Network Working Group B. Constantine, Ed.
Internet-Draft JDSU
Intended status: Informational G. Forget
Expires: September 7, 2010 Bell Canada (Ext. Consultant)
L. Jorgenson
Apparent Networks
Reinhard Schrage
Schrage Consulting
Mar 07, 2010
TCP Throughput Testing Methodology
draft-constantine-ippm-tcp-throughput-tm-02
Abstract
This memo describes a methodology for measuring sustained TCP
throughput performance in an end-to-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.
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 Mar 07, 2010.
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 September 7, 2010.
Constantine, et al. Expires September 7, 2010 [Page 1]
Internet-Draft TCP Throughput Testing Methodology March 2010
Copyright Notice
Copyright (c) 2010 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
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Goals of this Methodology. . . . . . . . . . . . . . . . . . . 4
2.1 TCP Equilibrium State Throughput . . . . . . . . . . . . . 5
3. TCP Throughput Testing Methodology . . . . . . . . . . . . . . 6
3.1. Baseline Round-trip Delay and Bandwidth. . . . . . . . . . 7
3.1.1 Techniques to Measure Round Trip Time . . . . . . . . 7
3.1.2 Techniques to Measure End-end Bandwidth . . . . . . . 8
3.2. Single TCP Connection Throughput Tests . . . . . . . . . . .9
3.2.1 Interpretation of the Single Connection TCP
Throughput Results . . . . . . . . . . . . . . . . . . 12
3.3. TCP MSS Throughput Testing . . . . . . . . . . . . . . . . 12
3.3.1 TCP Test for Network Path MTU . . . . . . . . . . . . 12
3.3.2 MSS Size Testing Method . . . . . . . . . . . . . . . 13
3.3.3 Interpretation of TCP MSS Throughput Results . . . . . 14
3.4. Multiple TCP Connection Throughput Tests. . . . . . . . . . 14
3.4.1 Multiple TCP Connections - below Link Capacity . . . . 14
3.4.2 Multiple TCP Connections - over Link Capacity. . . . . 15
3.5. TCP Sessions with Stateless Background traffic. . . . . . . 16
3.5.1. TCP Foreground Traffic Control. . . . . . . . . . . . 16
3.5.2 Stateless Backgound Traffic Control . . . . . . . . . 17
3.5.3. Test Methodology for TCP + Stateless Background
Traffic . . . . . . . . . . . . . . . . . . . . . . . 17
3.5.3.1. Prioritized Stateful TCP Traffic Test. . . . . 17
3.5.3.2. Prioritized Stateless Traffic Test . . . . . . 17
3.5.3.3. Other Traffic Test Cases . . . . . . . . . . . 18
4. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 18
5. References . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 19
Constantine, et al. Expires September 7, 2010 [Page 2]
Internet-Draft TCP Throughput Testing Methodology March 2010
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.
Network providers are coming to the realization that RFC2544 testing
and TCP layer testing are required to more adequately ensure end-user
satisfaction.
Therefore, the network provider community desires to measure network
throughput performance at the TCP layer. Measuring TCP throughput
provides a meaningful measure with respect to 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.). Network providers (and NEMs) are wrestling with end-end
complexities of the above and there is a strong interest in the
standardization of a test methodology to validate end-to-end TCP
performance (as this is the precursor to acceptable end-user
application performance).
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 sustained TCP layer performance. In this
document, sustained TCP throughput is that amount of data per unit
time that TCP transports during equilibrium (steady state), i.e.
after the initial slow start phase. We refer to this state as TCP
Equilibrium, and that the equalibrium throughput is the maximum
achievable for the TCP connection(s).
One other important note; the precursor to conducting the TCP tests
test methodlogy is to perform RFC2544 related Layer 2/3 tests. It
is highly recommended to run traditional RFC2544 type test to verify
the integrity of the network before conducting TCP testing.
Constantine et al. Expires September 7, 2010 [Page 3]
Internet-Draft TCP Throughput Testing Methodology March 2010
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.
- The methodology is not intended to predict TCP throughput
behavior during the transient stages of a TCP connection, such
as initial slow start.
