View source for OstDB DEV Foosic

{{TOCright}}

Protocol for foosic audio fingerprint matching related client<->server communication via UDP and TCP.

VERSION 2

Servers:
* dedicated Java matching server daemon TCP on anidb3 (sig) (does not accept client requests directly)
* dedicated Java redirector/load balancer daemon TCP and UDP on anidb2 (main/api) (UDP for single queries, TCP for batch runs)

Clients:
* cronjob on anidb2
* clients on user systems

== General Workflow ==
=== Workflow Draft ===
==== Involved Parties ====
* client (Avdump), locally on user machine
* main server (anidb2), keeps file meta data includes fingerprints
* matching redirector (anidb2), a simple load balancer which redirects all requests to the corresponding matching server(s) (MATCH goes to all, STORE to one), TCP and UDP interface (UDP for external queries, TCP for cron job/batch queries)
* matching server(s) (anidb3/sig server), Java standalone app does all the fingerprint<->fingerprint matching, TCP interface

==== Workflow - Matching ====
===== Synchronous Part =====
* user runs client on his local audio files
* client sends file meta data, including foosic audio fingerprint, to main server via UDP API
** optional intermediate step to speedup client processing: calculate content hash first and generate and submit the fingerprint only if the content hash is unknown to AniDB or no fingerprint is listed on AniDB for that file
* main server does not do any processing on the fingerprint and simply stores it in ostfiletb

===== Asynchronous Part =====
* a cronjob/daemon on the main server regularly checks for newly added (yet unmatched) foosic fingerprints and sends them, together with the ostfile id to the matching redirector via TCP with store set to 1
** flooding is impossible due to the synchronous nature of the TCP matching redirector API
** the cronjob could open multiple TCP connections to the matching redirector if one connection does not yield enough throughput.
* the matching redirector forwards the MATCH query to all matching servers via TCP
** (multiple) TCP connections to all matching servers should be kept alive in-between queries (to prevent TCP connection handshake overhead)
* the matching servers try to match the fingerprint to all potentially interesting fingerprints in their database
** pre-match filtering: via length, avg dom and avg fit
** post-match filtering: via hard cut-off-point for matching confidence value, this will definitely be >=0,50 anything lower would be useless. It will be possible to increase this value without problems at any point in time. However, reducing it would be a problem. We might therefore want to just start with 0,50 or 0,60 and increase it if we feel that it is necessary to reduce the amount of matches returned.
** potentially further internal optimizations, i.e. by identifying group representatives for certain well matching groups of files and only matching taking the representatives into account when matching
* each matching server replies with a list of matching ostfile ids together with the matching confidence per match and some general usage data (for load balancing).
** neither fingerprints nor matching results are stored on the matching servers at this point
** matching servers keep some internal usage statistics for potential future optimization
** if a matching server already stores this exact ofid, it will add a note about this fact to it's reply.
** the matching redirector does not keep any persistent storage. Other than some usage data, nothing is kept here.
* the matching redirector collects the replies from all matching servers and collates them into one reply for _later_ transmission to the main server.
* the matching redirector then decides which matching server should store the new fingerprint based on the usage statistics returned by each matching server and sends that server a STORE command. That servers identifier is then included with the reply to the main server.
** if one of the earlier replies from the matching servers indicated, that a server already has the fingerprint, no new STORE command is issued. The identifier of that server is used instead.
* the matching redirector replies to the main server (matching results + ident of matching server storing that fingerprint)
** for the main server/cron daemon it is not visible which _match_ came from which matching server
* the main server stores the matching data in the DB
** (ofidlow int4, ofidhigh int4, match int2) as ostfilefoosictb
** (...,foosicfp bytea,foosicsrv int2,...) in ostfiletb
* main server uses matching data to support manual file<->song matching via the web interface
** the user will be able select the cut-off-point for the matching value on-the-fly in order to reduce false-positives or increase recall

==== Workflow - Deletion ====
(external clients have no deletion permission)
* an ost file is deleted on the main server
* it is appended to a special deletion queue table
* a cron job processes the table in regular intervals and sends DELETE requests for each deleted ofid to the matching redirector.
** if the storage location of the ofids fingerprint is known to the main server, it will include that info in the DELETE request
* the matching redirector will either forward the DELETE request to the specified matching server or to all matching servers, if none was specified
* the matching servers will delete all local data for the given ofid
* the results are collated by the matching redirector and returned to the main server's cron job

== Possible Extension ==
(for client features; UDP API)
* command to fetch audio meta data by ostfile id or size+content hash
* command to add audio file to MyList by ostfile id or size+content hash
* same commands as available via TCP (used by the anidb2 cronjob) also available via UDP for use by other clients.
** i.e. to allow lookups by foosic fingerprint. For that a client would first contact the UDP API (anidb2) with the content hash and if the content hash is unknown to AniDB it would send the fingerprint to the matching redirector (anidb2), which would delegate to the matching servers (anidb3), to get one or more ostfile id(s) and then use those to query song data from the UDP API.
*** this is not meant for Avdump, but it might be interesting for direct integration into player software, i.e. a winamp/amarok plugin, would work somewhat like the already available musicbrainz plugins.

