Network Working Group B. Constantine
Internet-Draft JDSU
Intended status: Informational
Expires: April 19, 2010
Oct. 19, 2009
TCP Throughput Testing Methodology
draft-constantine-ippm-tcp-throughput-tm-00
Status of this Memo
This Internet-Draft is submitted to IETF in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as Internet-
Drafts. Creation date October 19, 2009.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
This Internet-Draft will expire on April 19, 2010.
Copyright Notice
Copyright (c) 2009 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents in effect on the date of
publication of this document (http://trustee.ietf.org/license-info).
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document.
Constantine Expires April 19, 2010 [Page 1]
Internet-Draft TCP Throughput Testing Methodology October 2009
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents in effect on the date of
publication of this document (http://trustee.ietf.org/license-info).
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document.
Abstract
This memo describes a methodology for measuring TCP throughput
performance in an end-end managed network environment. This memo is
intended to provide a practical approach to help users validate the
TCP layer performance of a managed network, which should provide a
better indication of end-user application level experience. In the
methodology, various TCP and network parameters are identified that
should be tested as part of the network verification at the TCP layer.
Requirements Language
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119 [RFC2119].
Constantine Expires April 19, 2010 [Page 2]
Internet-Draft TCP Throughput Testing Methodology October 2009
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4
2. Goals of this Methodology. . . . . . . . . . . . . . . . . . . 5
3. TCP Throughput Testing Methodology . . . . . . . . . . . . . . 6
3.1. Baseline Round-trip Delay and Bandwidth. . . . . . . . . . 6
3.1.1 Techniques to Measure Round Trip Delay. . . . . . . . 6
3.1.2 Techniques to Measure End-end Bandwidth . . . . . . . 7
3.2. Single TCP Connection Throughput Tests . . . . . . . . . . 7
3.3. Multiple TCP Connection Throughput Tests . . . . . . . . . 8
3.3.1 Multiple TCP Connections - below Link Capacity. . . . 9
3.3.2 Multiple TCP Connections - over Link Capacity . . . . 9
3.4. Varying MSS per Connection. . . . . . . . . . . . . . . . . 10
3.5. TCP Sessions with Stateless Background traffic. . . . . . . 10
3.5.1. TCP Foreground Traffic Control. . . . . . . . . . . . 11
3.5.2 Background Traffic Control. . . . . . . . . . . . . . 11
3.5.3. Test Methodology for TCP + Background Traffic . . . . 11
3.5.3.1. Prioritized Stateful TCP Traffic Test. . . . . 12
3.5.3.2. Prioritized Stateless Traffic Test . . . . . . 12
3.5.3.3. Other Traffic Test Cases . . . . . . . . . . . 13
4. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 13
5. References . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 14
Constantine Expires April 19, 2010 [Page 3]
Internet-Draft TCP Throughput Testing Methodology October 2009
1. Introduction
Even though RFC2544 was meant to benchmark network equipment and
used by network equipment manufacturers (NEMs), network providers
have used it to benchmark operational networks in order to
provide SLAs (Service Level Agreements) to their business customers.
Ultimately, network providers have come to the realization that a
successful RFC2544 test result does not guarantee end-user
satisfaction.
Therefore, the network provider community desires to measure network
throughput performance at the TCP layer. Measuring TCP throughput
will provide a much more meaningful measure that the network can meet
the end user's application SLA (and ultimately reach some level of
TCP testing interoperability which does not exist today).
The complexity of the network grows and the various queuing
mechanisms in the network greatly affect TCP layer performance (i.e.
improper default router settings for queuing, etc.) and devices such
as firewalls, proxies, load-balancers can actively alter the TCP
settings as a TCP session traverses the network (such as window size,
MSS, etc.). These are all very complex topics to the general network
community, and there is a strong interest to standardize a test
methodology at the TCP layer (especially from an end-end
perspective).