- 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, although a results interpretation
section for each test step may provide insight into potential
issues within the network
In contrast to the above exclusions, the goals of this methodology
are to define a method to conduct a structured, end-to-end
assessment of sustained TCP performance within a managed business
class IP network. A key goal is to establish a set of "best
practices" that an engineer should apply when validating the
ability of a managed network to carry end-user TCP applications.
Some specific goals are to:
- 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
- Provide specific test conditions (link speed, RTD, window size,
etc.) and maximum achievable TCP throughput under TCP Equilbrium
conditions. For guideline purposes, provide examples of these test
conditions and the maximum achievable TCP throughput during the
equilbrium state. Section 2.1 provides specific details concerning
the definition of TCP Equilibrium within the context of this draft.
- In test situations where the recommended procedure does not yield
the maximum achievable TCP throughput result, this draft provides some
possible areas within the end host or network that should be
considered for investigation (although again, this draft is not
intended to provide a detailed diagnosis of these issues)
Constantine, et al. Expires September 7, 2010 [Page 4]
Internet-Draft TCP Throughput Testing Methodology March 2010
- 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
only stateless traffic. Further more, many networks will not tolerate
TCP "traffic blasting" to emulate the TCP application traffic.
2.1 TCP Equilibrium State Throughput
TCP connections have three (3) fundamental congestion window phases
as documented in RFC-TBD. These states are:
- Slow Start, which occurs during the beginning of a TCP transmission
or after a retransmission time out event
- Congestion avoidance, which is the phase during which TCP ramps up
to establish the maximum attainable throughput on an end-end network
path. Retransmissions are a natural by-product of the TCP congestion
avoidance algorithm as it seeks to achieve maximum throughput on
the network path.
- Retransmission phase, which include Fast Retransmit (Tahoe) and Fast
Recovery (Reno and New Reno). When a packet is lost, the Congestion
avoidance phase transitions to a Fast Retransmission or Recovery
Phase dependent upon the TCP implementation.
The following diagram depicts these states.
| ssthresh
TCP | |
Through- | | Equilibrium
put | |\ /\/\/\/\/\ Retransmit /\/\ ...
| | \ / | Time-out /
| | \ / | _______ _/
| Slow _/ |/ | / | Slow _/
| Start _/ Congestion |/ |Start_/ Congestion
| _/ Avoidance Loss | _/ Avoidance
| _/ Event | _/
| _/ |/
|/___________________________________________________________
Time
This TCP methodology provides guidelines to measure the equilibrium
throughput which refers to the maximum sustained rate obtained by
congestion avoidance before packet loss conditions occur (which would
cause the state change from congestion avoidance to a retransmission
phase). All maximum achievable throughputs specified in Section 3 are
with respect to this Equilibrium state.
Constantine, et al. Expires September 7, 2010 [Page 5]
Internet-Draft TCP Throughput Testing Methodology March 2010
3. TCP Throughput Testing Methodology
This section summarizes the specific test methodology to achieve the
goals listed in Section 2.
As stated in Section 1, it is considered best practice to verify
the integrity of the network from a Layer2/3 perspective by first
conducting RFC2544 type testing. If the network is not performing
properly in terms of packet loss, jitter, etc. when running RFC2544
tests, then the TCP layer testing will not be meaningful since the
equalibrium throughput would be very difficult to achieve (in a
"dysfunctional" network).
The following provides the sequential order of steps to conduct the
TCP throughput testing methodology:
1. Baseline Round-trip Delay and Bandwidth. These measurements provide
estimates of the ideal TCP window size, which will be used in
subsequent test steps.
2. Single TCP Connection Throughput Tests. With baseline measurements
of round trip delay and bandwidth, a series of single connection TCP
throughput tests can be conducted to baseline the performance of the
network against expectations.
3. TCP MSS Throughput Testing. By varying the MSS size of the TCP
connection, the ability of the network to sustain expected TCP
throughput can be verified.
