icostan / yawndb

YAWNDB is an in-memory circular array database written in Erlang.

YAWNDB

YAWNDB is an in-memory circular array database written in Erlang. This language is based upon the lightweight process model allowing to process large amounts of data without consuming extra system resources. In YAWNDB the data are stored in a circular buffer named a conveyor in YAWNDB terminology. All the incoming data are dispersed and put into the archives named buckets according to certain rules. A rule is a set of properties for a set of statistical data (i.e. data size, data collection frequency, etc.)

The data in YAWNDB are represented as triplets composed of time, value and key. The key (in YAWNDB terminology it is also called path) is a sequence of lowercase letters and numeric symbols defining on which conveyor the triplet will be put.

Our approach allows to gain the following benefits:

• the obsolete data are deleted without excessive consumption of resources;
• fixed memory usage for a fixed number of keys;
• fixed access time for any records.

Architecture

YAWNDB is based on a circular array algorithm implemented in Ecirca library written in C. YAWNDB modules written in Erlang interact with Ecirca via native implemented functions (NIF). Data retention is provided by Bitcask application written in Erlang. The REST API is based on Cowboy web server. The data is written via the socket and read via the REST API interface.

Installation

Clone the repository:

$ git clone https://github.com/selectel/yawndb.git

and then execute the following commands:

$ cd yawndb
$ make all

then

$ ./start.sh

Copy the configuration file:

cp priv/yawndb.yml.example priv/yawndb.yml

or create an appropriate symlink.

Configuration

All YAWNDB settings are kept in the configuration file yawndb.yml.

Data Saving Settings

Parameter	Type	Description
flush_enabled	boolean	Enabling/disabling saving to disk
flush_dir	string	Directory to save data
flush_period	integer	Data saving frequency (in seconds)
flush_late_threshold	float	Write to log if saving takes more than flush_period * flush_late_threshold seconds
flush_cache	integer	Read cache
flush_write_buffer	integer	Write cache
flush_block_size	integer	Block size for saving to disk

Obsolete data deletion settings

Parameter	Type	Description
cleanup_enabled	boolean	Enable/disable obsolete data deletion
cleanup_period	integer	Deletion frequency, in seconds

User API settings

Parameter	Type	Description
listen_jsonapi_req	boolean	Enable/disable the user API
jsonapi_iface	string	Specify which network interface that will be used
jsonapi_port	integer	Specify which local port number to use
jsonapi_prespawned	integer	Number of workers previously created (can be increased)

Administrative API settings

Parameter	Type	Description
listen_tcpapi_req	boolean	Enabling/disabling the administrator API
tcpapi_iface	string	Specify which network interface that will be used
tcpapi_port	integer	Specify which local port number to use
tcpapi_prespawned	integer	Number of workers previously created (can be increased)

Data storing settings

Parameter	Type	Description
max_slice	integer	Maximum size of the slice to choose
rules	list	Rules regulating data retention

Data storing rules

Parameter	Type	Description
name	string	The name of the rule
prefix	string	The prefix referred by the rule
limit	integer	Conveyor size
split	string	Define the way the value will be divided in case it does not match the bucket: proportional - the value will be proportionally divided between the two nearest buckets; equal - the value will be divided into halves between the two nearest buckets; forward/backward - the whole value wil be put into the previous (or the nearest) bucket
type	string	Define the value to be saved in buckets: last - only the last value will be saved; max - the maximal incoming value will be saved; min - the minimal incoming value will be saved; sum - the sum of the incoming values will be saved; avg - the average value will be saved
value_size	integer	Define the value size for a bucket (16, 32 и 64 bits), possible values: small, medium, large
additional_values	list	A list of "foobar: weak/strong" pairs from 0 to 14 symbols in length. Sometimes it is necessary to save the additional values not being numbers ('timeout', for example). The list of additional values is used for this purpose. The values can be either strong or weak. The weak values are updated if the bucket is updated while the strong values are never updated.
autoremove	boolean	Enable/disable conveyor removal in case of data deterioration

Get data API

The data are captured with the HTTP API. There are two types of API in YAWNDB

• the administrative API and the user API. All API settings are specified in the configuration file in the sections Admin JSON API and User JSON API.

