From GEANT2-JRA1 Wiki

This page serves to present a study on "Recommendations and best practices for metric representation". The study is based on the metrics described on the section 6.2 of the document: GN2-04-150v14 (D.J.1.1.1): Requirements Report on the Design of the Measurement System.

The "Best Practices" are based on the representation of metrics used by many of the applications that already exist, mainly the ones identified in the JRA1 Requirements phase. The "Recommendations" part will be based on the visualisations available for the public in different NRENs and other possible ways of representing the given metrics.

Contents 1 Round-trip time (RTT) and one-way delay (OWD) 1.1 Round-trip time distribution on the NEMO tool ([1]) 1.2 Round trip times on the NEMO tool ([2]), with 8 intervals a 50 probes per hour 1.3 One-Way-Delay on DFN Erlangen G-Win ([3]) 1.4 Round-Trip Time on the MonALISA tool ([4]) 1.5 Round-Trip Time (real time) on the MonALISA tool ([5]) 1.6 Round trip times on the SmokePing tool at ISTF ([6]), with 20 ICMP Echo Pings (56 Bytes) every 60 seconds 2 Bandwidth utilization and IP available bandwidth 2.1 Bandwidth Utilization on the CNM tool ([7]), available for each network node 2.2 Throughput on the CNM tool ([8]), available for each network link 2.3 Total capacity of the link on the CNM tool ([9]), available for each network link 2.4 Percentage of utilization on the CNM tool ([10]), available for each network link 2.5 Throughput on the Nemo tool ([11]), plot graphic available for each network link 2.6 Traffic in percent of capacity on the NEMO tool ([12]), available for each network link 2.7 Traffic in packets per second on the NEMO tool ([13]), available for each network link 2.8 Bandwidth Utilization on FCCN ([14]) 2.9 Bandwidth Utilization (protocols used) on FCCN ([15]) 2.10 Bandwidth Utilization on the MonALISA tool ([16]), available for each MonALISA site 2.11 Bandwidth Utilization (real time) on the MonALISA tool ([17]) 3 Round-trip and one-way packet loss ratio 3.1 Percentage Loss on the NEMO tool ([18]), of 50 probes sent per hour to measure RTT 3.2 Packet Loss on DFN Erlangen G-Win ([19]) 3.3 Packet Loss on the MonALISA tool ([20]) 3.4 Packet Loss (real time) on the MonALISA tool ([21]), obtained by ABping module 4 Delay variation (jitter) 4.1 IP Delay Variation (Jitter) on DFN Erlangen G-Win ([22]) 4.2 Jitter on the MonALISA tool ([23]) 4.3 Jitter (real time) on the MonALISA tool ([24]), obtained by ABping module 5 Interface errors and drops 5.1 Percentage of ethernet errors on the NEMO tool ([25]), available for each network link 5.2 Percentage of input errors on the NEMO tool ([26]), available for each network link 5.3 Percentage of output discards / input ignores errors on the NEMO tool ([27]), available for each network link 5.4 Total of interface resets on the NEMO tool ([28]), available for each network link

Round-trip time (RTT) and one-way delay (OWD)

Round-trip time distribution on the NEMO tool ([1])

• Y-axis:
  • Number of hosts (this is a bar graph, that aggregates the number of hosts that had a certain average round-trip time)

• X-axis:
  • Average round-trip time in milliseconds

• Snapshot:

• Comments:'
  • It’s a very good representation, that allows for the user to quickly see if the network is having an “acceptable” behaviour.
  • It may be interresting to have, on the background, a curve representing the average (long term) situation or the optimum situation, so it would gives comparision indications against an ideal case.

