Router Support for Fine-Grained Latency Measurements – projects 2012
Projects 2012 Java Networking
Technology Used: Java
An increasing number of datacenter network applications, including automated trading and high-performance computing, have stringent end-to-end latency requirements where even microsecond variations may be intolerable. The resulting fine-grained measurement demands cannot be met effectively by existing technologies, such as SNMP, NetFlow, or active probing. Instrumenting routers with a hash-based primitive has been proposed that called as Lossy Difference Aggregator (LDA) to measure latencies down to tens of microseconds even in the presence of packet loss. Because LDA does not modify or encapsulate the packet, it can be deployed incrementally without changes along the forwarding path. When compared to Poisson-spaced active probing with similar overheads, LDA mechanism delivers orders of magnitude smaller relative error. Although ubiquitous deployment is ultimately desired, it may be hard to achieve in the shorter term.
To measure the latency accurately, to propose a method, LDA (Lossy Difference Aggregator), which accurately measures loss and delay over short timescales while providing strong bounds.
vÂ Â Â Â Current routers typically support two distinct accounting mechanisms: SNMP and NetFlow.
vÂ Â Â Â Operators of latency-critical networks are forced to use external monitoring mechanisms in order to collect a sufficient number of samples to compute accurate estimates. The simplest technique is to send end-to-end probes across the network.
vÂ Â Â Â Latency estimates computed in this fashion, however, can be grossly inaccurate in practice.
vÂ Â Â Â Unfortunately, placing hardware monitors between every pair of input and output ports is cost-prohibitive in many instances.
vÂ Â Â Â SNMP and NetFlow are not up to the task.
vÂ Â Â Â SNMP provides only cumulative counters that, while useful to estimate load, cannot provide latency estimates.
vÂ Â Â Â NetFlow, on the other hand, samples and timestamps a subset of all received packets; calculating latency requires coordinating samples at multiple routers (e.g., trajectory sampling).
vÂ Â Â Â In NetFlow, Samples and their timestamps have to be communicated to a measurement processor that subtracts the sent timestamp from the receive timestamp of each successfully delivered packet in order to estimate the average, a procedure with fundamentally high space complexity.
vÂ Â Â Â High NetFlow sampling rates significantly impact routersâ€™ forwarding performance and are frequently incompatible with operational throughput demands.
vÂ Â Â Â Lossy Difference Aggregator (LDA), fine-grain latency and loss measurement that can be cheaply incorporated within routers.
vÂ Â Â Â LDA accurately measures loss and delay over short timescales while providing strong bounds on its estimates, enabling operators to detect short-term deviations from long-term means within arbitrary confidence levels. Active probing requires 50â€“60 times as much bandwidth to deliver similar levels of accuracy.
vÂ Â Â Â Operators can use a classifier to configure an LDA to measure the delay of particular traffic classes to differing levels of precision, independent of others.
vÂ Â Â Â A low-overhead mechanism
vÂ Â Â Â Fine-granularity measurement