Monitoring and alerting manual

Answer Bid Ratio. For per-hour stats this is a ratio of all calls with duration completed in the hour to all attempted calls completed/rejected in the hour. For EMA stats the same formula is used, but over the EMA window. Codes, included into ABR calculation may be adjusted in Disconnect codes.

Names in table columns: In ABR, In Hr ABR, Out ABR, Out Hr ABR.

Exponential Moving Average. A way to calculate statistics for an EMA window of calls, rather than for a certain time period. With EMA, the more recent the call/attempt was, the more weight in calculation it had. Thus, EMA gives a good representation of statistical values over a short recent period, and can catch abrupt variations of those values.

EMA-calculated statistical data is represented by a line on a Chart canvas, while per-hour statistics are shown as bars. Per-hour statistics cannot very often give enough precision for monitoring some critical parameters that may change very quickly.

For example, if ACD starts dropping at the beginning of an hour, having only per-hour stats at hands, one will notice the drop only in roughly one hour, when the statistics for the whole hour is calculated. With EMA charts, the drop will be obvious in 5-10 minutes.

Without going into too much math, EMA window is roughly a certain number of calls within which the statistical parameter is calculated. The more the window - the smoother the statistics chart will be, the less the window - the easier it will be to catch small quick variations in statistics, but the chart will look quite like a saw.

Usually, good optimum is provided if the EMA window equals the number of current calls of the monitored object. For this reason, 5gVision is using an adaptive EMA window, that always adjusts itself to the number of calls. This way objects with 10 calls and 1000 calls will both have relatively smooth curves, still allowing you to quickly notice negative variations.

Collecting Per-hour statistics is the classical way of monitoring voice networks. All values are calculated for the given hour. Per-hour parameters always have an Hr indicator in their column names, like In Hr ACD and are underlined. On charts they are represented as bars, one per each hour. If there are too many bars (hours) to show them distinctly, bars are collapsed into jagged lines.

Depending on your switch, Active calls may provide information only on total calls (including the calls being established), only on connected calls, or both. Since a switch is usually polled for active calls info once a minute, these stats may have a lag of up to one minute from the real current calls in the switch.

Names in table columns: In Calls, In Conn, Out Calls, Out Conn. In and out calls may be different for Customers, Vendors and other objects (see Incoming and Outgoing), but will be same for Areas.

This type of stats is shown if your switch can provide information on capacity limits, and is usually collected for the following Objects:

• Contractors (Customers or Vendors)
• Equipment
• Equipment groups.

It simply reflects the simultaneous call limitations set in the switch.

Route load is showing a ratio of Active calls divided by current capacity.

Names in table columns: In Max, In Load, Out Max, Out Load.

A sum-up of all attempts per specified object hitting the switch in the given hour, both completed and failed calls. Attempts are included in the hour by the disconnect time.

Names in table columns: In Hr Atmpt, Out Hr Atmpt.

A sum-up of all connected calls per specified object in the given hour. Connected calls are included in the hour by the disconnect time.

Names in table columns: In Hr Conn, Out Hr Conn.

A sum-up of all connected minutes per specified object in the given hour. Connected minutes are included in the hour by the disconnect time of a corresponding call.

Names in table columns: In Hr Mins, Out Hr Mins.

Average Call Duration. For per-hour stats this is a ratio of all minutes completed in the hour to all connected calls completed in the hour. For EMA stats the same formula is used, but over the EMA window. Codes, included into ACD calculation may be adjusted in Disconnect codes.

Names in table columns: In ACD, In Hr ACD, Out ACD, Out Hr ACD.

Percentage of Short Calls to all connected calls. There are 2 parameters currently:

• PSC1 - % of calls less then 20 seconds
• PSC2 - % of calls less then 60 seconds

These settings can be adjusted upon request.

Names in table columns: In PSC1, In PSC2, Out PSC1, Out PSC2.

Answer Seizure Ratio. For per-hour stats this is a ratio of all calls with duration completed in the hour to all attempted calls completed/rejected in the hour minus calls that were rejected with certain disconnect codes, adjustable in Disconnect codes.

The codes that are usually ignored are SIP 503 and H.323 34 codes, as they usually indicate that a call was rejected right away due to no capacity, and should be rerouted to the next vendor by the sending party. Since SIP 503 or H.323 34 codes do not really indicate any errors in such situations, common trend in the industry was to exclude them from ASR calculation.

See also ABR which takes all codes into consideration (unless configured differently in Disconnect codes).