Round trip times on the NEMO tool ([2]), with 8 intervals a 50 probes per hour

• Y-axis:
  • Round trip in milliseconds (the values “median”, “min” and max” are available, in the day plot in raw resolution)

• X-axis:
  • The time in raw value

• Snapshot:

• Comments:'
  • It’s very good to see max, min and median values on the same graph.
  • A scatter plot is available, but the line plot is much better, the scatter is a bit confusing, primarily because of the "symbols" that are used in the graph. I think they are a bit bigger than necessary, making it difficult to understand what round-trip (msec) corresponds to each time value
  • Month plot with only one click
  • There is a percentage loss plot associated with this graph, accounting for the loss percentage for the 50 probes sent to measure RTT. It’s a very good representation, that allows for the user to quickly see if the network is having an “acceptable” behaviour.

One-Way-Delay on DFN Erlangen G-Win ([3])

• Y-axis:
  • Delay in milliseconds (the values “Minimum”, “Maximum” and “Median” are available).
  • The values are obtained sending groups of UDP packets with timestamps, in an active measurement task, through the network. Each group of packets has a certain configuration of TOS bits, packet size, group size and distance between packets.

• X-axis:
  • The time

• Snapshot:

• Comments:
  • This scatter plot is a very interesting way of representing delay, it gives a quick idea of the network state at a certain time.
  • It’s only possible to see the daily graphic and, beginning with the graphic for the current day, it’s possible to go back or forward one day, back or forward one week and back or forward one month with only one click. It makes no sense to be able to go forward from the present day, as it is obvious that no data exists. The “forward” links should only appear after the user goes backwards in time.
  • It’s possible to do “zoom in” or “zoom out” on the graph, by clicking on it, or using links outside the graph. It’s also possible to do so by choosing the beginning and ending values of time on two fields and pressing “Display”. Unfortunately, the “Display” button doesn’t seem to work properly. It’s a nice feature to be able to see, with detail to the second (the maximum resolution is 10 seconds), the interval of time that the graph is displaying (the “Observation Timeframe”).

Round-Trip Time on the MonALISA tool ([4])

• Y-axis:
  • I will assume that the unit is “milliseconds”, but it’s not specified anywhere on the graph

• X-axis:
  • The local time

• Snapshot:

• Comments:
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • The graphs should have an indication of the corresponding date (day, month and year) of the hours it shows on the X-axis.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean, Max or Min Heights for each value and to choose the time interval of the plot.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph.

Round-Trip Time (real time) on the MonALISA tool ([5])

• Y-axis:
  • I will assume that the unit is “milliseconds”, but it’s not specified anywhere on the graph.

• X-axis:
  • Node(s) in question

• Snapshot:

• Comments:
  • This is a “Real-Time Plot” representation of RTT between the “local host” chosen on the application and other hosts connected to it. It’s also possible to see the same graph on a 3D-Chart representation;
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean Height for the RTT value.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph.

Round trip times on the SmokePing tool at ISTF ([6]), with 20 ICMP Echo Pings (56 Bytes) every 60 seconds

• Y-axis
  • Median Round-Trip Time in milliseconds and Packet Loss (the values “average RTT”, and “average packet loss” are available)

• X-axis
  • The time

• Snapshot:

• Comments:
  • For every round of measurement, SmokePing sends several packets. It then sorts the different round trip times and selects the median, (ie. the middle one). This means when there are 10 time values, value number 5 is selected and drawn on the graphic. The other values are drawn as successively lighter shades of gray in the background (smoke). Sometimes a test packet is sent out but never returns. This packet-loss is represented by the colour of the median line changing accordingly to the number of packets lost.
  • It’s a good representation of network health, making it easy and intuitive to see the fluctuation of RTT and packet loss of the ICMP echo pings that travel the network.
  • SmokePing is not limited to testing just the roundtrip time of the packets. It can also perform some task at the remote end ("probe"), like download a webpage.
  • The first graph is a "daily summary" of a node, and by clicking on it the user will see the SmokePing graphics for last 90 minutes (1 minute average), last 8 hours (5 minutes average), last 24 hours (15 minutes average), last 72 hours (1 hour average), last 7 days (2 hours average), last 30 days (8 hours average), last 90 days (1 day average), and last year (1 day average).