A typical API response is given in the following form:

{
    "status": "ok",
    "code": "ok",
    "answer": {}
}

The status field may take the values ok or error. The code field contains "ok" on success. The answer is displayed in the answer field. If an error occurs, its code will be displayed in the code field; the human-readable description of the error will be displayed in the answer field.

Get a list of rules for a selected path

GET /paths/:path/rules

Example of an answer:

{
    "status": "ok",
    "code": "ok",
    "answer": ["stat"]
}

Get a list of time-value pairs for a given period:

POST /paths/slice

The conveyors are indicated in the body of the request: paths=path1/rule,path2/rule2,path3/rule3.

Request parameters:

Parameter	Type	Description
from	integer	Initial time, UTC timestamp
to	integer	End time, UTC timestamp

Example of an answer:

{
 "status": "ok",
 "code": "ok",
 "answer": [
    [63555098820, 10, 12],
    [63555098888, 16, "empty"],
    [63555098940, 12, 10],
    [63555099000, 3, 8],
    [63555099060, 10, 12],
    [63555099120, 2, 9],
    [63555099180, 7, 12],
    [63555099240, 3, 4],
    [63555099300, 20, 10],
    [63555099360, 2, 11]
 ]
}

Get a list of the last time-value pairs for a given period:

GET /paths/:path/:rule/slice

Request parameters:

Parameter	Type	Description
from	integer	Initial time UTC timestamp
to	integer	End time, UTC timestamp

Example of an answer:

{
 "status": "ok",
 "code": "ok",
 "answer": [
    [63555098820, 12],
    [63555098880, "empty"],
    [63555098940, 10],
    [63555099000, "empty"],
    [63555099060, 15],
    [63555099120, "empty"],
    [63555099180, 12],
    [63555099240, "empty"],
    [63555099300, 10],
    [63555099360, "empty"]
 ]
}

The answer contains a list of time-value pairs for a given period; if there is no value for a certain moment, the string empty is returned.

Get a list of the last time-value pairs for a conveyor:

Request parameters:

Parameter	Type	Description
n	integer	The number of last values

Example of an answer:

{
 "status": "ok",
 "code": "ok",
 "answer": [
    [63555103140, 11],
    [63555103200, 10],
    [63555103260, 10],
    [63555103320, 12],
    [63555103380, 10],
    [63555103440, 10],
    [63555103500, 11],
    [63555103560, 10],
    [63555103620, 10],
    [63555103680, 14]
 ]
}

The answer contains a list of time-value pairs for a given period; if there is no value for a certain moment, the string empty is returned.

Administrative API

Get a server status:

GET /status

Example of an answer:

{
 "status": "ok",
 "code": "ok",
 "answer": {
    "read_rpm": 37,
    "write_rpm": 816042,
    "read_rps": 1,
    "write_rps": 13601,
    "processes_now": 162836,
    "processes_max": 300000,
    "paths_count": 162256,
    "dirty_paths_count": 19864
 }
}

Answer fields:

Field	Type	Description
read_rpm	integer	Read requests per minute
write_rpm	integer	Write requests per minute
read_rps	integer	Read requests per second
write_rps	integer	Write requests per second
processes_now	integer	Number of active processes
processes_max	integer	Maximal number of active processes permitted
paths_count	integer	Number of paths
dirty_paths_count	integer	Number of the unsaved paths

Get the list of all paths saved

GET /paths/all

Example of an answer:

{
 "status": "ok",
 "code": "ok",
 "answer": [
    "some-stats-a",
    "some-stats-b"
 ]
}

Delete a path

DELETE /paths/:path

Example of an answer:

{
 "status": "ok",
 "code": "ok",
 "answer": "deleted"
}

Get an aggregated list of time-value pairs

POST /aggregate

The paths are indicated in the body of the request: paths=path1,path2,path3.