Names in table columns: In ASR, In Hr ASR, Out ASR, Out Hr ASR.

Percentage of SIP 487 codes and H.323 16 codes with no duration to **all attempted calls**. The codes above both indicate that the call was ended by the calling party before it was able to connect. High value of this parameter may indicate that the network is killing a lot of calls, for instance, trying to connect a call for too long, so that users are tired of waiting, or playing announcements before the connect, so that users are forced to end their calls. Codes, included into calculation of % of 487 codes may be adjusted in Disconnect codes.

Names in table columns: In 487, In Hr 487, Out 487, Out Hr 487.

Network Efficiency Ratio measures the ability of a network to deliver a call to the called terminal. Busy signals and other call failure due to user behaviour are counted as "successful" call delivery for NER calculation purposes.

By default, the following disconnect codes are considered "successful" (though for ASR or ABR they are not):

• SIP 480 - Temporarily Unavailable
• SIP 486 - Busy Here
• SIP 600 - Busy Everywhere
• H323 17 - User busy

The codes can be further adjusted in Disconnect codes.

Calls Per Second. A parameter showing how many call attempts per second were generated to/from the specified object. Real CPS is usually hard to obtain from most switches, at best, the switch will provide total CPS values or values per each signaling node, while 5gVision is showing CPS per any object or combination.

- 5gVision is estimating CPS via CDRs, looking for a second with the highest number of CDRs in it, and showing this value as the highest CPS for the whole minute. Since each CDR is written upon completion, not start of a call, the 5gVision CPS shows "calls ended per second", not "calls started per second". However, as very high CPS usually happens on systems with a lot of calls rejected, and calls are rejected within seconds from the setup, chances are, the estimated CPS will be very close to the real CPS.

Names in table columns: In CPS, Out CPS. Also, please note that there are 2 other useful parameters available:

• In Avg Atmpt/min - Incoming average attempts per minute.
• In Avg Conn/min - Incoming average connects per minute.

These parameters are averaged over a window of calls (see EMA and thus have a smoother nature than a bit jagged CPS parameter and can give you a good picture of how your attempts/connects grow within each hour.

Post Dial Delay. Measured as the difference between the alerting and setup times. High PDD values indicate that users wait for too long to hear the first ring from the called party.

PDD may not always provide realistic information in cases when alerting time is not available in CDRs, or a switch does not determine the alerting time correctly if multiple alerts were returned from different vendors while rerouting. We recommend using the TTC parameter in such cases.

Names in table columns: In PDD, In Hr PDD, Out PDD, Out Hr PDD.

Time To Connect. Measured as the difference between the connect and setup time. This parameter gives a slightly different perspective from PDD on how quickly your network connects calls and is useful when PDD can't be calculated for your type of switch.

Time To Connect will include the time when the end user's phone rings, so it should always be considered in comparison to the previous stats. For example, if TTC in the network was about 20 seconds (15 of them perhaps being just rings before connect), and then changed to 30 seconds, this may indicate a problem.

TTC is calculated only for incoming statistics, as this parameter can not show a true TTC for a Vendor, if there were other hunts that added up to the TTC before the call finally connected through this Vendor.

Names in table columns: In TTC, In Hr TTC.

Time To Reject. Measured as the difference between the disconnect and setup time only for non-connected calls. This parameter will show how quickly your network rejects the calls it cannot put through.

Calculated only for incoming statistics. Names in table columns: In TTR, In Hr TTR.

Average number of hunts (rerouting attempts). Calculated only for incoming statistics, as for the outgoing statistics this value will always be 1 (every Vendor or destination Equipment is getting only one rerouting attempt per call).

Names in table columns: In Hunts, In Hr Hunts.

Almost all statistical parameters in the system are calculated both for the incoming and outgoing traffic. Parameter names in column headers start with either In or Out to differentiate. They also are highlighted with different colors. An Object may have quite different values of the very same parameter for incoming and outgoing stats. For instance, a Contractor may be a Customer, and Vendor at the same time and have In ACD of 6 min. when sending calls to your network, but Out ACD of only 3 min. when accepting calls from your network.

One tricky situation is monitoring Customer->Area->Vendor and Customer->Vendor combinations. For example, in case of Customer->Area->Vendor the top parent of Customer will have stats in the In type of columns (let's say the customer is sending us 300 calls), the Area - again in In (say, 150 calls go to this very Area), but the Vendor will show the same calls in In and Out columns (say, 50 calls from this Customer to this Area go to this very Vendor). However, viewing them in both IN and OUT columns is convenient in some cases.