Bandwidth utilization and IP available bandwidth

Bandwidth Utilization on the CNM tool ([7]), available for each network node

• Y-axis:
  • Packets or thousand packets per second
  • Average, minimum and maximum are available.
  • Average, minimum and maximum are available.
  • An the window menu, it’s possible to select the graph mode: “forward and backward” “detailed forward” and “detailed backward”
  • In “forward and backward” mode, values are represented as two step plots over the aggregation interval, with one line for “forward” and one for “backward”.
  • In “detailed forward” mode, values are represented as three step plots over the aggregation interval, one line for “forward average values”, one for “forward minimal values”, and another for “forward maximal values”.
  • In “detailed backward” mode, values are represented as three step plots over the aggregation interval, one line for “backward average values”, one for “backward minimal values”, and another for “backward maximal values”.
  • It’s possible to define what is the highest and/or the lowest value of the Y-Axis, and to see the graph with that new scale.

• X-axis:
  • The time (in daily, weekly, monthly or yearly averaged value)

• Snapshot:

• Comments:
  • The “forward” and “backward” step plots should not be represented as lines and over one another, it gets confusing to distinguish each direction when they overlap. . Maybe if “forward” was in the upper Y-axis and “backward” in the lower Y-axis, the graphs would be easier to read. The three lines for average, maximal and minimal values also suffer from this “overlap” difficulty.
  • These graphs are much better with small periods of time, with longer periods of time, the graphs become almost unreadable.
  • It’s great to be able to go forwards or backwards in time, always seeing what happened before or after the time period initially selected. The step forward or backward is always of one unit of the time period chosen, one day, one week, one month, one year.

Throughput on the CNM tool ([8]), available for each network link

• Y-axis:
  • Kbits or Mbits per second
  • Average, minimum and maximum are available.
  • In the window menu, it’s possible to select the graph mode: “forward and backward” “detailed forward” and “detailed backward”
  • In “forward and backward” mode, values are represented as two step plots over the aggregation interval, with one line for “forward” and one for “backward”.
  • In “detailed forward” mode, values are represented as three step plots over the aggregation interval, one line for “forward average values”, one for “forward minimal values”, and another for “forward maximal values”.
  • In “detailed backward” mode, values are represented as three step plots over the aggregation interval, one line for “backward average values”, one for “backward minimal values”, and another for “backward maximal values”.
  • It’s possible to define what is the highest and/or the lowest value of the Y-Axis, and to see the graph with that new scale.

• X-axis:
  • The time (in daily, weekly, monthly or yearly averaged value)
  • It’s possible to go forwards or backwards in the X-axis, just by pressing the “»” or “«” buttons below the graph
  • It’s possible to re-draw the graph with the new X-axis in the same window, or to open a new window.

• Snapshot:

• Comments:
  • The “forward” and “backward” step plots should not be represented as lines and over one another, it gets confusing to distinguish each direction when they overlap
  • Maybe if “forward” was in the upper Y-axis and “backward” in the lower Y-axis, the graphs would be easier to read
  • The three lines for average, maximal and minimal values also suffer from this “overlap” difficulty.
  • These graphs are much better with small periods of time, with longer periods of time, the graphs become almost unreadable.
  • It’s great to be able to go forwards or backwards in time, always seeing what happened before or after the time period initially selected
  • The step forward or backward is always of one unit of the time period chosen, one day, one week, one month, one year.

Total capacity of the link on the CNM tool ([9]), available for each network link

• Y-axis:
  • Mbits per second
  • Average, minimum and maximum are available.
  • In the window menu, it’s possible to select the graph mode: “forward and backward” “detailed forward” and “detailed backward”
  • It’s possible to define what is the highest and/or the lowest value of the Y-Axis, and to see the graph with that new scale.

• X-axis:
  • The time (in daily, weekly, monthly or yearly averaged value)
  • It’s possible to go forwards or backwards in the X-axis, just by pressing the “»” or “«” buttons below the graph. It’s possible to re-draw the graph with the new X-axis in the same window, or to open a new window.

