Measuring control plane latency in SDN-enabled switches

Abstract

Timely interaction between an SDN controller and switches is crucial to many SDN applications-e.g., fast rerouting during link failure and fine-grained traffic engineering in data centers. However, it is not well understood how the control plane in SDN switches impacts these applications. To this end, we conduct a comprehensive measurement study using four types of production SDN switches. Our measurements show that control actions, such as rule installation, have surprisingly high latency, due to both software implementation inefficiencies and fundamental traits of switch hardware.

Integrated Optical Content Addressable Memories (CAM) and Optical Random Access Memories (RAM) for Ultra-Fast Address Look-Up Operations

“…The assignment of different wavelengths at the various T-CAM cells of an ML certainly adds an extra degree of freedom to system-designs for more efficient architectures, especially when considering multiple simultaneous WDM memory schemes with parallel wavelengths written to the T-CAM cells of an ML for fast memory updates. This can be of utter importance with the emergence of SDN and OpenFlow architectures that enforce a dynamic network operation with frequent updates of network topologies and multiple real time changes in the RIB-list [25,26]. Although the SDN controllers are increasingly optimized for swift policy updates, the T-CAM tables remain yet unoptimized for fast updates [27,28], which may trigger hundreds to thousands of table entry-moves and write-memory operations [26].…”

“…This performance disparity has been placing an increasingly heavy load on the shoulders of electronic CAMs, enforcing energy-hungry, cost-expensive optoelectronic header conversions with subsequent data-rate down-conversion [22][23][24], in order to perform AL searches in the MHz-regime. Furthermore, the migration towards Software Defined Networks (SDN) and OpenFlow networks enforces a dynamic operation with frequent updates of network topologies and multiple real time changes in the RIB-list [25]. Internet routers are experiencing 100 updates per second and potentially reaching upwards of 1000 per second [26], during which AL operations are stalled and the router remains idle, considerably limiting the performance.…”

“…Internet routers are experiencing 100 updates per second and potentially reaching upwards of 1000 per second [26], during which AL operations are stalled and the router remains idle, considerably limiting the performance. This, in-turn, requires fast Write operations to the CAM-table, with short latencies [25][26][27][28], to enable fast re-configurability of the network and rapid updating of the AL table.…”

Software Defined Networks in Industrial Automation

“…In each run, we use a randomized distribution of inbound (α in ) and outbound (α ou ) flow latency, control channel latency (β), data channel latency (α p ) and packet processing time (γ) by a controller. The distribution of α in and α ou is fabricated from the outcome of a comprehensive measurement study [27,28] conducted using four types of production SDN switches. The distribution of inbound latency is a Chi-squared distribution attributed by a mean of 1.853 ms, a median of 0.71 ms and a standard deviation of 6.71 ms.…”

Designing and building SDN testbeds for energy-aware traffic engineering services

“…Moreover, existing work has already shown that updating the data path forwarding rules is slow in current commercial SDN switches [24]. Google employees report 5 that their SDN-based WAN had an outage due to this issue on propagation of forwarding rules.…”