Before RFC2544 testing existed, network providers and NEMs deployed
a variety of ad hoc test techniques to verify the Layer 2/3
performance of the network. RFC2544 was a huge step forward in the
network test world, standardizing the Layer 2/3 test methodology
which greatly improved the quality of the network and reduced
operational test expenses. These managed networks are intended to be
predictable, but therein lies the problem. It is difficult if not
impossible, to extrapolate end user application layer performance
from RFC2544 results and the goal of RFC2544 was never intended
to do so.
So the intent behind this draft TCP throughput work is to define
a methodology for testing TCP layer performance, and guidelines for
expected TCP throughput results that should be observed in the
network under test. Network providers (and NEMs) are wrestling with
end-end complexities of the above (queuing, active proxy devices,
etc.); they desire to standardize the methodology to validate end-end
TCP performance, as this is the precursor to acceptable end-user
application performance.
Constantine Expires April 19, 2010 [Page 4]
Internet-Draft TCP Throughput Testing Methodology October 2009
2. Goals of this Methodology
Before defining the goals of this methodology, it is important to
clearly define the areas that are not intended to be measured or
analyzed by such a methodology. This is important to note, since
this methodology clearly does not intend to benchmark underlying
mechanisms within various flavors of TCP OS implementations:
- The methodology is not intended to definitively benchmark TCP
implementations of one OS to another (although some users may find
some value in conducting qualitative experiments)
- The methodology is not intended to provide detailed diagnosis
of problems within end-points or the network itself as related to
non-optimal TCP performance
- This methodology will not drill into the detailed inner behavior
of TCP implementations nor dissect the various timers and state
machines that can affect TCP performance
Concerning the goals of this methodology, it is clearly from the
perspective of a user that needs to conduct a structured, end-end
assessment of TCP performance within a managed business class IP
network. A key goal is that with the collective minds of this
working group, to establish a set of "best practices" that a user
should apply when validating the ability of a managed network to
carry end-user TCP applications. Some specific goals are to:
- Provide the logical, next-step testing methodology so that a
provider can test the network at Layer 4 (beyond the current Layer
2/3 RFC2544 testing approach)
- Provide a practical test approach that specifies the more well
understood (and end-user configurable) TCP parameters such as Window
size, MSS, # connections, and how these affect the outcome of TCP
performance over a network
- For networks that have been "tuned" with proper shaping and
queuing control mechanisms, it is desirable to define a TCP layer
test condition that can validate if the end-end network is tuned
as expected.
- Testing end-end prioritization of services is a key goal. This
draft proposes the use of stateful TCP connections in the midst of
stateless background traffic such as UDP, which is a very common
service condition in managed provider networks. The ability to test
stateful TCP with stateless traffic (with proper prioritization for
each), is a more thorough and realistic test than simply testing with
all stateless traffic. Further more, many networks will not tolerate
TCP "traffic blasting" to emulate the TCP application traffic.
Constantine Expires April 19, 2010 [Page 5]
Internet-Draft TCP Throughput Testing Methodology October 2009
3. TCP Throughput Testing Methodology
This section summarizes the specific test methodology to achieve the
goals listed in Section 2.
3.1. Baseline Round-trip Delay and Bandwidth
Before stateful TCP testing can begin, it is important to baseline
the round trip delay and bandwidth of the network to be tested.
These measurements provide estimates of the ideal TCP window size,
which will be used in subsequent test steps.
These latency and bandwidth tests should be run over a long enough
period of time to characterize the performance of the network over
the course of a meaningful time period. One example would
be to take samples during various times of the work day. The goal
would be to determine a representative minimum, average, and maximum
RTD and bandwidth for the network under test. Topology changes are
to be avoided during this time of initial convergence (e.g. in
crossing BGP4 boundaries).
In some cases, baselining bandwidth may not be required, since a
network provider's end-end topology may be well enough defined.
