1. Trang chủ
  2. » Công Nghệ Thông Tin

Tài liệu KRONE - White paper - TrueNet - Network Troubleshooting - 2001 pptx

8 285 0
Tài liệu đã được kiểm tra trùng lặp

Đang tải... (xem toàn văn)

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Tiêu đề Network Troubleshooting Using TrueNet Test Methodologies
Chuyên ngành Computer Networks
Thể loại White paper
Năm xuất bản 2001
Định dạng
Số trang 8
Dung lượng 250,49 KB

Các công cụ chuyển đổi và chỉnh sửa cho tài liệu này

Nội dung

Further, KRONE is the first and only structured connectivity manufacturer to ensure physical layer performance in the active domain through our exclusive post-installation active testing

Trang 1

KRONE: 800-775-KRONE www.kroneamericas.com

that the answer to network perform-ance problems lies

in top-down solu-tions: management software, more equipment, faster transmission speeds.

KRONE ® Incorporated suggests that you first look at the

foundation of your network — the structured cabling

system — and work your way up In fact, research

conducted in KRONE’s laboratories, and hundreds of

corporate site surveys conducted by KRONE technicians,

indicate that the physical layer can be a significant

source of difficult-to-diagnose network problems.

Field tests of corporate networks while they are running

have repeatedly verified that network signals can be

compromised by both poor cabling and out-of-specification

active hardware The result: corrupted Ethernet signals,

resulting in dropped frames, and poor application

performance The KRONE technical papers: “The Effect

of Errors on TCP Application Performance” and “Catching

up with TrueNet” detail what happens to networks, and

Transmission Control Protocol (TCP) performance in particular,

when errors are present in a network To summarize these

in-depth papers: no matter how well your network is

designed, or how much bandwidth you have, an error

is an error to TCP — and errors slow down applications

The extent to which your physical network is a cause of

errors has a tremendous amount of impact on how well

your network is capable of operating KRONE has found

that many corporate users have error-causing problems

hidden within their infrastructure, and they don’t even

know it KRONE has developed the TrueNet Structured

Cabling System as a solution to this problem

The TrueNet System

KRONE’s TrueNet Structured Cabling System is a complete

physical layer consisting of cabling and connectivity

components such as jacks, patch panels, termination

blocks, horizontal cable and patch cords The system was

designed and manufactured specifically to ensure optimum LAN transmission performance Further, KRONE is the first and only structured connectivity manufacturer to ensure physical layer performance in the active domain through our exclusive post-installation active testing program

The inherent advantages of post-installation active testing

We are often asked why KRONE places so much emphasis

on the active testing aspect of the TrueNet solution There are a number of answers, some obvious, and some less so

A quality check of the cabling system — This might

seem intuitive, but the quality check is not for the products themselves, it’s for the way they’re installed This takes on two forms: the way the product is installed

in your particular office space, with its own idiosyncrasies, and how well our certified installers performed the job

Your unique office environment — KRONE could

claim that we do active testing in the factory or the lab to ensure “data throughput,” but what relationship does our laboratory have to your office environment? The fact is, the myriad conditions that can be present in real facilities just can’t be duplicated in the lab What better test of how well it will work than testing what you actually have?

The KRONE Certified Installer’s responsibility —

KRONE Certified Installers are vigorously screened and trained to be the best in the industry We demand more from our installers than any other company — KRONE doesn’t certify a TrueNet warranty until we’re satisfied the job was done correctly We require our installers to send us passive test results for every node of your installation, and we check it to make sure it all passed We give our installers more than twenty additional installation requirements over and above the standards — installation requirements that ensure that data throughput won’t be compromised by excessive noise, over-tightened cable fasteners, incorrect punch-downs and so forth

MANY PEOPLE

B

BE EL LIIE EV VE E

Network Troubleshooting

Test Methodologies

Trang 2

Given that office environments vary, and installation practices have a definite effect on the performance of the cabling system, the active test process serves as a final check that the installation was done properly, and the office itself doesn’t present any throughput-degrading challenges (such as external RF noise)