• Snapshot:

• Comments:
  • This graph is basically one straight line with time on X-axis and total capacity of the link on the Y-axis, so it shouldn’t have the option of changing “graph modes” and maximum and minimum values.

Percentage of utilization on the CNM tool ([10]), available for each network link

• Y-axis
  • Percentage of link utilization, in respect to the maximum capacity of the link
  • Average, minimum and maximum are available.
  • In the window menu, it’s possible to select the graph mode: “forward and backward” “detailed forward” and “detailed backward”
  • In “forward and backward” mode, values are represented as two step plots over the aggregation interval, with one line for “forward” and one for “backward”.
  • In “detailed forward” mode, values are represented as three step plots over the aggregation interval, one line for “forward average values”, one for “forward minimal values”, and another for “forward maximal values”.
  • In “detailed backward” mode, values are represented as three step plots over the aggregation interval, one line for “backward average values”, one for “backward minimal values”, and another for “backward maximal values”.

.It’s possible to define what is the highest and/or the lowest value of the Y-Axis, and to see the graph with that new scale.

• X-axis:
  • The time (in daily, weekly, monthly or yearly averaged value)
  • It’s possible to go forwards or backwards in the X-axis, just by pressing the “»” or “«” buttons below the graph**It’s possible to re-draw the graph with the new X-axis in the same window, or to open a new window.

• Snapshot:

• Comments:
  • The “forward” and “backward” step plots should not be represented as lines and over one another, it gets confusing to distinguish each direction when they overlap. Maybe if “forward” was in the upper Y-axis and “backward” in the lower Y-axis, the graphs would be easier to read. The three lines for average, maximal and minimal values also suffer from this “overlap” difficulty.
  • These graphs are much better with small periods of time, with longer periods of time, the graphs become almost unreadable.
    • Maybe we should have an aggregation factor which would allow the user to aggregate the data to an acceptable level on the line (e.g. kind of zoom in/out but for aggregation: aggregate +/- with pre-defined aggregation level e.g. : raw, 1min, 5min, 20 min, 1h, 1d, 1w, 1m)
  • It’s great to be able to go forwards or backwards in time, always seeing what happened before or after the time period initially selected. The step forward or backward is always of one unit of the time period chosen, one day, one week, one month, one year.

Throughput on the Nemo tool ([11]), plot graphic available for each network link

• Y-axis:
  • Kbits per second

• X-axis:
  • The time in raw and hourly averaged value
  • It’s possible to see the plot for the period before the period the user is watching at that moment (for example, the day before the one in the plot, the week before the one in the plot, …), with the same resolution, and to progress, forwards or backwards, only with one mouse click.
  • Weekly or monthly plots are available with only one click.

• Snapshot:

• Comments:
  • It is great to be able to do “zoom in” or “zoom out” on the graphics, just by clicking on them. All the “zoom” plots show lines of “peek-in” and “peek-out” values.
  • “Zoom out” is done clicking on the upper part of the graph, and starting from the daily graph in “raw resolution”, one click shows the same graph with hour resolution, and lines of “peek-in” and “peek-out” values, two clicks show the week plot in hour resolution, three clicks show the monthly plot in day resolution, four clicks show the yearly plot in week resolution, and five clicks show the epoch plot in week resolution (all the data, since the inauguration of the network link), six clicks show the epoch plot in day resolution, as long as there is measured data in the database.
  • “Zoom in” is done clicking just below the “Zoom out” area. Holding the mouse still will show information about the next graph, in “zoom in” or “zoom out”.
  • It’s great to be able to go forwards or backwards in time, just by clicking the right side or the left side of the graph. If the user is seeing the daily plot, one click on the left shows the plot for the day before. One click on the right shows the plot for the day after. The same applies for the weekly, monthly or yearly plots.
  • A good thing is to be able to see a lot of other graphics, which focus on the same period of time the user is seeing at that moment. For instance, if he is seeing the week plot that shows the traffic in hour resolution, and he clicks on the link that says “Packets per second”, he will see a plot that shows the input errors of that link on that same week. The graphics available on each period of time vary, and are always listed on the section: “You may further inspect for this period/port”.
  • Examples of the graphs available are:
    • Traffic in percent of capacity
    • Ethernet errors
    • Input errors
    • Packets per second
    • Output discards / input ignores
    • Interface resets
    • Error hours above thresholds
  • Beyond the “further inspection” of the current period, on each graph page there are still other links that change according to the graph the user is seeing in each moment. Basically, the links give the possibility to go forward or backward in time ("zoom out" and see the "bigger picture" or "zoom in" and see the "finer details").
  • Like in CNM, NEMO should use dashed lines which form a grid in the background of each picture, so that it would become easier to guess the value of a date in question.
  • The Y-Axis in Kbps and the numbers expressed in the thousand is not readable. A "0" is quickly missed.

