The module is designed to collect and store Neflow and Clickstream data. Data is used to analyze QoE in DPIUI2.

Data from the VAS Experts DPI is received on several sockets (tcp or udp) using utility designed to collect IPFIX stream data.

The collected data is stored in the ClickHouse database.

For the subsystem, you can use hardware or virtual machines with the characteristics listed below:

1. Processor (CPU) 2.5 ГГц - 1 pce
2. RAM - at least 16 GB
3. Hard disk drive (SSD is highly desirable) - at least 500 GB
4. Operating system - Cent OS 6.4 - 7.x+
5. Network interface (NIC) - at least 1 Gbps

10 Gbps average daily traffic generates approximately 25 GB of data per hour in the QoE Store

SSE 4.2 instruction set support is required.

Choose processors with a large number of cores. Clock frequency is less important. For example, 16 cores with 2,600 MHz are better than 8 cores at 3,600 MHz.

RAM should be no less than the amount of data requested.

The more RAM server has, the better performance will be achieved when building reports.

The more memory, the less disk system is stressing.

Minimum prerequisites is 16 GB.

Always disable the swap file.

Required disk space is at least 16 GB for every storage day, it's actual value depends on daily traffic. It is estimated that 10 Gbit/s of average daily traffic generates approximately 25 GB of data per hour in the QoE Store.

If your budget makes you possible to use SSD, use it. Otherwise use HDD. SATA HDDs 7200 RPM will be suitable.

When using HDD, you can combine them into RAID-10, RAID-5, RAID-6 or RAID-50.

Most of the data is stored in the /var/lib/clickhouse/ directory. You can mount a drive/partition on this directory.

Temporary data (ipfix dumps) is stored in the /var/qoestor/backend/dump/ directory. 50 GB be sufficient for this.

You could familiarize yourself with the contents of maintenance advice from Yandex ClickHouse by following the https://clickhouse.yandex/docs/ru/operations/tips/ link.

• Bug concerning incorrect recognition of trunk switches is fixed
• protocols_dic dictionary is updated

• ipfix re-export feature is added and is available through the ipfixreceiver configuration file: IPFIX_FULLFLOW_EXPORT and IPFIX_CLICKSTREAM_EXPORT

• Bugs concerning the work of the subnets_local_dic dictionary (such as "A call to function range would produce 12884901882 array elements")
• ipfixreceiver2 receiver configuration has been fixed (FileWriter queue is full. Records dropped.)

• Host Category Dictionaries
• Transition to the ipfixreceiver2

• Preaggregation is implemented resulting in reducing the netflow by 6-7 times and clickstream by 3 times
• The following dictionaries are implemented: subscribers, switches, autonomous systems (AS), crc
• The following features are added: identification the traffic direction and subscribers filtering (separation of IP hosts and IP subscribers) by AS and CIDR. This option is meaningful only if the VAS Experts DPI is implemented using the mirror connection scheme.

This QoE Stor version works with the version of DPIUI2-2.1.5 and higher

clickhouse-client --query="drop database qoestor"

• A new QoE Stor module is implemented

For automatically installation or upgrading follow these steps:

1. Execute clickhouse-repos_config.sh.gz script. The repository will be prepared for the subsequent installation of the clickhouse database as a result.
2. Execute fastor-rpm_install.sh.gz script. It will cause the installation of the following packages: ipfixreceiver, clickhouse, fastor. All of them will be automatically configured according to the defaults.
3. Check whether the qoestor database in clickhouse is available. To do this, issue the command

   clickhouse-client --query="show databases" | grep qoestor

4. If there is no database (probably the database server is not running), you should to create it by issuing the following command

   clickhouse-client -n < /var/qoestor/backend/etc/db/qoestor.sql

Upgrading is performed using the same scripts as in the installation section.

If receivers stopped after you have executed the

yum -y update

command, pleas refer to the troubleshooting section following the link.

ipfix receivers configuration is implemented through the .env file:

/var/qoestor/backend/.env

Default configuration looks like:

#Ipfix form DPI 0
IPFIX_FULLFLOW_PORT_TYPE[0]=tcp
IPFIX_FULLFLOW_PORT[0]=1500
#IPFIX_FULLFLOW_EXPORT[0]=10.0.0.2/9920/tcp,10.0.0.3/3440/udp

IPFIX_CLICKSTREAM_PORT_TYPE[0]=tcp
IPFIX_CLICKSTREAM_PORT[0]=1501
#IPFIX_CLICKSTREAM_EXPORT[0]=10.0.0.2/9921/tcp,10.0.0.3/3441/udp