3.1.1 Techniques to Measure Round Trip Delay
This is not meant to provide an exhaustive list, but summarizes some
of the more common ways to determine round trip delay (RTD) through
the network. The desired resolution of the measurement (i.e. msec
versus usec) may dictate whether the RTD measurement can be achieved
with standard tools such as ICMP ping techniques or whether
specialized test equipment would be required with high precision
timers. The attempt in this section is to list several techniques in
order of decreasing accuracy.
- Use test equipment on each end of the network, "looping" the
far-end tester so that a packet stream can be measured end-end. This
test equipment RTD measurement may be compatible with delay
measurement protocols specified in RFC5357.
- Conduct packet captures of TCP test applications using for example
"iperf" or FTP, etc. By running multiple experiments, the packet
captures can be studied to estimate RTD based upon the SYN -> SYN-ACK
handshakes within the TCP connection set-up.
- ICMP Pings may also be adequate to provide round trip delay
estimations. Some limitations of ICMP Ping are the msec resolution
and whether the network elements / end points respond to pings (or
block them).
Constantine Expires April 19, 2010 [Page 6]
Internet-Draft TCP Throughput Testing Methodology October 2009
3.1.2 Techniques to Measure End-end Bandwidth
There are many well established techniques available to provide
estimated measures of bandwidth over a network. This measurement
should be conducted in both directions of the network, especially for
access networks which are inherently asymmetrical. Some of the
asymmetric implications to TCP performance are documented in RFC-3449
and the results of this work will be further studied to determine
relevance to this draft.
The bandwidth measurement test must be run with stateless IP streams
(not stateful TCP) in order to determine the available bandwidth in
each direction. And this test should obviously be performed at
various intervals throughout a business day (or even across a week).
Ideally, the bandwidth test should produce a log output of the
bandwidth achieved across the test interval AND the round trip delay.
And during the actual TCP level performance measurements (Sections
3.2 - 3.5), the test tool should also be able to track round trip
delay of the TCP connection(s) during the test. This can provide
insight into the potential effects of congestive delay to the
throughput achieved for the TCP layer test.
3.2. Single TCP Connection Throughput Tests
With a reasonable representation of round trip delay and bandwidth
from section 3.1, a series of single connection TCP throughput tests
can be conducted to baseline the performance of the network against
expectations. The optimum TCP window size can be calculated from
the bandwidth delay product (BDP), which is:
BDP = RTD x Bandwidth
By dividing the BDP by 8, the "ideal" TCP window size is calculated.
An example would be a T3 link with 25 msec RTD. The BDP would equal
~1,125,000 bits and the ideal TCP window would equal ~140,000 bytes.
There are several TCP tools that are commonly used in the network
provider world and one of the most common is the "iperf" tool. With
this tool, hosts are installed at each end of the network segment;
one as client and the other as server. The TCP Window size of both
the client and the server can be set and the achieved throughput is
measured, either uni-directionally or bi-directionally. For higher
BDP situations in lossy networks (satellite links, etc.), TCP options
such as Selective Acknowledgment should be considered and also become
part of the window size / throughput characterization.
Constantine Expires April 19, 2010 [Page 7]
Internet-Draft TCP Throughput Testing Methodology October 2009
Ideally, the single connection TCP throughput test should be run over a
a long duration and results logged at the desired interval. The
test should record RTD and TCP retransmissions at each interval. The
combination of throughput, RTD, and retransmissions per interval
can provide valuable insight into the resulting TCP throughput
results.
One important aspect of this test relates to the capabilities of the
end test tools or hosts. Variation in host hardware performance can
cause the results to be erroneous. Underperforming host hardware may
not be able to transfer payload fast enough to the NIC card, and
create the unintended bottleneck to TCP throughput performance.
Host hardware performance must be well understood before conducting
this TCP single connection test and other tests in this section.
Dedicated test equipment may be required, especially for line rates
of GigE and 10 GigE.