Traffic in percent of capacity on the NEMO tool ([12]), available for each network link

• Y-axis:
  • Percentage of link utilization, in respect to the maximum capacity of the link (the values “in” and “out” are available, in the day plot in raw resolution, and the values “peek-in” and “peek-out” in the day and week plot in hour resolution, and the values “peek hour in”, ”peek hour out”, “busy hour in”, “busy hour out”, “average-in” and “average-out” in the month and year plot in day and week resolution);

• X-axis:
  • The time in raw and hourly averaged value
  • It’s possible to see the plot for the period before the period the user is watching at that moment (for example, the day before the one in the plot, the week before the one in the plot, …), with the same resolution, and to progress, forwards or backwards, only with one mouse click.
  • Weekly or monthly plots are available with only one click.

• Snapshot:

• Comments:'
  • The same comments as the previous graph apply to this graph.

Traffic in packets per second on the NEMO tool ([13]), available for each network link

• Y-axis:
  • Percentage of link utilization, in respect to the maximum capacity of the link (the values “in” and “out” are available, in day and week plot in raw and hour resolution);

• X-axis:
  • The time in raw and hourly averaged value
  • It’s possible to see the plot for the period before the period the user is watching at that moment (for example, the day before the one in the plot, the week before the one in the plot, …), with the same resolution, and to progress, forwards or backwards, only with one mouse click.
  • Weekly or monthly plots are available with only one click.

• Snapshot:

• Comments:'
  • The same comments as the previous graph apply to this graph.

Bandwidth Utilization on FCCN ([14])

• Y-axis:
  • Bits per second, on a 5 minute average
  • Two different types of traffic are available in the out direction, their last value, their minimum, median and maximum, and the accumulated value for all of them, on that direction. Dropped traffic is also available.
  • The “out” direction is on the upper side of the Y-axis, while the “in” direction is on the lower side.

• X-axis:
  • The time (in daily, weekly, monthly or yearly averaged value)

• Snapshot:

• Comments:
  • By clicking the graphic, a page appears with a daily graph (5 minute average), a weekly graph (30 minute average), a monthly graph (2 hour average) and a yearly graph (1 day average).
  • In this representation, it’s easy to differentiate between inbound and outbound traffic, and to show how traffic is composed.
  • Unfortunately, this graph is not publicly available.

Bandwidth Utilization (protocols used) on FCCN ([15])

• Y-axis:
  • Bits per second, on a 5 minute average
  • The values available are the protocols that Cisco NBAR is able to recognise, and some user-customised protocols. Some scripts read the NBAR results from time to time, and create/update the RRD-files, that then are graphed. For each protocol, is available the last value, the minimum, median and maximum, and accumulated values, on each direction.
  • The “out” direction is on the upper side of the Y-axis, while the “in” direction is on the lower side.

• X-axis:
  • The time (in daily, weekly, monthly or yearly averaged value)

• Snapshot:

• Comments:
  • By clicking the graphic, a page appears with a daily graph (5 minute average), a weekly graph (30 minute average), a monthly graph (2 hour average) and a yearly graph (1 day average).
  • In this representation, it’s easy to differentiate between inbound and outbound traffic, and to show how traffic is composed.
  • One problem is about the protocols that appear and then quickly disappear, leaving the graph over-populated with values that are all zero. From time to time, one has to “clean” the graph.
  • Unfortunately, this graph is not publicly available.