#Traffic direction definition
# 0 - as is
# 1 - by AS (for fullflow only)
# 2 - by CIDR (for fullflow and clickstream)
# 3 - by both: AS and CIDR
# 4 - any: AS or CIDR
TRAFFIC_DIR_DEF_MODE=0

#Subscriber filter
# 0 - no filter
# 1 - by AS (for fullflow only)
# 2 - by CIDR (for fullflow and clickstream)
# 3 - by both: AS and CIDR
# 4 - any: AS or CIDR
SUBSCRIBER_FILTER_MODE=0

In the configuration above fullflow and clickstream receivers are listening 1500 and 1501 socket respectively. «0» in array subscript means that the receiver get the data from DPI number 0.

The values TRAFFIC_DIR_DEF_MODE = 0 and SUBSCRIBER_FILTER_MODE = 0 mean that there is no need to calculate the traffic direction and apply filters to subscribers.

The IPFIX_FULLFLOW_EXPORT and IPFIX_CLICKSTREAM_EXPORT parameters allow you to configure export to third-party receivers. Format for use: ip/port/proto[,ip/port/proto.

The following example shows how to configure data reception from several DPIs

#Ipfix form DPI 0
IPFIX_FULLFLOW_PORT_TYPE[0]=tcp
IPFIX_FULLFLOW_PORT[0]=1500

IPFIX_CLICKSTREAM_PORT_TYPE[0]=tcp
IPFIX_CLICKSTREAM_PORT[0]=1501

#Ipfix form DPI 1
IPFIX_FULLFLOW_PORT_TYPE[1]=tcp
IPFIX_FULLFLOW_PORT[1]=1510

IPFIX_CLICKSTREAM_PORT_TYPE[1]=tcp
IPFIX_CLICKSTREAM_PORT[1]=1511

#Ipfix form DPI 2
IPFIX_FULLFLOW_PORT_TYPE[2]=tcp
IPFIX_FULLFLOW_PORT[2]=1520

IPFIX_CLICKSTREAM_PORT_TYPE[2]=tcp
IPFIX_CLICKSTREAM_PORT[2]=1521

The following example corresponds to the situation when you need to identify subscribers by CIDR

This configuration makes sense only when the VAS Experts DPI is installed using port mirroring.

TRAFFIC_DIR_DEF_MODE=2
SUBSCRIBER_FILTER_MODE=2

Be sure to configure subnets_local_dic dictionary for this configuration example!

The following example corresponds to the situation when export to third-party receivers is configured

IPFIX_FULLFLOW_PORT_TYPE[0]=tcp
IPFIX_FULLFLOW_PORT[0]=1500
IPFIX_FULLFLOW_EXPORT[0]=10.0.0.2/1600/tcp

IPFIX_CLICKSTREAM_PORT_TYPE[0]=tcp
IPFIX_CLICKSTREAM_PORT[0]=1501
IPFIX_CLICKSTREAM_EXPORT[0]=10.0.0.2/1601/tcp

All receivers can be restarted using the command:

/var/qoestor/backend/qoestor-config.sh

If you need to restart the receivers one by one, you can do this by restarting corresponding systemd service units, for example

• For CentOS 7

  systemctl restart qoestor_fullflow_0.service
  systemctl restart qoestor_clickstream_0.service

• For CentOS 6

  service qoestor_fullflow_0 stop
  service qoestor_clickstream_0 stop
  service qoestor_fullflow_0 start
  service qoestor_clickstream_0 start

• For CentOS 7

  systemctl stop qoestor_fullflow_0.service
  systemctl stop qoestor_clickstream_0.service

• For CentOS 6

  service qoestor_clickstream_0 stop
  service qoestor_fullflow_0 stop

• Stop

  sudo /etc/init.d/clickhouse-server stop

• Start

  sudo /etc/init.d/clickhouse-server restart

The DPI version must be at least 8.1.

You can configure ipfix export by editing the fastdpi.conf configuration file on your DPI device.

netflow=8
netflow_dev=em1
netflow_timeout=10
netflow_as_direction=3
netflow_full_collector_type=2
netflow_full_port_swap=0
netflow_full_collector=YOUR_QOESTOR_IP:1500
netflow_passive_timeout=20
netflow_active_timeout=60
netflow_rate_limit=120
ipfix_dev=em1
ipfix_tcp_collectors=YOUR_QOESTOR_IP:1501

This can be achieved also using DPIUI2.The dpiui2 version must be at least 2.1.0.