At the end of this step, the user will document the theoretical BDP
and a set of Window size experiments with measured TCP throughput for
each TCP window size setting. If network conditions (RTD and
retransmissions) are determined to be acceptable during the test
interval and the TCP throughput is not, then this MAY point to active
devices within the network that are altering window size (or other)
TCP parameters.
3.3. Multiple TCP Connection Throughput Tests
After baselining the network under test with a single TCP connection
(Section 3.2), the notional capacity of the network has been
determined. The capacity measured in section 3.2 may be a capacity
range and it is reasonable that some level of tuning may have been
required (i.e. router shaping techniques employed, intermediary
proxy like devices tuned, etc.).
Single connection TCP testing is a useful first step to measure
expected versus actual TCP performance and as a means to diagnose
/ tune issues in the network and active elements. However, the
ultimate goal of this methodology is to more closely emulate customer
traffic, which will be many TCP connections over a network link.
This methodology inevitably seeks to provide the framework for
testing stateful TCP connections in concurrence with stateless
traffic streams, and this is described in Section 3.5.
Constantine Expires April 19, 2010 [Page 8]
Internet-Draft TCP Throughput Testing Methodology October 2009
3.3.1 Multiple TCP Connections - below Link Capacity
First, the ability of the network to carry multiple TCP connections
to full network capacity should be tested. Prioritization and QoS
settings are not considered during this step, since the network
capacity is not to be exceeded by the test traffic (section 3.3.2
covers the over capacity test case).
For this multiple connection TCP throughput test, the number of
connections will more than likely be limited by the test tool (host
vs. dedicated test equipment). As an example, for a GigE link with
1 msec RTD, the optimum TCP window would equal ~128 KBytes. So under
this condition, 8 concurrent connections with window size equal to
16KB would fill the GigE link. For 10G, 80 connections would be
required to accomplish the same.
Just as in section 3.2, the end host or test tool can not be the
processing bottleneck or the throughput measurements will not be
valid. The test tool must be benchmarked in ideal lab conditions to
verify it's ability to transfer stateful TCP traffic at the given
network line rate.
For this test step, it should be conducted over a reasonable test
duration and results should be logged per interval such as throughput
per connection, RTD, and retransmissions.
Since the network is not to be driven into over capacity (by nature
of the BDP allocated evenly to each connection), this test verifies
the ability of the network to carry multiple TCP connections up to
the link speed of the network.
3.3.2 Multiple TCP Connections - over Link Capacity
In this step, the network bandwidth is intentionally exceeded with
multiple TCP connections to test expected prioritization and queuing
within the network.
All conditions related to Section 3.3 set-up apply, especially the
ability of the test hosts to transfer stateful TCP traffic at network
line rates.
Using the same example from Section 3.2, a GigE link with 1 msec
RTD would require a window size of 128 KB to fill the link (with
one TCP connection). Assuming a 16KB window, 8 concurrent
connections would fill the GigE link capacity and values higher than
8 would over-subscribe the network capacity. The user would select
values to over-subscribe the network (i.e. possibly 10 15, 20, etc.)
to conduct experiments to verify proper prioritization and queuing
within the network.
Constantine Expires April 19, 2010 [Page 9]
Internet-Draft TCP Throughput Testing Methodology October 2009
Without any prioritization in the network, the over subscribed test
results could assist in the queuing studies. With proper queuing,
the bandwidth should be shared in a reasonable manner. The author
understands that the term "reasonable" is too wide open, and future
draft versions of this memo would attempt to quantify this sharing
in more tangible terms. It is known that if a network element
is not set for proper queuing (i.e. FIFO), then an oversubscribed
TCP connection test will generally show a very uneven distribution of
bandwidth.
With prioritization in the network, different TCP connections can be
assigned various QoS settings via the various mechanisms (i.e. per
VLAN, DSCP, etc.), and the higher priority connections must be
verified to achieve the expected throughput.
3.4. Varying MSS per Connection
This test step can be run either on a single TCP connection test
(Section 3.2) or a multiple TCP connection test (section 3.3).