A check of how your network actually runs —

Another obvious benefit of KRONE®active testing is that it’s an actual test of how well the as-configured network actually works Since we wait until all the active components are installed before we do our testing, we have the opportunity to evaluate real traffic on your actual network Traffic generators such as Smartbits™can simulate network activity and give us some information on the cabling system, but KRONE does not typically use traffic generators for the warranty verification for a couple of reasons First, for reasons explained in “Catching up with TrueNet™, ” devices such as Smartbits are not a good indicator of how the real network will perform, and second,

a traffic generator has its own network interface cards (NICs), so it isn’t really your network we would be testing

See what the active equipment is doing —

As you’ll see later, active equipment has a variety of transmission characteristics that can be faulty in almost any NIC or port Finding these faults is an added benefit

of the active testing, and often the one that our customers find the most useful Consider: a company moving offices from location A to B is almost certain to buy a new cabling system, but it’s very likely they will move all the computers, servers, hubs and switches with them If any of those “imported” pieces of equipment have poor transmitters in them — you’ve just imported a problem into your brand new cabling system! The beauty of the KRONE warranty test methodology is that we can find these problems for the customer in the normal course of performing the warranty check It’s essentially getting a

“health check” on your active hardware for free

Making sure you got what you paid for — The

suggestion that “any” Category 5e or 6 cabling system will perform flawlessly in the active domain so long as it meets the standards is simply not correct In the “Catching

up with TrueNet” paper, KRONE shows two Category 5e cabling systems, one TrueNet, the other “mix-and-match”

that have radically different error rates In fact, the

”mix-and-match“ solution dropped as many as 148 frames per

second, compared with the KRONE solution, which only

dropped one frame in 12 minutes.

What does KRONE’s bit error warranty entail —

The bit error aspect of the KRONE warranty has been the subject of many questions, so we’d like to clarify a point or two The nature of the warranty is to ensure that the cabling system will not cause bit errors that degrade performance; provided the active hardware in use is operating within the specifications defined by IEEE 802.3xx The operative terms here are “cabling system” and “degrade performance.” As shown in the KRONE white paper “Catching Up With TrueNet,” a TrueNet Category 5e system performs dramatically different from the alternative Category 5e system tested In that case, the TrueNet Cabling System only clocked one bit error in

nearly 12 minutes — and that error could have been due

to the switch or the NIC! The alternative cabling system

had four errors per second or worse, using the exact

same port and NIC The critical thing to understand

here is the time factor KRONE samples actual traffic

on the network and pinpoints possible problems with the active testing program In doing so, we ensure that operating bit error rates in the installed system are 10-10

(0.000000001%) or better As a function of elapsed time, this error rate is so low that the effect on TCP (and applications, by extension) is imperceptible to the user

TrueNet System active testing

Let’s take a look at some of the real-world networks that KRONE has encountered in site surveys First, however, it’s instructive to see what a well-matched system should look like In Figure 1, we see an

“impedance over distance” trace, which basically shows the electrical signature of the cabling channel over its distance The meter scale along the X-axis shows us this channel is approximately 81 meters long The electrical trace is smooth and all pairs line up on top of each other This is a good channel Figure 2 shows the same channel in the frequency (MegaHertz) domain This could never be a precisely smooth trace, but the lines should be as compact as possible, without any evidence of cyclical oscillation Figures 1 and 2 are a TrueNet cabling channel

Now, we’ll also look at some characteristics of an active Ethernet signal that also have a bearing on how well signals are transmitted They are the parameters that define the shape of the transmitted signal, and include rise time/fall time, amplitude and jitter

2

An obvious benefit of

KRONE active testing

is that it’s an actual

test of how well the

as-configured network

actually works.

