Letter from the Editor
As we ring in the New Year, it is a good time to scale back some of the clutter in your life. A good place to start with is your network.
One major cause of clutter in your network is ad-hoc growth, adding links and bandwidth without a scalable plan. By scalability, we mean a plan and architecture which allow for future growth without congestion, because congestion in the wrong choke point will reduce bandwidth and increase latency for everyone in your network. It's not an easy task to accomplish without insight into your current network, but it's even harder if your plan doesn't include provisions for maintaining visibility in your growing network.
For this Peeks, we will outline some scalability solutions that WildPackets offers in order to understand and optimize your network's performance.
Hope everyone survived the holidays, we're looking forward to working with you in 2013!
Marketing and Events Manager
OnDemand Webcast: Scalability and Omni Distributed Analysis Platform
Watch this OnDemand Webcast to see how WildPackets scales across all facets of network analysis, and continues to push the boundaries in high-speed, highly distributed network analysis and troubleshooting with a single distributed cost-effective solution.
In this OnDemand Webcast, we cover:
- The key areas of scalability that must be addressed by a network analysis solution
- Best practices in addressing key areas of scalability
- Practical distributed network analysis scenarios
You will learn how to:
- Deal with highly interconnected 10G and 40G networks
- Eliminate choke-points without eliminating visibility
- Design distributed network analysis solutions to meet various scenarios
Focus On: Best Practices for Network Management in Distributed Environments
Networks today are composed of an amalgamation of resources across wired and wireless infrastructures spanning different regions. This complexity can make it difficult to identify the root cause of network latency, packet loss, and wireless problems. The solution? Start with a plan:
- Monitor at both the client and the server.
- Benchmark your application response time – the time it takes an application to respond to a specific user request.
- Simplify your configuration changes, and use alarms and filters to monitor changes.
- Store your data for historical analysis.
- And most importantly, if your company runs applications over the web, don't forget to monitor and analyze those cloud-hosted services.
Below are some best practice guidelines to get you started:
Product Tip: Setting Analysis Options for Performance
Modern network analysis tactics require a constant collection of network data across multiple endpoints. Being able to manage simultaneous captures from multiple network adapters runs the risk of inadvertently asking your network analysis solution to collect packets more rapidly than the CPU and system bus can process and more quickly than the hard drive controller can write to disk.
Here are three tips for optimizing the performance of your network analysis solution.
Tip 1: Match the solution to your environment.
- For a NOC: choose a solution specifically designed for 10G, such as TimeLine, with enough disk space to cover the history you need, overnight, a weekend, 30 days, or something else.
- For server farms: select a solution that is able to see traffic traversing within the virtual server – "hidden traffic" that never crosses a physical NIC.
- For remote offices and widespread campuses: pick a solution that can handle mobility with very dense yet ephemeral usage.
For more details, read Distributed Networks: Best Practices for Selecting Analysis Options.
Tip 2: Anticipate hardware resource needs and know your limits.
Today's highly-utilized high speed networks can quickly consume a lot of resources. If you're capturing steady-state traffic of 1 Gbps, you need at least 6TB (460GB/hr) to maintain a record of overnight activity and at least 22TB (11TB/day) for weekend activity. For steady-state traffic of 10 Gbps, you're looking at 60TB overnight and 220TB for the weekend. Also, don't forget to consider RAM; searching 10 128MB files with no limits on analysis could use 1.3GB.
In this short video clip, Managing the Network Data Collected in a Distributed Environment, Jay Botelho, Director of Product Management, walks you through starting points for your buffer size, file size, and more. (To see how to apply these suggestions to your OmniEngine or Omnipliance, read this Product Tip.) If you're growing your network analysis solution, here are some resources to help you choose your hardware:
Tip 3: View only the ports you need.
GACs can aggregate up to 4 data streams. If you only need to see traffic from 1 to 2 ports, use channel filters during captures from highly utilized network segments.