GDSTR

Had an extremly productive meeting in Brad's office with the Heterogenous Experimental Network HEN lead developers Adam Greenhalgh & Felipe Huici.

Our objectives:

• Integrating our Moat testbed into the HEN infrastructure
• Find out what has or hasn't been done in HEN development and so decide on what tools to develop.

Currently Hardware status

2 Soekris boxes & 2 USB hubs with approximatly 30 Moats available.

Adam:  

Both Soekris boxes have been configured to run on HEN with USB and Power connections.  Prgrogramming Moats and Soekris boxes done via serial ports.  Adam has sent an email detailing how to connect to Soekris boxes via Arkell (Gatebox that allows ssh access to HEN infrastructure).

No Management or reservation software available for HEN in general or WSN,  Although research conducted on existing tools developed by Ramesh Govindan at USC, HEN team thought it wouldn't be feasible.  Brad will research into current USC testbed offering and find out what is on offer.

HEN uses XML for node reservations

Interface – multiple users – Suggest a login/reservation system:

Darren: Issues may be apparent if experiements are run by multiple personel.  Issues will be present in terms of:

1. Reservation system for Multiple users
2. Addressing individual nodes
3. Overlapping of Ranges 

1. Reservation system for Multiple Users

A Look into pre built management and reservation systems is required.  Was highlighted that, even though in a small scale use, with only a few users, reservations wouldn't be a problem, with a larger deployment and as the testbed develops reservations and manipulation of system would be an issue.

Theres a need to know, whats running, have images available and track experiements.

ToDo:

• Make a list of fine grained requirements for a reservation system
• Look at current tools available or develop own tools.

2. Addressing individual nodes

Nodes will be deployed in static deployment.  We need to uniquly address each moat.  Various techniques discussed:

Soekris Boxes have their own unique addresses – Adam has a script that queries a box – (Soekris or other) – and finds out what it is connected to.  Addressing moats could initially start with Soekris Address, then moat local ID - E.g. Soekris box 1, Moat 4. – Problems if box rebooted as id's would be reset.

Moats may have some form of MAC address that may be addressable need to look into this.  Deployment may also require development of a default moat image that can go on all moats.

3. Overlapping of Ranges 

On a testbed with many potential tests, overlapping radio ranges may be an issue.  Will need to look into frequencies and possibly setting tests on various frequency bands.

All the above points lead to the development of a Management "Middleware" level designed for easy querying of Moat information, detailing whats connected, whats available.

Theres a need for a static "store" (Database, XML storage),  that:

• Records Moat information
• Can allow for querying of Moat
• Displays information on Moat and its position / usage
• allows commands to be invoked

Phases:

Development of the Wireless Sensor Network Testbed components can be seperated into the following areas:

Phase 1: Push to Moat

• Push images to Moat
• Find Moat by Unique ID
• Programatically figure which Moat is connected to which node a form of "Moat DNS" – Use MAC address for addressing.

Phase 2: Detect and Write Moat Configuration

• Run Commands on Soekris boxes
• Develop auto detection functionality

Phase 3: Reservations

• Record length of time, experiment types, moats in use, frequency settings

Look and feel: better for look/feel to be different between HEN & WSN.

Brad: Orders been placed for the following:

• 6 x Soekris boxes
• 10 x USB Hubs
• Various lengths of USB cable
• 4 Pin cable for motherboard

ToDO:

• Brad – looks at current USC code and contacts Ramesh to find out what is available or could be reused.
• Read Vera's report – decide on power levels for network to be connected and communication between moats
• Darren: Aquires guide on how to set different channels & find out whether channels overlap.
• Check operation of Moat List and other Moat tools to see what is available for unique node addressing.

Had an productive meeting with UCL’s Moat experts, Vera Cady & Vladimir Dyo today in our regular 2pm meeting with Brad.

Key points from our chat were:

Root privilages are better for installation purposes, though not ideal.

Vera: Tmote’s have 15 power levels… 1 gives a range of 30cm’s, 15 give a range of the whole 7th floor.

Clock drift is apparent when multiple power over ethernet.

Brad: We have on hand in the department:

• Approximatly 30 Moats
• 2 USB Hubs
• 2 Soekris Boxes

Orders for more hardware have been placed and will be confirmed next week. USB Hub will arrive in days, Soekris boxes will take 1-2 weeks.

We’ll be using the D-Link 7 port DUB H7 hub…

Problem highlighted in experiments that when 4 – 7 motes in operation after a week they die

Soekris boxes that boot on HEN have no disk. Boxes need to be set up to netboot over PXE so there’s no need to install anything on the boxes.

As an overview the topology will be as follows:

• Soekris boxes will be placed in a region
• Moats will hang off Soekris boxes.
• Multiple hubs will have long run length – provide potential to daisy chain devices.

ToDo:

• Look at Chip Con for power level settings.
• Make Contact with HEN team – Adam / Felipe
• Develop remote login capabilities on Moats
• Find out the run length limit of USB cables
• Update & Include Topological diagram from presentation on blog

ToDo Aside:

• Tiny OS – Version 2.0 – Check features
• Contiki as opposed to Tiny OS
• Stargate
• Power Over Ethernet (POE)