Parameters of the request

Parameter	Type	Description
from	integer	Initial time, UTC timestamp
to	integer	Ending time, UTC timestamp
rule	string	The rule used
aggregate	string	Aggregation function (max, min, avg, sum)

Example of an answer:

{
 "status": "ok",
 "code": "ok",
 "answer": [
    [63555096900, 12],
    [63555096960, "empty"],
    [63555097020, 10],
    [63555097080, 16],
    [63555097140, 10],
    [63555097200, "empty"],
    [63555097260, 11],
    [63555097320, 17],
    [63555097380,16],
    [63555097440, "empty"]
 ]
}

Error codes

Error	HTTP-code	Description
no_aggregate	400	The aggregate request parameter is not indicated or improperly formulated
no_path	400	The path request parameter is not indicated or improperly formulated
no_rule	400	The rule request parameter is not indicated or improperly formulated
no_from	400	The from request parameter is not indicated or improperly formulated
no_to	400	The to request parameter is not indicated or improperly formulated
no_n	400	The n request parameter is not indicated or improperly formulated
no_paths	400	The paths request parameter is not indicated or improperly formulated
no_body	400	The request body is expected but absent
no_fun	400	Unknown request
from_to_order	400	The value of the from parameter is greater than that of the to parameter
page_not_found	404	Requested path not found
rule_not_found	404	The requested rule for the given path not found
slice_too_big	413	The number of request results is greater than the value of the max_slice parameter
process_limit	500	Limit of active processes achieved
memory_limit	500	Write limit achieved
internal	500	Internal error

Save data API

This API is used for saving new data to YAWNDB. To save the data connect to the TCP port indicated in the configuration file (TCP API options section) and send the packets in the indicated format. No reply from the server is expected.

The main types of data used

Type	Description
uint8	Unsigned 8-bit number
uint16	Unsigned 16-bit number in big-endian
uint64	Unsigned 64-bit number in big-endian
byte*	Byte sequence

Packet structure

Field	Value
uint16	Packet size (except this field)
uint8	Protocole version
uint8	Flag indicating special values (0 - the value is not special)
unit64	Number of seconds from the beginning of UNIX era
unit64	Value (or the number of the special value)
byte*	Byte sequence

Constructing a packet in Python

def encode_yawndb_packet(is_special, path, time, value):
    """
    Construct a packet to be sent to yawndb.

    :param bool is_special: special value? used for additional_values
    :param str path: statistics identifier
    :param int value: statistics value
    """
    is_special_int = 1 if is_special else 0
    pck_tail = struct.pack(">BBQQ", YAWNDB_PROTOCOL_VERSION, is_special_int,
                           time, value) + path
    pck_head = struct.pack(">H", len(pck_tail))
    return pck_head + pck_tail

Constructing a packet in C

// indicate the protocol version
#define YAWNDB_PROTOCOL_VERSION 3

// describe the packet structure
struct yawndb_packet_struct {
    uint16_t    length;
    uint8_t     version;
    int8_t      isSpecial;
    uint64_t    timestamp;
    uint64_t    value;
    char        path[];
};

// construct the packet
yawndb_packet_struct *encode_yawndb_packet(int8_t isSpecial,
    uint64_t timestamp, uint64_t value, const char * path)
{
    yawndb_packet_struct *packet;
    uint16_t length;

    length = sizeof(uint8_t) + sizeof(int8_t) + sizeof(uint64_t) + sizeof(uint64_t) + strlen(path);
    packet = malloc(length + sizeof(uint16_t));
    packet->length = htobe16(length);
    packet->version = YAWNDB_PROTOCOL_VERSION;
    packet->isSpecial = isSpecial;
    packet->timestamp = htobe64(timestamp);
    packet->value = htobe64(value);
    strncpy(packet->path, path, strlen(path));

    return packet;
}

Authors

YAWNDB was written by Dmitry Groshev in 2012. Alexander Neganov, Kagami Hiiragi, Maxim Mitroshin, Pavel Abalikhin and Fedor Gogolev have also contributed to the project.