Trang 3

KRONE: 800-775-KRONE www.kroneamericas.com

Figure 1 shows the impedance of a TrueNet structured cabling channel over the distance of the entire channel length All the components in the channel: patch cords, jacks, patch panels, interconnect, cross-connect, horizontal cable and so forth, appear on this trace, but it is nearly impossible to tell where the individual components are located All four cable pairs are impedance matched throughout the length of the channel

In the same channel, impedance is charted against frequency (in MegaHertz) to show that the impedance holds a tight band at 100±15 ohms for the critical frequency range of 1–125 MHz where the energy of Ethernet signals is concentrated (Figure 2)

Beyond the channel’s impedance characteristics, however, the following active test characteristics are important to proper network operation: network load, jitter, rise/fall time, minimum voltage/amplitude, interference, CRC errors, alignment fragments, interframe gap and oversized/undersized packets

Within this document, we’ll examine three active test characteristics: rise/fall time, amplitude and jitter

Rise and fall time — These characteristics are important

to system performance, as synchronization requires “change

of state” to occur within a specified time frame Even if the signal reaches proper amplitude — but too slowly — the signal is misinterpreted (Figure 3)

For example, TrueNet Warranty Certification Tests run at

a large, nationwide law firm showed that some sections

of their newly-installed network were running slowly The problem: the 100BASE-T port on a brand new switch was out of compliance for both rise time and fall time (Figures

4 and 5) Imagine if this had been the connection to a server!

3

Nanoseconds

Rise Time Fall Time

Rise and fall time is the length of time, in nanoseconds, that it takes for a signal to rise or fall from one state to another, signaling a “one.”

Figure 3: Rise and fall time.

7.8

6.0

4.3

2.5

0.8

IEEE LIMIT

Time (Nanoseconds)

GOOD WARNING ERROR

Figure 4: Rise time.

Distance (Meters)

40 50 60 70 80 90 100 110 120 130 140

KRONE IMPEDANCE MATCHED CHANNEL

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 1: A TrueNet matched impedance channel.

0 26 52 78 104 130 156 182 208 234 260

Frequency (MHz)

40 50 60 70 80 90 100 110 120 130 140

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 2: Frequency trace of a TrueNet channel.

7.8

6.0

4.3

2.5

0.8

IEEE LIMIT

Time (Nanoseconds)

GOOD WARNING ERROR

Figure 5: Fall time.

Rise and fall

time is the

length of time,

in nanoseconds,

that it takes for

a signal to rise

or fall from one

state to another,

signaling a “one.”

Trang 4

Amplitude — The amplitude of a wave (Figure 6) is

equivalent to the height of the physical wavelength — also signaling to the active device a “change in state”

corresponding to a one or zero If the amplitude is not within the specified range, errors can occur Using actual test data, we “see” an active device that is well within the specified limits (Figure 7, yellow lines) The NIC is barely within compliance, however, and over time may drift out of specification (Figure 8)

Jitter — In yet another installation, the 10BASE-T card

shows significant instability in meeting the IEEE jitter specification (Figure 9) Over time, this card may begin

to emit signals not within specifications

How are system errors found?

The previous information gives some idea of the type of errors that the TrueNet™warranty testing brings to light, but how are these errors actually found?

First, the entire network must be polled for errors There are various products that poll SNMP (Simple Network Management Protocol) reports for errors Alternatively, specific nodes that are known to be problems can

be polled directly By attaching a diagnostic device to problem nodes, KRONE®technicians are able to capture specific error events and filter them by type Experience shows that certain errors tend to be attributable to cabling problems, and others to poorly performing active equipment Of course, once error events are detected on a specific node, it is relatively simple to determine the health of the active devices If the active devices check out, then it is time to analyze the cabling traces for problems

Figure 10 shows a series of errors In this particular test situation, we are seeing a number of fragment errors, depicted on the report as “Jams” or “Runt” packets These packets are not complete and cannot

be interpreted by the receiving end At the extreme bottom of the chart, we see two errors that indicate

a wrong value in the destination prefix This means that the “addressee” is not known Again, the packet cannot be used if it’s not known where to send it

4

Amplitude signals

a “change in state”

to the active device.

