Router Support for
Fine-Grained Latency Measurements
ABSTRACT:
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. We propose instrumenting routers with a
hash-based primitive that we call a 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, our LDA mechanism
delivers orders of magnitude smaller relative error; active probing requires 50–60
times as much bandwidth to deliver similar levels of accuracy. Although
ubiquitous deployment is ultimately desired, it may be hard to achieve in the
shorter term; we discuss a partial deployment architecture called mPlane using
LDAs for intra router measurements and localized segment measurements for inter
router measurements.
EXISTING SYSTEM:
AN
INCREASING number of datacenter-based applications require end-to-end latencies
on the order of milliseconds or even microseconds. Moreover, many of them further
demand that latency remain stable, i.e., low jitter, for optimal performance.
These applications range from storage-area networks (SANs) to interactive Web
services that depend on large numbers of back-end services to niche—but
commercially important—markets like automated trading and high-performance computing.
Currently, most of these latency-sensitive applications are deployed on
specialized and often boutique hardware technologies like InfiniBand and
FibreChannel. Technology advances and market pressures, however, are leading to
increasing use of hybrid technologies like FibreChannel. over Etherent (FCoE)
that have the potential for increased performance crosstalk and complex partial
failure modes. Vendors and operators are under increasing pressure to be able to
provision and manage converged datacenter networks that meet these stringent
specifications. Unfortunately, most of the currently available tools are unable
to accurately measure latencies of these magnitudes, nor can they detect or
localize transient delay variations or loss spikes. Hence, we propose a new
mechanism to measure latency and loss at extremely small timescales, even tens
of microseconds.
DISADVANTAGES OF
EXISTING SYSTEM:
Moreover, computing accurate time
averages requires a high sampling rate, and detecting short-term deviations
from the mean requires even more. Unfortunately, high NetFlow sampling rates
significantly impact routers’ forwarding performance and are frequently
incompatible with operational throughput demands.
PROPOSED SYSTEM:
We propose the Lossy Difference
Aggregator (LDA), a low-overhead mechanism for fine-grain latency and loss
measurement that can be cheaply incorporated within routers to achieve the same
effect.
LDA forms the key building block of a
network-wide architecture we propose for collecting fine-grained latency
measurements called MPLANE. By using LDA for intra-router measurements and
active probes for link measurements, MPLANE enables fine-grained latency
measurements across the entire network for network operators to localize any
end-to-end latency spikes. We also propose an incremental deployment strategy
so that parts of the network can be upgraded in a more gradual fashion. Without
requiring all routers to be upgraded at the same time.
ADVANTAGES OF PROPOSED
SYSTEM:
LDA has the following features.
• Fine-granularity measurement: 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 50–60 times as much bandwidth to deliver similar levels of accuracy.
• Low overhead: Our suggested
40-Gb/s LDA implementation uses less than 1% of a standard networking ASIC and 72
kb of control traffic per second.
• Customizability: 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
MODULES:
·
Coordinated
Streaming
·
Internal Measurements
·
Segmented
Measurement
·
LDA
MODULES DESCRIPTION:
Coordinated Streaming
We measure the
goodness of a measurement scheme by its accuracy for each metric (in terms of
relative error), its storage overhead, bandwidth requirements, and its
computational overhead. A naive solution to the measurement problem is for the
sender to store a hash and timestamp of each sent packet and for the receiver
to do the same for each received packet. At the end of the interval, the sender
sends the hashes and timestamps for all packets to the receiver, who then
matches the send and receive timestamps of successfully received packets using
the packet hashes and computes the average.
Internal
Measurements
In many real
routers, forwarding metrics (e.g., loss, delay) depend on the forwarding class more
than the particular flow. For example, all flows traveling between the same
input and output ports of a router in a given QoS class are often treated
identically in terms of queuing and switch scheduling. Thus, we group such
flows into what we call a measurement equivalence class (MEC). We propose the use
of a scalable measurement primitive that can report latency measurements per
MEC within the router.
Segmented Measurement
The majority of operators
today employ active measurement techniques that inject synthetic probe traffic
into their network to measure loss and latency on an end-to-end basis. While
these tools are based on sound statistical foundations, active measurement
approaches are inherently intrusive and can incur substantial bandwidth
overhead when tuned to collect accurate fine-grained measurements. Rather than
conduct end-to-end measurements and then attempt to use tomography or inference
techniques to isolate the latency of individual segments, we propose to
instrument each segment of the network with our new measurement primitive.
Thus, in our model, every end-to-end path can be broken up into what we call
measurement segments.
LDA
A Lossy Difference Aggregator is a
measurement data structure that supports efficiently measuring the average
delay and standard deviation of delay. Both sender and receiver maintain an
LDA; at the end of a measurement period—in our experiments, we consider 1 s—the
sender sends its LDA to the receiver and the receiver computes the desired
statistics. The only additional requirements are tight time synchronization
between sender and receiver, a requirement shared by all one-way delay measurement
mechanisms, and consistent packet ordering (i.e., no packet reordering) at the
sender and receiver. To better explain the LDA, we begin with the simplest average
delay measurement primitive—a pair of counters—and then develop the full LDA.
HARDWARE REQUIREMENTS
•
SYSTEM : Pentium IV 2.4 GHz
•
HARD
DISK : 40 GB
•
MONITOR : 15 VGA colour
•
MOUSE : Logitech.
•
RAM : 256 MB
•
KEYBOARD :
110 keys enhanced.
SOFTWARE REQUIREMENTS
•
Operating system : Windows XP Professional
•
Front End : JAVA,
RMI, Swing
•
Tool : ECLIPSE
REFERENCE:
Ramana Rao Kompella, Kirill
Levchenko, Alex C. Snoeren, and George Varghese, “Router Support
for Fine-Grained Latency Measurements”, IEEE/ACM
TRANSACTIONS ON NETWORKING, VOL. 20, NO. 3, JUNE 2012.
**************************
Router Support for Fine-Grained Latency Measurements, Router Support for Fine-Grained Latency Measurements base paper, Router Support for Fine-Grained Latency Measurements ppt, Router Support for Fine-Grained Latency Measurements source code, Router Support for Fine-Grained Latency Measurements coding, Router Support for Fine-Grained Latency Measurements code, Router Support for Fine-Grained Latency Measurements 2012 ieee project, Router Support for Fine-Grained Latency Measurements for sale, Router Support for Fine-Grained Latency Measurements video file, Router Support for Fine-Grained Latency Measurements demo project, Router Support for Fine-Grained Latency Measurements document, Router Support for Fine-Grained Latency Measurements details