap// capturing stack to %%FreeBSD%%-8
  * Measurements and performance comparison between the standard packet capturing stack of %%FreeBSD%%-8 and //ringmap//.

\\ 

====== Project Schedule ======

  * **May 24 - May 28: Ringmap refactoring**
    * Refactoring //ringmap// code to achieve more portability
      * I'm planning at first to split out the existing //ringmap// code into hardware-dependent and hardware-independent code. This will first be done by splitting the data structures and then the functions. This must be done because the present //ringmap// code was implemented only for //em//-driver and only for one type of Ethernet controller, and as a result the hardware-dependent and -independent data is mixed together within the same structures. Moreover, I have to extend the existing data structures to make the //ringmap// code more portable.

\\ 

  * **May 31 - June 4: Porting ringmap to %%FreeBSD%%-CURRENT**
    * Installing %%FreeBSD%%-8 and update it to CURRENT
    * Porting //ringmap// to %%FreeBSD%%-CURRENT

\\ 

  * **June 7 - June 20: Porting ringmap to 10Gb**
    * The //ringmap// capturing stack is currently based on //em//-driver and is usable only with Intel 1Gb Ethernet adapters. Therefore, the goal is to modify the //ringmap// capturing stack for the 10Gb Ethernet Intel network adapters based on //82598// and //82599// processors. To achieve this goal, the //ringmap// code must be ported to **ixgbe**-driver. To do it I will first insert the "hooks" for //ringmap// functionalities into the //ixgbe//-driver code  (through #ifdef). The first places where I will port the //ringmap// functions are "attach" and "detach" functions of //ixgbe//-driver. In addition to their ordinary tasks, these functions must be able to allocate and deallocate the //ringmap// data structures properly. After this step I will begin porting most complicated part of the code that has to be executed as a result of interrupt.

\\ 

  * **June 21 - June 23: Debugging**
    * These three days should be for debugging the code and also for implementing in case I didn't manage something in previous steps 

\\ 

  * **June 24 - July 4: Functionality tests**
    * Deploying a tesbed network of three hosts and implementing the software for tests control and measurements (the software is already partly implemented). These three hosts should be for: 
      * Traffic generation 
        * OS Linux. Will be done using Linux Kernel Packet Generator
      * Capturing
        * OS %%FreeBSD%%. //ringmap// vs. standard %%FreeBSD%% capturing software
      * Test control
        * Scripts for the central coordination of tests
    * Conducting experiments

\\ 

  * **July 5 - July 16: Porting ringmap to %%FreeBSD%%-8**
    * This step will be the port to %%FreeBSD%%-8, but it requires that all previous steps are finished (at least the coding part, but not necessarily all testing work)

\\ 

  * **July 19 - July 23: Functionality and Performance Tests**
    * Functionality test and cosmetic debugging of %%FreeBSD%%-8 //ringmap// capturing stack.
    * Conducting performance analysis and evaluation between the standard packet capturing stack of %%FreeBSD%%-8 and //ringmap//. To perform effective performance evaluation, several servers for traffic generation will be used together in order to reach greater-than 1Gibt/sec traffic.

\\ 

  * **July 26 - August 15: Documentation and the remaining work**
    * Debugging, implementing small requested features, documentation

\\ 

====== External Links ======
  * [[http://code.google.com/p/ringmap/|Project-Page on Google-Code]]
  * [[http://ringmap.googlecode.com/files/ringmap_slides.pdf|Ringmap Presentation]]
  * [[http://www.FreeBSD.org/cgi/man.cgi?query=bpf&apropos=0&sektion=0&manpath=FreeBSD+8-current&format=html|FreeBSD 8.x BPF(4) man page, including description of zero-copy BPF buffers]]
...