Jitter relates to the

stability of the NIC.

Wavelength

Figure 6: Sine wave amplitude.

10.0

7.5

5.0

2.5

0.0

Time (Nanoseconds)

GOOD WARNING ERROR

Figure 7: This active device from an actual installation shows amplitude well within the specified limits (yellow lines).

4

3

2

1

0

Time (Nanoseconds)

GOOD WARNING ERROR

Figure 8: This NIC is barely within compliance, and overtime may drift out of specification.

11

0

-11

-22

Time (Nanoseconds)

GOOD WARNING ERROR

Figure 9: Significant instability in meeting the IEEE jitter is apparent in this 10BASE-T card.

Trang 5

KRONE: 800-775-KRONE www.kroneamericas.com

The exclusive TrueNet™test methodology extends to the active devices in use, evaluating if NICs, hubs and switches are performing within IEEE specifications The end result is

a LAN that is proven to transmit data efficiently, allowing network managers to concentrate on other concerns

NIC concerns

In site after site, we have noticed a tremendous variation

in the performance of NICs The rule of thumb we’ve found is: once a NIC becomes marginal, it never gets any better And some NICs are marginal right out of the box —

in fact, to some users it has become commonplace to view the NIC as an expendible resource Some companies are even known to keep dozens of “cheap” NICs around, replacing them constantly, at the first sign of trouble

We also have found that the probability of finding a marginal NIC has nothing to do with the brand or expense of the card Inexpensive and expensive brands alike are just as likely to fail; and a weak NIC (Figures 11, 12 and 13) has very little chance of successful transmission when the poorly transmitted

5

8.0

6.5

5.0

3.5

2.0

Time (Nanoseconds)

GOOD WARNING ERROR

Figure 11: Rise time of a weak NIC.

8.0

6.5

5.0

3.5

2.0

Time (Nanoseconds)

GOOD WARNING ERROR

Figure 12: Fall time of a weak NIC.

Figure 10: This screen shot shows a series of errors, depicted in the “Failure Type” column.

22

11

0

-11

-22

Time (Nanoseconds)

BORDER LINE JITTER

GOOD WARNING ERROR

Figure 13: Jitter characteristics of a weak NIC.

A weak NIC has

very little chance

of successful

transmission

when the signal

must also

navigate system

mismatches.

Trang 6

signal reaches the closet (and must then navigate the usual system mismatches in its weakened/attenuated state) This

is why we so often see 10/100 nodes auto-negotiate down

to 10 Mb/s (because they cannot maintain the transmission

at 100 Mb/s) Misguided MIS managers often attempt to turn off auto-negotiation to prevent this from occurring, the consequence being a virtual meltdown of the network

— nothing can get through at all

TruePatch Cords are a key component to the TrueNet™

Structured Cabling System They provide a vital system upgrade and solve a problem found in many networks — unmatched, low performance/low reliability patch cords

Research has shown that even one marginal patch cord can substantially slow down an entire network But the overall performance of the network can be significantly improved just by replacing the existing patch cords with TruePatch Patch Cords Figures 14, 15, 16 and 17 serve

as examples of how TruePatch Cords can advance the performance of a network These examples are not laboratory created simulations of theoretical problems,

they are based on real, active domain tests conducted

at various corporate sites

An electronic components manufacturer —

Tests showed that this manufacturer has good horizontal cable, but a poorly matched patch cord which has a much lower impedance than the horizontal cable (Figure 14) For such a seemingly minor part of the entire channel, that patch cord can cause big problems for data transmission Notice how the patch cord causes the impedance versus frequency chart to drop out on the low end of impedance (Figure 15) A TruePatch cord added to the channel is almost a perfect match (note the green pair still spikes in the jack — Figure 16) In this case, a TruePatch cord with the existing horizontal cabling yields a good trace for impedance versus frequency A TruePatch cord makes the difference between passing and failing (Figure 17)