4. Multiple TCP Connection Throughput Tests. Single connection TCP
testing is a useful first step to measure expected versus actual TCP
performance. The multiple connection test more closely emulates
customer traffic, which comprise many TCP connections over a network
link.
5. TCP Sessions with Stateless Background Traffic. This step assesses
the ability of the network to carry end user TCP application traffic
concurrent with stateless background traffic. The prioritization must
be configured properly throughout the network to allow either the TCP
traffic or the stateless traffic to receive the expected priority.
Important to note are some of the key characteristics and
considerations for the TCP test instrument. The test host may be a
standard computer or dedicated communications test instrument
and these TCP test hosts be capable of emulating both a client and a
server.
Whether the TCP test host is a standard computer or dedicated test
instrument, the following areas should be considered when selecting
a test host:
Constantine, et al. Expires September 7, 2010 [Page 6]
Internet-Draft TCP Throughput Testing Methodology March 2010
- TCP implementation used by the test host OS, i.e. Linux OS kernel
using TCP Reno, TCP options supported, etc. This will obviously be
more important when using custom test equipment where the TCP
implementation may be customized or tuned to run in higher
performance hardware
- Most importantly, the TCP test host must be capable of generating
and receiving stateful TCP test traffic at the full link speed of the
network under test. This requirement is very serious and may require
custom test equipment, especially on 1 GigE and 10 GigE networks.
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-to-end topology may be well enough defined.
3.1.1 Techniques to Measure Round Trip Time
We follow in the definitions used in the references of the appendix;
hence Round Trip Time (RTT) is the time elapsed between the clocking
in of the first bit of a payload packet to the receipt of the last
bit of the corresponding acknowledgement. Round Trip Delay (RTD)
is used synonymously to twice the Link Latency.
In any method used to baseline round trip delay between network
end-points, it is important to realize that network latency is the
sum of inherent network delay and congestion. The RTT should be
baselined during "off-peak" hours to obtain a reliable figure for
network latency (versus additional delay caused by congestion).
During the actual sustained TCP throughput tests, it is critical
to measure RTT along with measured TCP throughput. Congestive
effects can be isolated if RTT is concurrently measured.
Constantine, et al. Expires September 7, 2010 [Page 7]
Internet-Draft TCP Throughput Testing Methodology March 2010
This is not meant to provide an exhaustive list, but summarizes some
of the more common ways to determine round trip time (RTT) through
the network. The desired resolution of the measurement (i.e. msec
versus usec) may dictate whether the RTT 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 objective 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 RTT 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 RTT based upon the SYN -> SYN-ACK
handshakes within the TCP connection set-up.
- ICMP Pings may also be adequate to provide round trip time
estimations. Some limitations of ICMP Ping are the msec resolution
and whether the network elements / end points respond to pings (or
block them).
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 must be able to track round trip time
of the TCP connection(s) during the test. Measuring round trip time
variation (aka "jitter") provides insight into effects of congestive
delay on the sustained throughput achieved for the TCP layer test.
Constantine, et al. Expires September 7, 2010 [Page 8]
Internet-Draft TCP Throughput Testing Methodology March 2010
3.2. Single TCP Connection Throughput Tests
This draft specifically defines TCP throughput techniques to verify
sustained TCP performance in a managed business network. Defined
in section 2.1, the equalibrium throughput reflects the maximum
rate achieved by a TCP connection within the congestion avoidance
phase on a end-end network path. This section and others will define
the method to conduct these sustained throughput tests and guidelines
of the predicted results.
With baseline measurements of round trip time 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 = RTT 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 RTT. The BDP would equal
~1,105,000 bits and the ideal TCP window would equal ~138,000 bytes.
The following table provides some representative network link speeds,
latency, BDP, and associated "optimum" TCP window size. Sustained
TCP transfers should reach nearly 100% throughput, minus the overhead
of Layers 1-3 and the divisor of the MSS into the window.
For this single connection baseline test, the MSS size will effect
the achieved throughput (especially for smaller TCP window sizes).