To perform configuration using DPIUI2 you should open the section DPI CONTROL → CONFIGURATION. Open the tab Collection and analysis of statistics on protocols and directions.

Set neflow option to Export of complete statistics for sessions. See figure below.

Then specify socket for fullflow receiver within "netflow_full_collector" parameter: "IP address of the netflow collector with full statistics (netflow_full_collector)". "netflow_full_collector_type" should be set to "Export ipfix to udp header", whereas "netflow_full_port_swap" should be empty or equals to "Keep original port numbers". See the figure below.

Type in the clickstream receiver socket in "ipfix_udp_collectors" parameter under the "DPI CONTROL → CONFIGURATION → Operator's SORM → IP or the domain name (: port) of the ipfix collector clickstream (ipfix_udp_collectors)". See the figure below.

Press Save and then restart fast_dpi. See the figure below.

Open the section Hardware "Management → Equipment → Hardware settings". Type in an identifier within the "Id on ipfix collector" section (under the "Ipfix settings") for the ipfix collector. See the figure below.

In order to browse QoE reports, you should configure the DPIUI2 connection to the QoE Stor. See the Setting up a connection to the QoE Stor section.

All dictionaries are in the /var/qoestor/backend/etc/db/ directory and have .txt extension

Each dictionary has it's own sample.txt which can be used as a pattern.

All columns within the dictionaries are separated by a tab character (\t). The number of \t should be one less than the number of columns in dictionary. Please, pay close attention to this circumstance.

When files are changed, the data is loaded into the database automatically.

Some useful commands when working with dictionaries:

• Speed up data updates in directories

  clickhouse-client --database=qoestor --query="system reload dictionaries"

• Check dictionaries for errors

  clickhouse-client --database=qoestor --query="select * from system.dictionaries"

• Check if there is data in the directory, for example for the subnets_local_dic

  clickhouse-client --database=qoestor --query="select * from subnets_local_dic"

The lists of your local AS and local subnets are specified in these dictionaries. Dictionaries are used to identify the traffic direction (true in case the DPI is installed using mirroring) and to filter subscribers (so that the hosts IP addresses do not appear in the subscriber reports)

Example of asnum_local_dic dictionary

12345	LOCAL
65535	UNKNOWN

The first column is AS number, the second one is it's name (it is displayed in reports).

Example of subnets_local_dic dictionary

192.168.1.0/24	LOCAL
10.64.66.0/24	LOCAL
172.16.0.0	LOCAL
2a02:2168:aaa:bbbb::2	LOCAL

The first column is IP address or CIDR, the second one is the name (it is NOT displayed in reports, but it is required by format).

Dictionary of subscribers.

Dictionary example

10.64.66.100	login	5	port1	unit_vendor	cabel	contract	services	mac
10.64.66.101	login	2	port1	unit_vendor	cabel	contract	services	mac
10.64.66.102	login	3	port1	unit_vendor	cabel	contract	services	mac
10.64.66.103	login	4	port1	unit_vendor	cabel	contract	services	mac
10.64.66.104	login	5	port1	unit_vendor	cabel	contract	services	mac
10.64.66.105	login	5	port2	unit_vendor	cabel	contract	services	mac
10.64.66.106	login	5	port3	unit_vendor	cabel	contract	services	mac

Columns:

1. IP address
2. Login
3. Switch ID (access switch)
4. Switch port
5. Subscriber device vendor
6. Cable
7. Contract
8. Services
9. Subscriber device MAC address (is reserved for future purposes)

Hierarchical dictionary of devices (access switches and trunk switches)

Dictionary example

1	Switch_1	Ethernet	Region_1	Address_1	10.140.1.18	ISP_1	0	0		
2	Switch_2	Ethernet	Region_2	Address_2	10.140.2.18	ISP_1	0	0		
3	Switch_3	Ethernet	Region_3	Address_3	10.140.3.18	ISP_1	0	1	port1	
4	Switch_4	Ethernet	Region_4	Address_4	10.140.4.18	ISP_1	0	3	port1	
5	Switch_5	Ethernet	Region_5	Address_5	10.140.5.18	ISP_1	0	4	port1	

Columns:

1. Device ID UInt64
2. Device name
3. Device type
4. Region
5. Address
6. Switch IP address
7. Internet service provider
8. Indicator: trunk switch indicator (1 if so). Is not currently used, you can set 0 everywhere
9. Upstream Switch ID UInt64
10. Upstream Switch port
11. The owner

CRC Errors Dictionary (on switch ports)

Dictionary example

2	port_1	450
5	port_1	550
5	port_2	500
4	port_1	780

Columns

1. Switch ID
2. Switch port
3. CRC value

Host categories dictionaries. Designed to determine the ownership of a particular host category.