A top US airline — This channel was one of the worst

we had ever seen The user on this node complained that his computer was excruciatingly slow when down-loading files off the network We suspected an enormous impedance mismatch and a wide difference in the impedance between pairs to be the culprit (Figure 18) The next chart shows this channel to be way out of

6

40 50 60 70 80 90 100 110 120 130

Distance (Meters)

IMPEDANCE USING A TRUEPATCH PATCH CORD

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 16: A TruePatch cord added to the channel corrects the problem.

0 26 52 78 104 130 156 182 208 234 260 40

50 60 70 80 90 100 110 120 130 140

Frequency (MHz)

10BASE-T

100BASE-T

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 17: The addition of a TruePatch cord can be the difference between meeting or failing the spec.

For such a seemingly

minor part of the

entire channel, a bad

patch cord can cause

big problems for data

transmission.

0 26 52 78 104 130 156 182 208 234 260 40

50 60 70 80 90 100 110 120 130 140

Frequency (MHz)

10BASE-T

100BASE-T

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 15: Patch cord causes big problems for data transmission.

40 50 60 70 80 90 100 110 120 130

Distance (Meters)

IMPEDANCE MISMATCH OF THE PATCH CORD, THE BLOCK AND THE CABLE

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 14: Good horizontal cable with a poorly matched patch cord.

Trang 7

KRONE: 800-775-KRONE www.kroneamericas.com

specification It took this user eight minutes to download

a large PowerPoint®file over this node (Figure 19) Simply replacing the patch cord at the computer cleans up the channel quite a bit; but will it bring the frequency chart into compliance (Figure 20)? As you can see, the frequency curve isn’t completely normal, but the patch cord made a big difference nonetheless The same file that took eight minutes

to download, now took only two minutes Of course, actual results on other networks will vary, but this is a compelling example of the value of good infrastructure (Figure 21)

A software and engineering firm — Sometimes the

TruePatch™Cord will smooth out network problems, but other system components may also need to be replaced Tests

at this firm showed that the patch cord was ill-matched to the horizontal cable (Figure 22) Note in particular how the green and brown pairs differ in impedance from each other The impedance spikes occurring at five meters into the run (viewing left to right) are the poorly matched interface between the plug on the patch cord and the jack Simply replacing the existing patch cord with a TruePatch cord yields a noticeable improvement in the trace Note, however, that since the jacks were not part of a matched system, the same spike occurs at the plug/jack interface (Figure 23)

7

40 60 80 100 120

Distance (Meters)

IMPEDANCE MISMATCH OF THE PATCH CORD, THE PANEL AND THE CABLE

IMPEDANCE MISMATCH OF THE CABLE CONDUCTORS

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 18: A channel with an impedance mismatch and mismatched cable pairs.

60 70 80 90 100 110 120 130 140 150 160

Frequency (MHz)

<<10BT 100BT

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 19: Viewed in the frequency domain, it is apparent that the channel is way out of specification.

40 60 80 100 120 140

Distance (Meters)

IMPEDANCE IMPROVEMENT USING A TRUEPATCH PATCH CORD

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 20: The same channel with a TruePatch cord.

50 60 70 80 90 100 110 120 130 140

Frequency (MHz)

<<10BT 100BT

IMPROVEMENT USING A TRUEPATCH PATCH CORD

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 21: The addition of TruePatch greatly improves channel performance.

40 60 80 100 120 140

Distance (Meters)

IMPEDANCE MISMATCH OF THE PATCH CORD, THE PANEL AND THE CABLE

IMPEDANCE MISMATCH OF THE CABLE CONNECTORS

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 22: Patch cord is ill-matched to horizontal cables.