Table 3.2 provides the achievable, equalibrium TCP
throughput (at Layer 4) using 1000 byte MSS. Also in this table,
the case of 58 byte L1-L4 overhead including the Ethernet CRC32 is
used for simplicity.
Table 3.2: Link Speed, RTT and calculated BDP, TCP Throughput
Link Ideal TCP Maximum Achievable
Speed* RTT (ms) BDP (bits) Window (kbytes) TCP Throughput(Mbps)
----------------------------------------------------------------------
T1 20 30,720 3.84 1.20
T1 50 76,800 9.60 1.44
T1 100 153,600 19.20 1.44
T3 10 442,100 55.26 41.60
T3 15 663,150 82.89 41.13
T3 25 1,105,250 138.16 41.92
T3(ATM) 10 407,040 50.88 32.44
T3(ATM) 15 610,560 76.32 32.44
T3(ATM) 25 1,017,600 127.20 32.44
100M 1 100,000 12.50 90.699
100M 2 200,000 25.00 92.815
Constantine, et al. Expires September 7, 2010 [Page 9]
Internet-Draft TCP Throughput Testing Methodology March 2010
Link Ideal TCP Maximum Achievable
Speed* RTT (ms) BDP (bits) Window (kbytes) TCP Throughput (Mbps)
----------------------------------------------------------------------
100M 5 500,000 62.50 90.699
1Gig 0.1 100,000 12.50 906.991
1Gig 0.5 500,000 62.50 906.991
1Gig 1 1,000,000 125.00 906.991
10Gig 0.05 500,000 62.50 9,069.912
10Gig 0.3 3,000,000 375.00 9,069.912
* Note that link speed is the minimum link speed throughput a network;
i.e. WAN with T1 link, etc.
Also, the following link speeds (available payload bandwidth) were
used for the WAN entries:
- T1 = 1.536 Mbits/sec (B8ZS line encoding facility)
- T3 = 44.21 Mbits/sec (C-Bit Framing)
- T3(ATM) = 36.86 Mbits/sec (C-Bit Framing & PLCP, 96000 Cells per
second)
The calculation method used in this document is a 3 step process :
1 - We determine what should be the optimal TCP Window size value
based on the optimal quantity of "in-flight" octets discovered by
the BDP calculation. We take into consideration that the TCP
Window size has to be an exact multiple value of the MSS.
2 - Then we calculate the achievable layer 2 throughput by multiplying
the value determined in step 1 with the MSS & (MSS + L2 + L3 + L4
Overheads) divided by the RTT.
3 - Finally, we multiply the calculated value of step 2 by the MSS
versus (MSS + L2 + L3 + L4 Overheads) ratio.
This gives us the achievable TCP Throughput value. Sometimes, the
maximum achievable throughput is limited by the maximum achievable
quantity of Ethernet Frames per second on the physical media. Then
this value is used in step 2 instead of the calculated one.
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 maunally set and the achieved
throughput is measured, either uni-directionally or bi-directionally.
For higher BDP situations in lossy networks (long fat networks or
satellite links, etc.), TCP options such as Selective Acknowledgment
should be considered and also become part of the window
size / throughput characterization.
Constantine, et al. Expires September 7, 2010 [Page 10]
Internet-Draft TCP Throughput Testing Methodology March 2010
The following diagram shows the achievable TCP throughput on a T3 with
the default Windows2000/XP TCP Window size of 17520 Bytes.
45|
|
40|
TCP |
Throughput 35|
in Mbps |
30|
|
25|
|
20|
|
15| _______ 14.48M
| | |
10| | | +-----+ 9.65M
| | | | | _______ 5.79M
5| | | | | | |
|_________+_____+_________+_____+________+____ +___________
10 15 25
RTT in milliseconds
The following diagram shows the achievable TCP throughput on a 25ms T3
when the TCP Window size is increased and with the RFC1323 TCP Window
scaling option.
