Skip to content

netobserv/pano-research

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Payload Aware Network Observability (PANO)

Welcome to the project site for Payload Aware Network Observability (PANO).

The goal of this project is to provide observability to networking levels 5, 6, and 7. Typical observability tools only look at layers 3 and 4, outputing flow logs, which summarize traffic flow by 5-tuple. A great deal of information can be captured looking at higher level protocols, such as DNS, HTTP, LDAP, SSL/TLS, SIP, etc. As an example of PANO's utility, we present a demo using DNS.

While several approaches to this problem are possible, we sought to take a balanced one, in terms of trading off efficiency vs. generality. For example, while we could do everything in kernel space with eBPF, that would require constant changes to the kernel. Similarly, using python for everything would be fast and simple, but slow due to the generality. Instead, we strike a middle ground: we use eBPF in the kernel for packet filtering and capture, for performance reasons, and C++ or Go in user-space for efficient flexibility. Where possible, where it makes sense, we re-use existing industry-standard tools.

Background

As an observability tool, PANO consists of a pipeline of service components. The pipeline captures raw packets, analyzes them, outputs metrics, and displays them on a dashboard. A major difference with other tools is that it does not depend on the support of a specific server (e.g., a specific DNS server such as coredns). It uses raw packets captured on the network to reconstruct the behavior of the server.

More detail can be found in a PPT deck here.

Pipeline Components

Changes We Made

Changes needed to the components consisted of the following:

Use Case: DNS

To illustrate PANO, we demonstrate a use case with DNS. We configure the NetObserv eBPF Agent to capture all packets on UDP Port 53, the default DNS port.

Workload

In addition to our pipeline, we use two instances of a DNS load generation tool, DNS-OARC's dnsperf. One instance queries 8.8.8.8 (Google's DNS server) and another instance queries 9.0.0.1 (IBM's DNS server) simultaneously, using a sample DNS querylist from a trace taken at yorktown.ibm.com.

Running the Demo

We have a demo of PANO's capabilities running in Docker.

A recording of the demo is available here.

Requirements

You will need the following to run the demo:

  • A Linux bare-metal machine or VM (we tested on Ubuntu 22.04.3 LTS).
  • A relatively new Linux Kernel that supports CAP_BPF in Docker (tested on 5.15.0-78-generic).
  • Git - to check out the source tree (should be included in your Linux).
  • Bash - to build the docker images from standard components (should be included in your Linux).
  • Docker and docker-compose (or "docker compose") that support "host" networking (for Ubuntu get them here and here).

Check-out the Demo

Run

git clone [email protected]:netobserv/pano-research.git

to check out the source tree.

Building the Demo

Run build-me.sh in the demo1 subdirectory:

build-me.sh

You should see various component images downloading and some building.

Running the Demo

In the demo1 subdirectory:

Run docker compose up -d

docker compose up -d

OR docker-compose up -d

docker-compose up -d

depending on which version of docker and docker-compose you have.

You should see the various components come up:

Creating netobserv-ebpf-agent ... done
Creating zeek                 ... done
Creating kafka                ... done
Creating flowlogs-pipeline    ... done
Creating prometheus           ... done
Creating grafana              ... done
Creating dnsperf1             ... done
Creating dnsperf2             ... done

Now in a browser window, go to:

http://pano.sl.cloud9.ibm.com:3000/

(if you are running on a different machine, substitute your-machine-name for pano.sl.cloud9.ibm.com)

The first time you do this, you will see the following:

Grafana Login Page

Grafana is asking you to log in. The credentials are admin and admin. If you change them, you must remember them for next time. Our advice is to not change them and skip the password change.

You may be presented with a Dashboard page. If so, click the Home button in the upper left.

You should see a Welcome page:

Grafana Welcome Page

In the lower left corner, click on the PANO DNS Metrics.

You should see something like the following:

Grafana DNS Page

At first, the page will not display anything, but in about 10-15 seconds it should start to populate.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published