An Object refers to any element that 5gVision is capable of collecting information about. Currently, the following objects and their combinations are supported:

• Contractors (Customers or Vendors), c
• Customers->Vendors, c->c
• Destination Areas, a
• Customers/Vendors->Areas, c->a, a->c
• Customers->Areas->Vendors, c->a->c
• Disconnect Codes (Reasons), r
• Customers/Vendors->Disconnect Codes, c->r
• Areas->Disconnect Codes, a->r
• Customers->Areas->Disconnect Codes, c->a->r
• Source Areas, b
• Source Areas->Customers, b->c
• Source Areas->Customers->Destination Areas, b->c->a
• Products/Rate plans/Partitions, p
• Customers/Vendors->Products/Rate plans/Partitions, c->p
• Areas->Products/Rate plans/Partitions, a->p
• Customers/Vendors->Areas->Products/Rate plans/Partitions, c->a->p
• Customers->Vendors->Products/Rate plans/Partitions, c->c->p
• Equipment, e
• Equipment groups or Access numbers, g
• Accounts, s
• Switch nodes, n
• IP addresses from signaling logs, j
• U-objects, c, a, e, g
• Flex combinations, f
• SNMP Devices, d
• External DB objects, t, x, y, z
• Poller internal objects, i

Objects have IDs and Names. Usually, they both reflect the ones in the VoIP switch configuration. All IDs have an object-specific prefix, added to easily distinguish objects of different types:

• c - for Customers or Vendors (cXXX.XXX may be used for separate Trunks)
• a - for destination Areas (a2, a3 for Area sets 2 and 3)
• p - for Products/Rate plans/Partitions
• b - for source Areas
• e - for Equipment
• g - for Equipment Groups or Access numbers
• s - for Accounts
• n - for switch Nodes
• r - for Disconnect Codes (Reasons)
• f - for Flex combinations
• d - for SNMP Devices, see Config-SNMP
• i - for Poller internal stats
• t - for External DB tables
• x, y, z - for objects in External DB tables

You may use these initial letters to quickly filter tables for the object types needed, for instance a filter like c||a will show only Customers/Vendors and Areas, see more in Filtering objects.

An Object statistical parameter that is monitored (Calls, ACD, etc.), or a CDR/Calls/Signaling collector, Media collector records field. For the main parameters monitored, please go to Introduction. Complete list of parameters/fields will differ per Module and may be found by pressing the C (Columns) menu button in each Module table menu.

If a mouse is clicked and dragged over the Chart canvas containing any charts, the light blue zoom zone will appear. Clicking on the zone will bring the chart data for the period zoomed. To get back to the original chart you may simply click on the Chart Interval again. Zoom will only work for the intervals over 30 minutes. If a chart was zoomed, the Cust interval will be highlighted, and the Custom interval will be automatically set to the zoomed one.

When the Cust option is clicked on the Interval strip, the Interval picker will pop up. You may choose the custom interval you need from the picker, or, in the Chart module, by using the Zoom feature. Since the Chart module caches recently fetched charts, clicking on the Cust will first load the chart for the last selected custom interval, the second click will then pop up Interval picker window. You may also Ctrl-click to call the Interval picker right away.

When the Cust option is clicked on the Interval strip, the Interval picker window is invoked to let one choose non-standard intervals. The window consists of 4 sections:

• Calendar to choose a day. When a beginning day is chosen, the time is automatically set to 00:00:00, when the ending one - to 23:59:59. Clicking at MonthYear at the top (like Jan 2012) will set the current time.
• Hour strip. When a beginning hour is chosen, the time is automatically set to HOUR:00:00, when the ending one - to HOUR:59:59.
• Minute strip. Choosing a beginning minute is setting the seconds to 00, the ending one - to 59.
• Date-time string. You may edit or paste the string directly. Once the mouse is clicked outside of the string window, all calendar values update accordingly.

There are 2 extra options in the CDR and DC modules: Local time and CDR time. Let's suppose that your CDRs are stored in UTC, and your local time is EST and, moreover, there is a 30 second time discrepancy between your local machine and the CDR server. Here is how those settings will affect your results if you choose an interval from 2012-01-29 09:00:00 to 2012-01-29 10:59:59:

• Local time. The CDR date range you will see in the result set will be from 2012-01-29 14:00:30 to 2012-01-29 16:00:29 because of both the time zone difference and the time discrepancy.
• CDR time. The CDR date range will be exactly as chosen, from 2012-01-29 09:00:00 to 2012-01-29 10:59:59. (Which will correspond to something like 4am-6am EST). You need to, however, manually recalculate your local time into the CDR time zone.