45|
| +-----+42.47M
40| | |
TCP | | |
Throughput 35| | |
in Mbps | | |
30| | |
| | |
25| | |
| ______ 21.23M | |
20| | | | |
| | | | |
15| | | | |
| | | | |
10| +----+10.62M | | | |
| _______5.31M | | | | | |
5| | | | | | | | |
|__+_____+______+____+___________+____+________+_____+___
16 32 64 128
TCP Window size in Kili Bytes
Constantine, et al. Expires September 7, 2010 [Page 11]
Internet-Draft TCP Throughput Testing Methodology March 2010
The single connection TCP throughput test must be run over a
a long duration and results must be logged at the desired interval.
The test must record RTT and TCP retransmissions at each interval.
This correlation of retransmissions and RTT over the course of the
test will clearly identify which portions of the transfer reached
TCP Equilbrium state and to what effect increased RTT (congestive
effects) may have been the cause of reduced equilibrium 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.
3.2.1 Interpretation of the Single Connection TCP Throughput Results
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. For cases where the sustained TCP
throughput does not equal the predicted value, some possible causes
are listed:
- Network congestion causing packet loss
- Network congestion not causing packet loss, but effectively
increasing the size of the required TCP window during the transfer
- Network fragmentation at the IP layer
- Intermediate network devices which actively regenerate the TCP
connection and can alter window size, MSS, etc.
3.3. TCP MSS Throughput Testing
This test setup should be conducted as a single TCP connection test.
By varying the MSS size of the TCP connection, 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.).
3.3.1 TCP Test for Network Path MTU
TCP implementations should use Path MTU Discovery techniques (PMTUD),
but this technique does not always prove reliable in real world
situations. Since PMTUD relies on ICMP messages (to inform the host
that unfragmented transmission cannot occur), PMTUD is not always
reliable since many network managers completely disable ICMP.
Constantine, et al. Expires September 7, 2010 [Page 12]
Internet-Draft TCP Throughput Testing Methodology March 2010
Increasingly network providers and enterprises are instituting fixed
MTU sizes on the hosts to eliminate TCP fragmentation issues in the
application.
Packetization Layer Path MTU Discovery or PLPMTUD (RFC4821) should
be conducted to verify the minimum network path MTU. Conducting
the PLPMTUD test establishes the upper limit upon the MTU, which in
turn establishes the upper limit for the MSS testing of section 3.3.2.
MSS refers specifically to the payload size of the TCP packet and does
not include TCP or IP headers.
3.3.2 MSS Size Testing Method
The single connection testing listed in Section 3.2 should be
repeated, using the appropriate window size and collecting
throughput measurements per various MSS sizes.
The following are the typical sizes of MSS settings for various
link speeds:
- 256 bytes for very low speed links such as 9.6Kbps (per RFC1144).
- 536 bytes for low speed links (per RFC879) .
- 966 bytes for SLIP high speed (per RFC1055).
- 1380 bytes for IPSec VPN Tunnel testing
- 1452 bytes for PPPoE connectivity (per RFC2516)
- 1460 for Ethernet and Fast Ethernet (per RFC895).
- 8960 byte jumbo frames for GigE
Using the optimum window size determined by conducting steps 3.1 and
3.2, a variety of window sizes should be tested according to the link
speed under test. Using Fast Ethernet with 5 msec RTT as an example,
the optimum TCP window size would be 62.5 kbytes and the recommended
MSS for Fast Ethernet is 1460 bytes.
Link Achievable TCP Throughput (Mbps) for
Speed RTT(ms) MSS=1000 MSS=1260 MSS=1300 MSS=1380 MSS=1420 MSS=1460
----------------------------------------------------------------------
T1 20 | 1.20 1.008 1.040 1.104 1.136 1.168
T1 50 | 1.44 1.411 1.456 1.335 1.363 1.402
T1 100 | 1.44 1.512 1.456 1.435 1.477 1.402
T3 10 | 41.60 42.336 42.640 41.952 40.032 42.048
T3 15 | 42.13 42.336 42.293 42.688 42.411 42.048
T3 25 | 41.92 42.336 42.432 42.394 42.714 42.515
T3(ATM) 10 | 32.44 33.815 34.477 35.482 36.022 36.495
T3(ATM) 15 | 32.44 34.120 34.477 35.820 36.022 36.127
T3(ATM) 25 | 32.44 34.363 34.860 35.684 36.022 36.274
100M 1 | 90.699 89.093 91.970 86.866 89.424 91.982
100M 2 | 92.815 93.226 93.275 88.505 90.973 93.442
100M 5 | 90.699 92.481 92.697 88.245 90.844 93.442
Constantine, et al. Expires September 7, 2010 [Page 13]
Internet-Draft TCP Throughput Testing Methodology March 2010
For GigE and 10GigE, Jumbo frames (9000 bytes) are becoming more
common. The following table adds jumbo frames to the possible MSS
values.