By varying the MSS size of the TCP connection(s), the ability of the
network to sustain expected TCP throughput can be verified. This is
similar to frame and packet size techniques within RFC2-2544, which
aim to determine the ability of the routing/switching devices to
handle loads in term of packets/frames per second at various frame
and packet sizes. This test can also further characterize the
performance of a network in the presence of active TCP elements
(proxies, etc.), devices that fragment IP packets, and the actual
end hosts themselves (servers, etc.).
The single connection testing listed in Section 3.2 should be
repeated first, using the appropriate window size and collecting
throughput measurements per various MSS sizes. It would be reasonable
to run single TCP connection tests with MSS sizes of 24, 88, 216,
472, 984, and 1460 bytes. These tests should also be run over a
predetermined test interval and the throughput, retransmissions, and
RTD logged during the entire test interval.
3.5. TCP Sessions with Stateless Background Traffic
The ultimate intent of this methodology is to more accurately assess
the ability of the network to carry end user TCP application traffic
in the midst of stateless background traffic. The background traffic
may be of a lower priority than the TCP traffic (i.e. background is
best effort Internet), or the background traffic may be of higher
priority (i.e.UDP representing voice). The prioritization must be
configured properly throughout the network to allow either the TCP
foreground traffic or the stateless background traffic to receive
the expected priority.
Constantine Expires April 19, 2010 [Page 10]
Internet-Draft TCP Throughput Testing Methodology October 2009
For this test, each stateful TCP connection must be able to have a
unique prioritization. Depending upon the network prioritization
scheme (i.e. VLAN, DSCP, MPLS, etc.), the test system must allow for
unique identification of each connection. The same applies for the
stateless background streams. The individual background streams must
also be tagged or identified based upon the prioritization mechanism.
3.5.1. TCP Foreground Traffic Control
In addition to the prioritization of each individual TCP connection,
the desired bandwidth for each TCP connection must be configurable.
Depending upon the capabilities of the test system, this bandwidth
rate may be discretely controlled (or shaped) by the test system or
may also be controlled by limiting the window size of each
connection. The sophistication of the test system will dictate which
bandwidth control mechanism is used for the TCP connections. The
window based approach allows the TCP traffic to reach the maximum
achievable within the bandwidth capacity of the link (BDP), which
may be the intended test configuration. Each TCP foreground
connection should also have configurable MSS sizes as well.
3.5.2 Stateless Background Traffic Control
Each stateless background traffic stream must be configurable in
terms of the offered network bandwidth. The frame / packet sizes
of the background traffic streams should be individually
configurable. Ideally, the test system should allow for the
stateless background traffic to ramp up from a configurable starting
bandwidth to the final bandwidth setting (i.e. start at 10 Mbps,
incrementing by 5 Mbps, until reaching 50 Mbps). The time step of
the ramping function should also be programmable. Ramping of the
background traffic facilitates the study of the effect of stateless
background traffic on foreground TCP traffic.
3.5.3. Test Methodology for TCP + Stateless Background Traffic
Depending upon the prioritization within the provider's network,
there are many permutations of prioritization between TCP sessions,
background streams, and combinations between. This memo will
summarize two common use cases: 1) higher priority TCP stateful
traffic in the midst of best effort background traffic; 2) higher
priority stateless traffic (i.e. VoIP) in the midst of lower
priority TCP stateful traffic.
Constantine Expires April 19, 2010 [Page 11]
Internet-Draft TCP Throughput Testing Methodology October 2009
3.5.3.1. Prioritized Stateful TCP Traffic Test
In this test, the intent is to verify that a business class data
service (i.e. thin client, web-based application traffic) is given
proper priority in times of high utilization. For this test
case, the TCP connections are given priority via the prioritization
mechanism used in the network (VLAN, DSCP, etc..) and the stateless
background traffic is given lower priority (or best effort).
