Recording traffic at high rates (10/40/100 Gbit and above) usually requires ad-hoc/manual
configurations, as in that case accelerated capture technologies like PF_RING ZC with 
Intel, Mellanox or FPGA adapters are required, and n2disk needs to be properly configured
and tuned to take fully advantage of the hardware architecture (CPU affinity, storage, etc).

In order to support custom n2disk configurations, ntopng can also attach to external
n2disk instances/services which have been configured manually using systemd, as opposite
to using ntopng for automatically configure and start the recording using the GUI.
In fact, ntopng can bind to an external traffic recording provider by selecting it from
the interface settings page. A dropdown menu with the list of available recording providers
is shown there.

.. figure:: ../img/web_gui_interfaces_recording_external_providers.png
  :align: center
  :alt: Traffic Recording External Providers Selection

  The Selection of External Traffic Recording Providers

.. note::

   In case n2disk processes are managed manually using configuration
   files, ntopng will not show a settings tab nor it will allow any
   configuration change. However, extractions will still be possible
   as described in the following section.

Please also note that ZC and FPGA adapters are based on kernel bypass technologies,
which provides the best capture acceleration, with the drawback that they do not (usually)
allow multiple applications to capture the same data stream at the same time (with some
exceptions like Silicom/Fiberblaze adapters and Mellanox, depending on the configuration).
This means that ntopng and n2disk cannot capture and process the same traffic at the same
time. For this reason, in order to overcome this limitation, n2disk is able to export 
flow metadata to ntopng over ZMQ, similar to what nProbe does (as explained in the 
Using ntopng with nProbe section). In this configuration n2disk can be configured to 
capture raw packets, dump PCAP data to disk, and export flow metadata in JSON or TLV 
format through ZMQ to ntopng for visibility.

In order to process flow metadata through ZMQ in ntopng, the collector interface needs
to be added in the ntopng configuration file (/etc/ntopng/ntopng.conf):

.. code:: bash

   -i=tcp://*:5556c

The same endpoint needs to be added to the n2disk configuration file (e.g. /etc/n2disk/n2disk-eth0.conf):

.. code:: bash

   [...]
   --zmq=tcp://127.0.0.1:5556
   --zmq-probe-mode
   --zmq-export-flows
   -u=ntopng

Please note that it is a good practice to run n2disk using the ntopng user (see -u)
in order to make sure that ntopng is able to access the PCAP data recorded by n2disk
and run traffic extractions.

Please check the n2disk User’s Guide for a full example of n2disk configuration and
further information about other options.

Traffic Extraction
------------------

All pcap files dumped to disk are indexed on-the-fly by *n2disk* to improve traffic 
extraction speed when recorded data need to be retrieved.
It is possible to extract traffic from multiple places in *ntopng*, including the interface
and the host *Historical Traffic Statistics* pages. 

After enabling continuous traffic recording on an interface, a new button for extracting
traffic appears at the top right corner of the *Historical Traffic Statistics* page.

.. figure:: ../img/web_gui_interfaces_extract_pcap.png
  :align: center
  :alt: Extract pcap button

  The Extract Pcap Button in the Interface Historical Traffic Statistics page

.. figure:: ../img/web_gui_hosts_extract_pcap.png
  :align: center
  :alt: Extract pcap button

  The Extract Pcap Button in the Host Historical Traffic Statistics Page

By clicking on the button, a dialog box will let you run an extraction to retrieve the 
traffic matching the time interval selected on the chart. It is possible to download the
extracted traffic directly (this should be used when the expected amount of extracted 
traffic is low) or queue the extraction job to process traffic in background (this should 
be used for extractions taking too long, or to archive extracted data on the machine running 
ntopng).

.. figure:: ../img/web_gui_interfaces_extract_pcap_dialog.png
  :align: center
  :alt: Extract pcap dialog

  The Extract Pcap Dialog

In addition to the time constraint, it is possible to configure a BPF-like filter, 
to further reduce the extracted amount of data, by clicking on the *Advanced*. button 
The filter format is described at `Packet Filtering <https://www.ntop.org/guides/n2disk/filters.html>`_.

.. figure:: ../img/web_gui_interfaces_extract_pcap_dialog_filter.png
  :align: center
  :alt: Extract pcap dialog filter

  The Extract Pcap Dialog Filter

The extraction button is also available in several other places while browsing the
historical data, an example is the list of the *Top Receivers* or *Top Senders* available 
at the bottom of the *Interface Historical Traffic Statistics* page. In this case, a button
on the right side of the row lets you download the traffic matching a specific host in the
selected time interval.

.. figure:: ../img/web_gui_interfaces_extract_pcap_from_list.png
  :align: center
  :alt: Extract pcap button

  The Extract Pcap Button in the Top Receivers in the Interface Historical Traffic Statistics Page

When an extraction job is scheduled for background processing by selecting the *Queue as Job* option, 
ntopng extracts the traffic and creates new *pcap* files with the traffic. This usually requires a few 
seconds, depending on a few factors, including: the time interval, the amount of recorded data, the 
extraction filter. 

A reference for the extraction job (a link to the *Traffic Extraction Jobs* page with the list of 
scheduled extractions, and the extraction *ID*) is provided after starting the extraction, in order 
to control the status and download the pcap file(s) as soon as the extraction is completed.
Extraction jobs can be stopped anytime using the *Stop* button, in case of extractions taking too 
long, or removed using the *Delete* button (this will also delete the corresponding pcap files).