Time interval for which data is requested. All intervals can be chosen in Interval strip. 30m in intervals stands for 30 minutes, 12h for 12 hours, and 7d for 7 days.

If the interval is a single value, like 1d, it is measured from the interval value in the past to current time (one day in this example). If the interval is a range - it gives the time span within the range boundaries, so 31-60d means from beginning of 60th day from now to end of 31st day from now (effectively skipping the most recent 30 days).

There are also a couple of special intervals like Cust (see Custom interval, Zoom) or ThisH, PrevH for This hour, Previous hour in Tables and Reports.

Ctrl-clicking the Chart canvas interval of less than 2 days will change the interval for all visible Chart modules of all Blocks on the screen. This is especially valuable when working with lots of charts in the Thumbnail mode.

Provides a way to choose the Interval for data retrieved in various modules, for instance: Chart, Report, CDR, Signaling collector, Media collector. In Charts and Tables the data shown previously is not discarded, but kept in cache, so that you may quickly return to it without querying the DB again. In Charts the Interval with cached data is highlighted in color.

A special Cust option stands for a Custom interval.

The Row limit strip intends to limit the number of rows returned by the system to only the top X rows in the Report, Alerts, CDR, Signaling collector and Media collector modules. You may choose to fetch 10, 100, 300, 1000, 3000, or 10000 rows depending on the module.

Which rows are returned first is determined by the column to which the sort is applied. In the Signaling logs and Media logs module you are limited to sort only by the Capture time column before data request. This limitation is made because the Signaling logs and Media logs table contents a huge amount of data and its sorting during the request can lead to overloading the DB. You are not restricted to sort by any column after the data request because this is performed right in the Web browser only with the fetched rows.

Shows current number of visible rows at the top of tables in the Table, Report, Alerts and other modules. If some rows are filtered out, Row count indicator will adjust accordingly.

In the CDR, CDRbill, Calls, Signaling collector and Media collector modules the row count is also showing total number of records found in the DB, and allows to choose which records to fetch next time. See Row count strip.

Time in the web interface in the Chart, Report, and Config modules is always local time of the user's PC. Time in the CDR module is taken as it was recorded in a VoIP switch DB or dump files.

In the 5gVision DB stats are stored with UTC timestamps. EMA stats are using 5gVision server time, and Per-hour stats are using CDR time, taken from a switch DB, and converted to UTC. The reason for this is that Per-hour stats are rotated every hour, and each hour will have info from the CDRs with this hour's "cdr_date" field, even if the 5gVision server time and the switch DB time have some difference.

The Chart module will not show the last hour of Per-hour stats (the last bar in a sequence) for the first 3 min in a new hour (by user time), because there would be not enough data to gather realistic stats (like ACD), and in order to compensate for any possible differences in time between servers, and delays in CDR processing (CDRs are processed once a minute with some offset from the current time to allow CDRs to be fully written to the Oracle DB, this delay is usually 30-90 seconds depending on the system load).

Stands for Absolute alerts, when the absolute value of a parameter is compared to the threshold. ABS alerts can be current (for stats presented as lines in charts) and Per-hour (for stats presented as bars). See more in Alerts and the configuration section Alerts ABS. Also see DIFF.

Stands for Differential alerts, when the difference between the previous and the current value of a parameter is compared to the delta threshold in %%. DIFF alerts can be current and Per-hour. See more in Alerts and the configuration section Alerts DIFF. Also see ABS.

This parameter calculates a percentage of calls with media losses more than 5%. If a call had duration, but no media in one of the directions, it is also considered as the one with media losses.

Names in table columns: In Loss, Out Loss.

There are 2 parameters for % of Codec conversions - one to show % of all conversion happening in the switch, and another one to show only the heavy conversions, when codecs are converted twice with G.711 at the intermediary stage (for example G.729 to G.723 conversion).

Names in table columns: In Conv, Out Conv, In Heavy, Out Heavy.

An API was developed to get 5gVision charts and other data from any external application by sending a URL with requested objects and parameters to a 5gVision server. Please contact support if you need more info. This feature is available for an additional fee.

Normally, the areas one sees in the interface are destination number areas with Area object IDs starting with a. This module allows to additionally calculate statistics for the source number areas, whose IDs will start with b. Object combinations, like Customer->Area, are not calculated for the source areas.

For SRC area statistics ba, bc, bca combinations are used.