By one of the traffic control techniques listed in Section 3.5.1, the
stateful TCP connections are allocated bandwidth within the test
system and ideally the background traffic will perform a ramp traffic
function. The results of this test should show that the
prioritized stateful TCP traffic reaches the designated throughput
and is not disturbed when the best effort background traffic exceeds
the link capacity. The test results should be logged during the test
interval, with each TCP connection's throughput, retransmissions, and
RTD recorded (along with the background traffic levels).
3.5.3.2. Prioritized Stateless Traffic Test
The corollary to section 3.5.3.1 is the case where the stateless
traffic is higher priority than the stateful TCP traffic. This may
be the case where a network provider is offering VoIP services in
addition to regular IP data service. Even though the TCP traffic is
prioritized lower than the stateless traffic, it is important to
determine how the TCP traffic reacts in the presence of over
subscription (this can again point to non-optimized queuing
techniques in the network for the TCP traffic as discussed in
Section 3.3.2.).
With the TCP offered bandwidth set by one of the traffic control
mechanisms listed in Section 3.5.1., the higher priority stateless
traffic should be ramped up to exceed the link capacity. The results
of this test should show that the higher priority stateless traffic
achieves the designated bandwidth and that the TCP connection
bandwidth is reduced. By studying the logged TCP and stateless
traffic throughput over the test interval (and the retransmissions
+ RTD for the TCP traffic), the manner in which the TCP connections
shared the remaining bandwidth may provide insight into possible
queuing optimizations in the network.
Constantine Expires April 19, 2010 [Page 12]
Internet-Draft TCP Throughput Testing Methodology October 2009
3.5.3.3. Other Traffic Test Cases
Sections 3.5.3.2 and 3.5.3.3 lay out the basic foundation for
testing the prioritization effects of TCP traffic and background
stateless traffic. There are many hybrids that can also be
pertinent, dependant upon the network provider's offering. An
example would be strictly an IP service type test. In this case,
the network provider seeks to test various prioritizations of each
stateful TCP connection and verify that the higher priority TCP
connection(s) achieve the SLA bandwidth while the others do not.
Regardless of the prioritization profile of the TCP connections
and the background streams, the same test results should be
recorded across the entire test interval (as specified in Section
3.5.3.1 and 3.5.3.2)
4. Acknowledgements
The author would like to thank Gilles Forget, Mike Hamilton,
and Reinhard Schrage for technical review and contributions to this
draft-00 memo.
Also thanks to Matt Mathis and Matt Zekauskas for many good comments
through email exchange and for pointing me to great sources of
information pertaining to past works in the TCP capacity area.
5. References
[RFC2581] Allman, M., Paxson, V., Stevens W., "TCP Congestion
Control", RFC 2581, April 1999.
[RFC3148] Mathis M., Allman, M., "A Framework for Defining
Empirical Bulk Transfer Capacity Metrics", RFC 3148, July
2001.
[RFC2544] Bradner, S., McQuaid, J., "Benchmarking Methodology for
Network Interconnect Devices", RFC 2544, March 1999
[RFC3449] Balakrishnan, H., Padmanabhan, V. N., Fairhurst, G.,
Sooriyabandara, M., "TCP Performance Implications of
Network Path Asymmetry", RFC 3449, December 2002
[RFC5357] Hedayat, K., Krzanowski, R., Morton, A., Yum, K., Babiarz,
J., "A Two-Way Active Measurement Protocol (TWAMP)",
RFC 5357, October 2008
draft-ietf-ippm-btc-cap-00.txt Allman, M., "A Bulk Transfer Capacity
Methodology for Cooperating Hosts", August 2001
Constantine Expires April 19, 2010 [Page 13]
Internet-Draft TCP Throughput Testing Methodology October 2009
Author's Address
Barry Constantine
JDSU, Test and Measurement Division
One Milesone Center Court
Germantown, MD 20876-7100
USA
Phone: +1 240 404 2227
Email: barry.constantine@jdsu.com
Constantine Expires April 19, 2010 [Page 14]