.. figure:: ../img/web_gui_interfaces_extraction_jobs.png
  :align: center
  :alt: Traffic Extraction Jobs

  The Traffic Extraction Jobs page

It is possible to access the *Traffic Extraction Jobs* page also by clicking on the badge that 
appears on the top status bar when there is at least one extraction job scheduled.

.. figure:: ../img/web_gui_interfaces_extraction_badge.png
  :align: center
  :alt: Traffic Extraction Jobs Badge

  The Traffic Extraction Jobs Badge in the Top Status Bar

REST API
~~~~~~~~

The pcap file can also be downloaded directly through http, running a live extraction. 
It is possible to use a command line tool such as `wget` or `curl` for this.
The direct URL for downloading the pcap is 
:code:`http://<ntopng IP>:3000/lua/rest/get/pcap/live_extraction.lua?ifid=<id>&epoch_begin=<epoch>&epoch_end=<epoch>[&bpf_filter=<filter>]`

Where:

- *ifid* is the interface Id as reported by *ntopng* in the interface page
- *epoch_begin* is the start of the time interval to be extracted (epoch)>
- *epoch_end* is the end of the time interval to be extracted (epoch)>
- *bpf_filter* is a filter in `nBPF <https://www.ntop.org/guides/n2disk/filters.html>`_ format

Please note that you should use cookies for authentication, as explained in the documentation. 
For example with `curl` you can specify username and password with :code:`--cookie "user=<user>; password=<password>"`

Command line tools are useful for example to process a pcap stream and pipe it to an analysis tool such as `tcpdump` or `tshark`/`wireshark`. 
For example, to process the extracted traffic directly with `wireshark`, it is possible to use `curl` as in the example below:

.. code:: bash
	  
   curl -s --cookie "user=admin; password=admin" "http://192.168.1.1:3000/lua/rest/get/pcap/live_extraction.lua?ifid=1&epoch_begin=1542183525&epoch_end=1542184200" | wireshark -k -i -

.. _ZCSupportSection:

ZC/FPGA Support
---------------

If you need to process traffic at high rate at 10/40Gbit and above, you are probably looking for
capture technologies like `PF_RING ZC <https://www.ntop.org/guides/pf_ring/zc.html>`_ for Intel
or `FPGA <https://www.ntop.org/guides/pf_ring/modules/index.html>`_ adapters.

As both PF_RING ZC and FPGA adapters are based on kernel bypass, the drawback is that they do not 
allow you to capture the same stream from multiple applications at the same time. This means that
you cannot run ntopng for traffic analysis and n2disk for raw traffic recording at the same time
on the same interface.

In order to overcome this, n2disk is able to export flow metadata to ntopng over ZMQ, similar to
what nProbe does as explained in the `Using ntopng with nProbe <https://www.ntop.org/guides/ntopng/using_with_nprobe.html>`_ section. 
As depicted below, in this configuration n2disk can be configured to capture raw packets, dump PCAP 
data to disk, and export flow metadata in JSON format through ZMQ to ntopng at the same time. 

.. figure:: ../img/n2disk_zmq_export.png
  :align: center
  :alt: Support for n2disk ZMQ export

  Support for n2disk ZMQ export

Following is a sample configuration of n2disk and ntopng to achieve what has been depicted above. 
This example assumes that both n2disk and ntopng are running on the same host.

*ntopng Configuration File*

In order to process flow metadata through ZMQ in ntopng, you need to add a collector interface to
the configuration file (*/etc/ntopng/ntopng.conf*):

.. code:: bash

   -i=tcp://*:5556c

*n2disk Configuration File*

The ntopng endpoint should be added to the n2disk configuration file (e.g. */etc/n2disk/n2disk-nt01.conf)* 
using the :code:`--zmq` option.
The :code:`--zmq-probe-mode` option (if ntopng is running as a collector like in this example: notice 
the small :code:`c` in the ntopng endpoint) and the :code:`--zmq-export-flows` option (to export flow 
metadata in addition to traffic statistics) are also required.

It is a good practice to run n2disk using the *ntopng* user (see :code:`-u`) in order to make sure that
ntopng is able to access the PCAP data recorded by n2disk and run traffic extractions.

Please see the `n2disk User's Guide <https://www.ntop.org/guides/n2disk/index.html>`_ for further information
about the other options. Please note that in the example below n2disk is aggregating traffic in hardware
from 2 ports of a Napatech adapter, please see the `Napatech configuration <https://www.ntop.org/guides/n2disk/napatech.html>`_
for configuring the adapter.

.. code:: bash

   --interface=nt:0,1
   --dump-directory=/storage/n2disk/pcap
   --timeline-dir=/storage/n2disk/timeline
   --disk-limit=80%
   --max-file-len=1000
   --buffer-len=4000
   --max-file-duration=60
   --index
   --snaplen=1536
   --writer-cpu-affinity=0
   --reader-cpu-affinity=1
   --compressor-cpu-affinity=2,3
   --index-on-compressor-threads
   -u=ntopng
   --zmq=tcp://127.0.0.1:5556
   --zmq-probe-mode
   --zmq-export-flows

At this point you should start both the ntopng service (e.g. *systemctl start ntopng*) and the n2disk
service (e.g. *systemctl start n2disk@nt01*), and configure the n2disk instance as external PCAP source
for the collector interface as explained in the `External Traffic Recording Providers`_ section in order
to be able to check the n2disk service status and run traffic extractions.