Preface
This period of time has been studying how to improve the utilization of flow tablespace. I have never been able to think of good idea. One article compares the methods mentioned in the existing research, and here records the types of methods and their shortcomings. These methods are not limitedHow to improve the utilization of flow tablespaceTo extend the scopeHow to improve the matching success rate of the vswitch flow table.
Background
Software Defined Network (SDN), as a new architecture, separates the control plane from the data plane using the concept of hierarchy, it provides great flexibility and scalability for network deployment and configuration.
However, the current SDN network can only ~ The information on the L4 layer is identified. When the Controller sets the timeout value of the stream table item, it treats all data streams equally, resulting in only one matching success rate for the table in the switch.Under 20%, Greatly increasing the burden on the controller and reducing the overall performance of the network.
Solution
- Increase the capacity of the stream table
- Ideas:You can add cache in the vswitch to storeHard_timeoutFrequently matched stream table items deleted when the time arrives to improve the matching rate of the stream table.
- Disadvantages:Essentially, it only increases the capacity of the stream table, which obviously cannot keep up with the speed of SDN development. Meanwhile, the matching rate of the stream table items in the cache is much worse than that of TCAM.
- Related Literature:Flow table management scheme applying an LRU caching algorithm, ICTC, 2014
- Add a cache module to the Controller.
- Ideas:The added cache module records the expiration time of the streaming table item, and uses the difference between the time when the streaming table item is re-installed and the expiration time as the basis for updating the streaming table item.
- Disadvantages:This method is applicable becauseIdle_timeoutFrequent updating of stream tables due to timeout has the disadvantages of increasing or decreasing only.
- Related Literature:Intelligent timeout master: Dynamic timeout for SDN-based data centers, ifip/IEEE, 2015
- Prediction of useless stream table items
- Ideas:PassPrediction MethodIn advance, the table items that will not be matched in the recent period of time are deleted, and the stream tablespace is reserved for the coming data stream.
- Disadvantages:This method will inevitably lead to misjudgment, which may lead to the deletion of useful stream table items, but the useless method occupies the flow tablespace.HoweverIf we can improve the prediction mechanism, this method is still an ideal idea.
- Related Literature:Flowmaster: Early eviction of deactive flow on SDN switche, the 15th International Conference on distributed computing and networking, 2014
A dynamic timeout control algorithmin Software Defined networks, ijfcc, 2014
- The Workflow table works like a queuing system.
- Ideas:The stream table model is converted into a state-dependent queuing model, and the queuing theory is used to quantitatively analyzeHard_timeoutInfluence of the number of convection truncation and blocking probability.
- Disadvantages:Based onHard_timeoutThe timeout mechanism model has poor flexibility. For packets with uneven data transmission time in the network, it cannot meet the transmission requirements well.
- Related Literature:Ahtm: achieving efficient flow table utilization in Software Defined networks, the 2014 IEEE Global Communications Conference, 2014
- Consider the processing capability of the Controller
- Ideas:The methods mentioned above do not use the Controller's processing capability as a dynamic modification timeout.
Value constraints. Which can be processed by the Controller Unit
The number of packet-in requests is used as the constraint to satisfy the minimum condition.Idle_timeoutValue as the final result to meet the processing capability of the controller.
To maximize the matching rate of a stream table.
- Disadvantages:This process assumes that all data streams have the same distribution parameter and does not consider the feature differences between different data stream types.
- Related Literature:Impulsive idle_timeout value for instant messaging in Software Defined network, the 2015 IEEE International Conference on communication workshop, 2015
- Added data stream Sensing Technology in SDN Networks
- This technology is hard to understand and needs further reading.
- Related Literature:SDN-based application-aware networking on the example of YouTube video, 2013 Second European workshop on Software Defined networks, 2013
Application-awareness in SDN, the ACM sigcomm 2013 conference on sigcomm, 2013
- The Workflow table works like a queuing system.
- Ideas:The stream table model is converted into a state-dependent queuing model, and the queuing theory is used to quantitatively analyzeHard_timeoutInfluence of the number of convection truncation and blocking probability.
- Disadvantages:Based onHard_timeoutThe timeout mechanism model has poor flexibility. For packets with uneven data transmission time in the network, it cannot meet the transmission requirements well.
- Related Literature:Ahtm: achieving efficient flow table utilization in Software Defined networks, the 2014 IEEE Global Communications Conference, 2014
- Comprehensive utilization of data stream type information and network resource usage
- Ideas:
In L2 ~ Adds data stream feature information differentiation based on L4 Network InformationAudio streams, video streams, and plain text streamsSuch as the data stream type, so as to set different timeout values for the stream data packet table items of different stream types, to maximize the use of stream table resources while meeting the processing capabilities of the controller, this improves the matching rate of the stream table.
- Disadvantages:Only data streams are divided into three categories, without further division of data streams, so as to implement more fine-grained control over the update of streaming table items.
- Related Literature:Intelligent update method for flow table in switch through analyzing data flow characteristics, ijca, 2016
Summary
- Currently, the documents related to stream table update are mainly divided into two types:
- Search for the matching rule of the stream table to dynamically update the timeout value, thus improving the matching rate of the stream table.
- Remove invalid stream table items in advance. Whether using the prediction method or the data packet itself, you can clearly determine that the stream is about to close the connection.
- The stream table is updated based on the data stream type information and network resource usage.
How to improve the matching success rate of vswitch flow tables in openflow Protocol