Bandwidth Utilization on the MonALISA tool ([16]), available for each MonALISA site

• Y-axis:
  • I will assume that this is Mbits/s, but it’s not specified anywhere in the graph.

• X-axis:
  • The local time

• Snapshot:

• Comments:
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • The graphs should have an indication of the corresponding date (day, month and year) of the hours it shows on the X-axis.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean, Max or Min Heights for each value and to choose the time interval of the plot.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph.
  • The scatter graphic is a bit confusing, the symbols used are too big, making it difficult to sometimes understand the time that corresponds to a given value.

Bandwidth Utilization (real time) on the MonALISA tool ([17])

• Y-axis:
  • I will assume that this is Mbits/s, but it’s not specified anywhere in the graph.

• X-axis:
  • Node(s) in question

• Snapshot:

• Comments:
  • This is a “Real-Time Plot” representation of the bandwidth utilization on the chosen interface on the host selected in the application. It’s also possible to see the same graph on a 3D-Chart representation;
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean Height for each of the values presented.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph.

Round-trip and one-way packet loss ratio

Percentage Loss on the NEMO tool ([18]), of 50 probes sent per hour to measure RTT

• Y-axis:
  • Loss in percentage of the 50 probes sent

• X-axis:
  • The time

• Snapshot:

• Comments:'
  • Monthly plot with only one click;
  • This graph is associated with another that measures round trip times, and this graph measures the loss of the probes sent when measuring RTT.

Packet Loss on DFN Erlangen G-Win ([19])

• Y-axis:
  • Number of lost packets (the value “Lost Packets (cumulative)” is available)
  • The number of lost packets is obtained sending groups of UDP packets with timestamps, in an active measurement task, through the network, and counting how many got lost before reaching the destination. Each group of packets has a certain configuration of TOS bits, packet size, group size and distance between packets.

• X-axis:
  • The time

• Snapshot:

• Comments:
  • Line plot with the cumulative number of lost packets at a given time. It’s a simple and effective representation. However, the line should be continuous (on the picture shown, the transition from 0 to 2 lost packets should have a line connecting the two values), and a grid on the image would help identifying the value of the Y-axis that corresponds to a given time on the X-axis.
  • It’s only possible to see the daily graphic and, beginning with the graphic for the current day, it’s possible to go back or forward one day, back or forward one week and back or forward one month with only one click. It makes no sense to be able to go forward from the present day, as it is obvious that no data exists. The “forward” links should only appear after the user goes backwards in time.
  • It’s possible to do “zoom in” or “zoom out” on the graph, by clicking on it, or using links outside the graph. It’s a nice feature to be able to see, with detail to the second (the maximum resolution is 10 seconds), the interval of time that the graph is displaying (the “Observation Timeframe”).
  • There are also two links that enable the user to see the same graph with the beginning and ending timeframe advancing (“Later”) or receding (“Earlier”) three hours.
  • It’s possible to limit the Y-values within a certain range, either by clicking on the left are of the graph, or by choosing the beginning and ending values on two fields and pressing “Display”. Unfortunately, the “Display” button doesn’t seem to work properly.

Packet Loss on the MonALISA tool ([20])

• Y-axis:
  • I will assume that the unit is “number of packets”, but it’s not specified anywhere on the graph

• X-axis:
  • The local time

• Snapshot:

• Comments:
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • The graphs should have an indication of the corresponding date (day, month and year) of the hours it shows on the X-axis.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean, Max or Min Heights for each value and to choose the time interval of the plot.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph

Packet Loss (real time) on the MonALISA tool ([21]), obtained by ABping module

• Y-axis:
  • I will assume that the unit is “number of packets”, but it’s not specified anywhere on the graph.