Link Achievable TCP Throughput (Mbps) for
Speed RTT(ms) MSS=1260 MSS=1300 MSS=1380 MSS=1420 MSS=1460 MSS=8960
----------------------------------------------------------------------
1Gig 0.1 | 924.812 926.966 882.495 894.240 919.819 713.786
1Gig 0.5 | 924.812 926.966 930.922 932.743 934.467 856.543
1Gig 1.0 | 924.812 926.966 930.922 932.743 934.467 927.922
10Gig 0.05| 9248.125 9269.655 9309.218 9839.790 9344.671 8565.435
10Gig 0.3 | 9248.125 9269.655 9309.218 9839.790 9344.671 9755.079
Each row in the table is a separate test that should be conducted
over a predetermined test interval and the throughput, retransmissions,
and RTT logged during the entire test interval.
3.3.3 Interpretation of TCP MSS Throughput Results
For cases where the predicted TCP throughput does not equal the
predicted throughput predicted for a given MSS, some possible causes
are listed:
- TBD
3.4. Multiple TCP Connection Throughput Tests
After baselining the network under test with a single TCP connection
(Section 3.2), the nominal 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 comprise 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.
3.4.1 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
Constantine, et al. Expires September 7, 2010 [Page 14]
Internet-Draft TCP Throughput Testing Methodology March 2010
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 RTT, 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, RTT, 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.
3.4.2 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
RTT 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.
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
Constantine, et al. Expires September 7, 2010 [Page 15]
Internet-Draft TCP Throughput Testing Methodology March 2010
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.
3.5. 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
concurrent with 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.
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 size as well.
Constantine, et al. Expires September 7, 2010 [Page 16]
Internet-Draft TCP Throughput Testing Methodology March 2010
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.
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
RTT 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
Constantine, et al. Expires September 7, 2010 [Page 17]
Internet-Draft TCP Throughput Testing Methodology March 2010
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
+ RTT 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.
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.
Constantine, et al. Expires September 7, 2010 [Page 18]
Internet-Draft TCP Throughput Testing Methodology March 2010
[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
[RFC4821] Mathis, M., Heffner, J., "Packetization Layer Path MTU
Discovery", RFC 4821, March 2007
draft-ietf-ippm-btc-cap-00.txt Allman, M., "A Bulk
Transfer Capacity Methodology for Cooperating Hosts",
August 2001
[MSMO] The Macroscopic Behavior of the TCP Congestion Avoidance
Algorithm Mathis, M.,Semke, J, Mahdavi, J, Ott, T
July 1997 SIGCOMM Computer Communication Review,
Volume 27 Issue 3
[Stevens Vol1] TCP/IP Illustrated, Vol1, The Protocols
Addison-Wesley
Authors' Addresses
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
Gilles Forget
Independent Consultant to Bell Canada.
308, rue de Monaco, St-Eustache
Qc. CANADA, Postal Code : J7P-4T5
Phone: (514) 895-8212
gilles.forget@sympatico.ca
Loki Jorgenson
Apparent Networks
Phone: (604) 433-2333 ext 105
ljorgenson@apparentnetworks.com
Constantine, et al. Expires September 7, 2010 [Page 19]
Internet-Draft TCP Throughput Testing Methodology March 2010
Reinhard Schrage
Schrage Consulting
Phone: +49 (0) 5137 909540
reinhard@schrageconsult.com
Constantine, et al. Expires September 7, 2010 [Page 20]