Changes between Version 25 and Version 26 of NewDelayNodeNs
- Timestamp:
- Dec 14, 2011 10:26:05 AM (13 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
NewDelayNodeNs
v25 v26 1 1 = New Delay Node Commands = 2 2 == Introduction == 3 To simulate real networks Emulab offers a choice of injecting delays and dropping packets. This is done by inserting intercepting nodes, running dummynet, that delay packets and induce some artificial losses. DETER inherited this feature from [http://www.emulab.com Emulab], however, real network's behaviour is far different from static loss and delays. This new delay node tries to emulate real delays that fits into some well-know probability distribution. Also, it allows the user to configure a more sophisticated loss model. This new delay node is similar to dummynet by the mean of intercepting connections. These nodes runs Click modular routers, with some new modules developed in DETER. Below is the description of the command that sets up networks and configure nodes with the new features. The new command syntax is similar to the old make-lan to make an easy transition. However, it doesn't require the user to memorise the position of the parameters. 3 4 [http://www.isi.deterlab/net DETER] extends the [http://www.emulab.net Emulab] model of links and LANs with static delay and loss properties to include probabalistic losses and delays. This allows researchers to model large networks with fewer resources. Researchers can change the distributions on the fly in similar ways to changing static delay and loss parameters. The facility is based on the Click modular router, though this is transparent to researchers. 5 6 Delays can be distributed Normally, Exponentially, or using a Poisson distribution. Losses follow a two-state model with threshold poisson probabilities for entering and leaving the burst dropping state. The two-state loss model is described in more detail below. Delays can also be static and losses uniform to replicate the existing behavior. Similarly, the new facility will limit bandwidths. 7 8 Researchers can make use of this facility using the commands below. The new command syntax is similar to the old make-lan to make an easy transition. However, it doesn't require the user to memorise the position of the parameters. 4 9 5 10 == make-deter-lan == … … 51 56 52 57 }}} 53 Delay types are : static, normal , poisson and exponential. Note that normal requires standard deviation, which is set to zero if not provided.54 loss dropmodes are static and poisson. Also notice that the default bandwidth is 1Gb.58 Delay types are : '''static''', '''normal''', '''poisson''' and '''exponential'''. Note that normal requires standard deviation, which is set to zero if not provided. 59 Loss dropmodes are '''static''' and '''poisson'''. Also notice that the default bandwidth is 1Gb. 55 60 56 61 By choosing a static delay with 100ms, the LAN would delay each packet by 100ms. Normal delays packets based on a normal distribution fashion, with a mean and a standard variation. Plotting the travel time of the packet you would get a bell-curved graph. Poisson delays packet based on a discrete poisson distribution with Lambda=mean, and so does the exponential. 57 The loss model is more sophisticated.It adds the ability to drop packets based on different parameters. The user can adjust the parameters to achieve any desirable pattern. Dropping packet is based on a threshold x; x ≥ 0.000001 and x ≤ 1 which is given by the user. For each received packet, if not in the dropping state, the element generates a uniform random number r ∈ [0.000001, 1] if r ≤ x then it enters the dropping mode. It has to drop σ packets to leave the dropping state. σ is either a static value given by the user or a random number belongs to a Poisson distribution. To set up a dropping mode with static dropping rate the dropmode should be set to "static". 62 63 === Loss Model === 64 65 The loss model is more sophisticated.It adds the ability to drop packets based on different parameters. The user can adjust the parameters to achieve any desirable pattern. Dropping packet is based on a threshold x; x ≥ 0.000001 and x ≤ 1 which is given by the user. For each received packet, if not in the dropping state, the element generates a uniform random number r ∈ [0.000001, 1] if r ≤ x then it enters the dropping mode. It has to drop σ packets to leave the dropping state. σ is either a static value given by the user or a random number belongs to a Poisson distribution. To set up a dropping mode with static dropping rate the dropmode should be set to '''static'''. 58 66 59 67 == Examples == … … 292 300 293 301 }}} 294 295 296 This is a placeholder