• X-axis:
  • Node(s) in question

• Snapshot:

• Comments:
  • This is a “Real-Time Plot” representation of Packet Loss between the “local host” chosen on the application and other hosts connected to it. It’s also possible to see the same graph on a 3D-Chart representation;
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean Height for the Packet Loss value.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph

Delay variation (jitter)

IP Delay Variation (Jitter) on DFN Erlangen G-Win ([22])

• Y-axis:
  • Delay variation in milliseconds (the values “Group Minimum IPDV”, “Group Maximum IPDV” and “Group Median IPDV” are available).
  • "Group" refers to the group of UDP packets with timestamp that are sent, in an active measurement task, through the network to test its jitter. Each group of packets has a certain configuration of TOS bits, packet size, group size and distance between packets.

• X-axis:
  • The time

• Snapshot:

• Comments:'
  • This scatter plot is a very interesting way of representing jitter, it gives a quick idea of the jitter on the network at a certain time.
  • It’s only possible to see the daily graphic and, beginning with the graphic for the current day, it’s possible to go back or forward one day, back or forward one week and back or forward one month with only one click. It makes no sense to be able to go forward from the present day, as it is obvious that no data exists. The “forward” links should only appear after the user goes backwards in time.
  • It’s possible to do “zoom in” or “zoom out” on the graph, by clicking on it, or using links outside the graph. It’s also possible to do so by choosing the beginning and ending values of time on two fields and pressing “Display”. Unfortunately, the “Display” button doesn’t seem to work properly. It’s a nice feature to be able to see, with detail to the second (the maximum resolution is 10 seconds), the interval of time that the graph is displaying (the “Observation Timeframe”).
  • There are also two links that enable the user to see the same graph with the beginning and ending timeframe advancing (“Later”) or receding (“Earlier”) two hours.

Jitter on the MonALISA tool ([23])

• Y-axis:
  • I will assume that the unit is seconds, but it’s not specified anywhere on the graph

• X-axis:
  • The local time

• Snapshot:

• Comments:
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • The graphs should have an indication of the corresponding date (day, month and year) of the hours it shows on the X-axis.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean, Max or Min Heights for each value and to choose the time interval of the plot.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph.
  • The scatter graphic is a bit confusing, the symbols used are too big, making it difficult to sometimes understand the time that corresponds to a given value.

Jitter (real time) on the MonALISA tool ([24]), obtained by ABping module

• Y-axis:
  • I will assume that the unit is seconds, but it’s not specified anywhere on the graph.

• X-axis:
  • Node in question

• Snapshot:

• Comments:
  • This is a “Real-Time Plot” representation of Jitter between the “local host” chosen on the application and other hosts connected to it. It’s also possible to see the same graph on a 3D-Chart representation;
  • A negative point is the absence of unit identification on the Y-axis. It’s very difficult for the user to understand what units are on that axis, he has to guess.
  • A positive point is the fact that the graphics are completely customizable. The colours, the type of graph (line, scatter, area, or the three types at the same time), the values shown (it’s possible to add or remove values).
  • It’s possible to get the Mean Height for the Jitter value.
  • By holding a mouse button and dragging it on the graph, it’s possible to zoom into that area. Waiting some moments does bring back the original graph.

Interface errors and drops

Percentage of ethernet errors on the NEMO tool ([25]), available for each network link

• Y-axis:

. Percentage of ethernet errors rate (the values “in-errors”, “out-errors”, “CRC”, “Frame”, “Runts” and “Giants” are available, in raw and hour resolution)

• X-axis:

. The time in raw and hourly averaged value . It’s possible to see the plot for the period before the period the user is watching at that moment (for example, the day before the one in the plot, the week before the one in the plot, …), with the same resolution, and to progress, forwards or backwards, only with one mouse click. . Weekly or monthly plots are available with only one click.