Directories are automatically downloaded from vasexperts.ru resources.

To speed up the initial load, issue the following commands

1. sh /var/qoestor/backend/etc/cron_daily.sh

2. clickhouse-client --database=qoestor --query="system reload dictionaries"

If you have installed and configured everything according to the instructions above, and the DPIUI2 "QoE Analytics" section is empty, below is a checklist of the steps to be taken before contacting our technical support:

1. Check the time and timezone settings on servers with dpiui2 and QoE Stor installed. Try to specify a long period in dpiui2. If it's about the timezone, the data will appear. Set the proper time on the servers with dpiui2 and QoE Stor module installed , restart the corresponding servers (on which the dpiui2 and QoE Stor module are installed).
2. Check if the database is created on the server with QoE Stor installed

   clickhouse-client --query="show databases" | grep qoestor

   If the database is not created, you should create it using the following command

   clickhouse-client -n < /var/qoestor/backend/etc/db/qoestor.sql

3. Check if there is data in the database on the server with QoE Stor installed

   clickhouse-client --query="select count(), min(flow_start_time), max(flow_start_time) from qoestor.fullflow"

   и

   clickhouse-client --query="select count(), min(time), max(time) from qoestor.clickstream"

4. Check the content of the receiver dumps on the server with QoE Stor installed

   /var/qoestor/backend/dump/fullflow

   and

   /var/qoestor/backend/dump/clickstream

5. Check the receivers logs under the

   /var/qoestor/backend/logs/

   directory. Is there something like "oops!”? Please contact technical support, because there is most likely the components are not installed from the proper repositories". If there is "Illegal IPFIX Message Version 0x0005", then once again check the export settings on the dpi: netflow_full_collector_type is specified improperly.

6. Check whether the 1500 and 1501 ports are listening on the server with QoE Stor installed

   netstat -nlp | grep 1500 и netstat -nlp | grep 1501

   Restart all the receivers, just to be safe, by issuing the command:

   /var/qoestor/backend/qoestor-config.sh

7. Check again ipfix export settings on the dpi device
8. Check the GUI connection details for QoE Stor on the server with DPIUI2 installed
9. Check if the ClickHouse database is running on the server with QoE Stor installed by issuing following command:

   ps aux | grep clickhouse

   Make sure the server has enough amount of RAM.

10. Check the clickhouse logs under the /var/log/clickhouse-server/ directory on the server with QoE Stor installed.

If there is a need to drop all data in the database, then you should take the following steps on the server with the QoE Stor installed:

1. Drop the DB by issuing:

   clickhouse-client --query="drop database qoestor"

2. Recreate DB by issuing

   clickhouse-client -n < /var/qoestor/backend/etc/db/qoestor.sql

When running yum –y update some libraries can be broken resulting in receivers are not running. Follow these steps to troubleshoot:

1. Remove fastor and its dependencies

    yum remove fastor ipfixreceiver libfixbuf netsa_silk netsa-python

2. Reinstall it using fastor-rpm_install.sh.gz script.

If necessary, you can create your own reports without additional tools and upload data in any format such as CSV, JSON, TabSeparated.

Data is stored in 4 main logs:

• qoestor.fullflow – full netflow log, storage period is 24 hours
• qoestor.clicksteam – full clickstream log, storage period is 24 hours
• qoestor.fullflow_agg – pre-aggregated neflow log, unlimited storage period
• qoestor.clicksteam_agg – aggregated clickstream log, unlimited storage period

Use the following command format

clickhouse-client --database=qoestor  --query="Your_SQL_query"

Data is uploaded using TabSeparated format by default.

Example. The client asked for a log of connections to a specific host in CSV format.

clickhouse-client --database=qoestor --query="select * from fullflow prewhere flow_start_date = '2018-10-04' where (source_ipv4 = '10.64.66.100' or destination_ipv4 = '10.64.66.100') and host = 'google.com' ORDER BY flow_start_time limit 10 format CSV"

For more information on ClickHouse SQL, follow the link: https://clickhouse.yandex/docs/ru/query_language/select/.