40 50 60 70 80 90 100 110 120 130 140

Distance (Meters)

IMPEDANCE IMPROVEMENT USING A TRUEPATCH PATCH CORD

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 23: Replacement of patch cord yields a noticeable improvement.

Replacing a

patch cord with

a TruePatch cord

can smooth out

network problems.

Trang 8

KRONE ® Incorporated

North America Headquarters

6950 South Tucson Way

Englewood, CO 80112-3922

Telephone: (303) 790.2619

Toll-Free: (800) 775.KRONE

Facsimile: (303) 790.2117

www.kroneamericas.com

The original patch cord causes the channel’s impedance over frequency to drop below 85 ohms at lower than

100 MHz, meaning potentially serious data transmission problems (Figure 24) The TruePatch™cord in the same channel smooths out the frequency trace somewhat, actually bringing most of the trace into compliance However, the lack of fine tuning throughout the whole channel means that even the addition of the TruePatch cords will not completely eliminate throughput problems (Figure 25)

A benefits administration company — This last

installation example, occurring at a benefits administration company, shows huge mismatches and inconsistencies, particularly with cross-connect blocks and poorly installed cable This customer with a 10BASE-T network was considering an upgrade to 100BASE-T (Figure 26) TruePatch cords aren’t able to overcome the cross-connect block, but they do manage to smooth out the trace somewhat (Figure 27) This node will never run 100BASE-T effectively, as the massive oscillations will cause tremendous signal degradation (Figure 28) The TruePatch cord makes some difference, but not enough Our recommendation: don’t waste money on 100BASE-T active equipment without first fixing the infrastructure (Figure 29)

8

Distance (Meters)

40 50 60 70 80 90 100 110 120 130 140

IMPEDANCE MISMATCH OF THE PATCH CORD, THE BLOCK AND THE CABLE HARSH BENDS AND OVER TIGHTENED CABLE TIES

TYPICAL CHARACTERISTIC OF OVER PULLING TENSION

IMPEDANCE MISMATCH OF THE CABLE, THE STATION OUTLET AND THE PATCH CORD

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 26: Without TruePatch, this channel has huge mismatches and inconsistencies.

0 26 52 78 104 130 156 182 208 234 260

Frequency (MHz)

40 50 60 70 80 90 100 110 120 130

10BASE-T

100BASE-T

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 28: This node will not run 100BASE-T effectively.

Distance (Meters)

40 50 60 70 80 90 100 110 120 130 140 REDUCED IMPEDANCE MISMATCH OF THE BLOCK AND THE CABLE USING A TRUEPATCH PATCH CORD

REDUCED REFLECTION POINTS REDUCED IMPEDANCE MISMATCH

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 27: With TruePatch, the traces smooth out.

0 26 52 78 104 130 156 182 208 234 260 40

50 60 70 80 90 100 110 120 130 140

Frequency (MHz)

10BASE-T

100BASE-T

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 29: The TruePatch cord helps this node, but cannot remedy all the infrastructure problems.

Check Your

Infrastructure First

Armed with this

kind of evidence, we

would hope that most

companies would heed

our warning and look

first at network

infrastructure before

investing in more

software, active

equipment and/or

more bandwidth An

active system test is

a completely plausible

way to diagnose

network problems and,

as evidenced, it can

go a long way toward

pointing out some

network performance

deficiencies.

0 26 52 78 104 130 156 182 208 234 260

Frequency (MHz)

40 50 60 70 80 90 100 110 120 130

10BASE-T

100BASE-T

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 24: Original patch cord causes the impedance to drop.

0 26 52 78 104 130 156 182 208 234 260 40

50 60 70 80 90 100 110 120 130 140

Frequency (MHz)

10BASE-T

100BASE-T

—— PAIR 1 —— PAIR 2 —— —— PAIR 3 —— PAIR 4

Figure 25: The TruePatch cord helps to smooth out the frequency trace.

Ngày đăng: 16/01/2014, 21:20

w