• Snapshot:

• Comments:'
  • It is great to be able to do “zoom in” or “zoom out” on the graphics, just by clicking on them. All the “zoom” plots show lines of “peek-in” and “peek-out” values.
  • “Zoom out” is done clicking on the upper part of the graph, and starting from the daily graph in “raw resolution”, one click shows the same graph with hour resolution, and lines of “peek-in” and “peek-out” values, two clicks show the week plot in hour resolution, three clicks show the monthly plot in day resolution, four clicks show the yearly plot in week resolution, and five clicks show the epoch plot in week resolution (all the data, since the inauguration of the network link), six clicks show the epoch plot in day resolution, as long as there is measured data in the database.
  • “Zoom in” is done clicking just below the “Zoom out” area. Holding the mouse still will show information about the next graph, in “zoom in” or “zoom out”.
  • It’s great to be able to go forwards or backwards in time, just by clicking the right side or the left side of the graph. If the user is seeing the daily plot, one click on the left shows the plot for the day before. One click on the right shows the plot for the day after. The same applies for the weekly, monthly or yearly plots.
  • A good thing is to be able to see a lot of other graphics, which focus on the same period of time the user is seeing at that moment. For instance, if he is seeing the week plot that shows the traffic in hour resolution, and he clicks on the link that says “Packets per second”, he will see a plot that shows the input errors of that link on that same week. The graphics available on each period of time vary, and are always listed on the section: “You may further inspect for this period/port”.
  • Examples of the graphs available are:
    • Traffic in percent of capacity
    • Ethernet errors
    • Input errors
    • Packets per second
    • Output discards / input ignores
    • Interface resets
    • Error hours above thresholds
  • Beyond the “further inspection” of the current period, on each graph page there are still other links that change according to the graph the user is seeing in each moment. Basically, the links give the possibility to go forward or backward in time ("zoom out" and see the "bigger picture" or "zoom in" and see the "finer details").

Percentage of input errors on the NEMO tool ([26]), available for each network link

• Y-axis:
  • Percentage of input error rate (the values “Errors”, “CRC”, “Frame”, and “Abort” are available, in raw and hour resolution, the value “Abort” is substituted by “Ignores” in the month plot in day resolution, and the values in the year plot with week resolution are “max day input errors”, “average input errors”, “max day ignores” and “average ignores”)

• X-axis:
  • The time in raw and hourly averaged value
  • It’s possible to see the plot for the period before the period the user is watching at that moment (for example, the day before the one in the plot, the week before the one in the plot, …), with the same resolution, and to progress, forwards or backwards, only with one mouse click.
  • Weekly or monthly plots are available with only one click.

• Snapshot:

• Comments:'
  • The same comments as the previous graph apply to this graph.

Percentage of output discards / input ignores errors on the NEMO tool ([27]), available for each network link

• Y-axis:
  • Percentage of output discards/ input ignores error rate (the values “Discards” and “Ignores” are available, in the day plot in raw and hour resolution, the values “peek-hour” and “day average” are available in the month plot in day resolution, the value “week-average” appears in the year plot in week resolution)

• X-axis:
  • The time in raw and hourly averaged value
  • It’s possible to see the plot for the period before the period the user is watching at that moment (for example, the day before the one in the plot, the week before the one in the plot, …), with the same resolution, and to progress, forwards or backwards, only with one mouse click.
  • Weekly or monthly plots are available with only one click.

• Snapshot:

• Comments:'
  • The same comments as the previous graph apply to this graph.

Total of interface resets on the NEMO tool ([28]), available for each network link

• Y-axis
  • Number of interface resets (the value “Resets” is available, in the day plot in raw and hour resolution, and in the week plot in hour resolution)

• X-axis
  • The time in raw and hourly averaged value
  • It’s possible to see the plot for the period before the period the user is watching at that moment (for example, the day before the one in the plot, the week before the one in the plot, …), with the same resolution, and to progress, forwards or backwards, only with one mouse click.
  • Weekly or monthly plots are available with only one click.

• Snapshot:

• Comments:'
  • The same comments as the previous graph apply to this graph.