== General Considerations ==
* it is very important to effectively limit the number of fingerprints which need to be taken into account for each lookup. As such the file length and the average dom and fit should be stored in a way which allows easy and fast filtering via range queries on those 3 dimensions. so that'd probably mean it will be a: (length int4, dom int4, fit int4, fingerprint bytea) kind of table
** the number of fingerprints to consider for each matching lookup could be reduced further by only taking one representative fingerprint for each closely matching group of fingerprints into account.
*** i.e. if there are 20 files for one song and the fingerprints of 18 of those files match with a confidence of NN% (the actual confidence number to use will be hard to decide on, might be something like 98%) then the median of that group (the file which has the closest cumulated match with all other files of that group) could be picked as a representative and all other 17 fingerprints could be skipped during matching lookups. If we always want to return all matches, then the remaining 17 fingerprints could be matched in a second run once the matching with the representative yielded a value above the cut-off-point.
*** depending on the number of encodes per song and the closeness of their match such an optimization might well reduce the number of fingerprints to consider per lookup by a factor of 20 or more.
*** possible storage: additional grouprep int4 field which stores ofid of group representative if an entry is part of a group. Group representatives and fingerprints not belonging to any group would have a value of 0. The initial matching lookup could simply restrict the SELECT to WHERE grouprep=0 (in addition to the len, dom and fit constraints).
* as further optimization may become necessary someday, we should collect some usage statistics per fingerprint in order to identify hotspots and areas of very low interest. Data to collect could be:
** added - the date this fingerprint was added
** bestmatch - the best match confidence this fingerprint ever achieved for a query (ignoring a self compare)
** lastmatchdate - date of last match of this fingerprint for a query
** matchedcnt - the number of times this fingerprint was returned as a match for a query
** as the foosic server would require no authentication, the same user sending a fingerprint multiple times would increase the counter every time
* it may also become necessary to split the processing over multiple servers someday. this can be greatly simplified if the protocol is designed in a way which would allow the following setup.
** loadbalancing matching redirector listens for foosic fingerprint lookups each received lookup is send to _all_ matching servers
** each matching server replies with ostfile ids and match confidence or with "unknown"
** matching redirector merges all ostfile id replies together (sorted by match confidence) and returns reply to client.
** if none of the matching servers indicated that it has the exact fingerprint in his local storage, the matching redirector tells a matching server with free resources to store the fingerprint.
*** the decision is made based on the observed fingerprint distribution of the matching servers. The MATCH reply from the matching servers lists the number of fingerprints which needed to be taken into account for the specific match. The server with the smallest number of fingerprints with similar length, avg. fit and avg. dom would be the best place to store the new fingerprint. Other factors could also be taken into account.
* this would mean that each query is processed in parallel by all available matching servers. The very nature of the search approach makes the entire approach very scalable.

== Protocol ==
=== Broken Clients ===
* we probably want to require a client string and client version in every query (similar to UDP API) to be able to ban badly broken clients, should the need arise someday.

=== Protocol Draft ===
Every query should contain as additional parameters:
 client={str client name}&clientver={int client version}

Access to the STORE command will be limited to the main server's cron job by this method.

==== Querying a foosic fingerprint (don't add if unknown) ====
Used by:
* external clients 
* main server's cron job
* matching redirector (forwarded)

Client:
 MATCH ofid={int4 ostfile id}&foosic={str ascii hex representation of fingerprint}[&store=1]
: The <tt>store=1</tt> parameter is filtered out and interpreted by the matching redirector, only the main server's cron job is allowed to set <tt>store=1</tt>.

Server Reply:
* Matchings found
: 200 MATCHED
: {int2 ident of matching server this fingerprint is stored on}
:* this line is inserted by the matching redirector, the data is only interesting for the main server's cron job. Done only if store=1.
: {int result count}|{int compare count}|{int time taken in ms}
:* this line will be suppressed by the matching redirector which processes it to decide where to store a new fingerprint (load balancing)
: ({int error}|{int ofid}\n)*

The reply starts with
: 201 MATCHED LOCAL
if the ofid being queried is stored on that specific matching server
:* the matching redirector will filter 201 replies, a normal client will never see them.

* No matchings found
: 300 UNMATCHED
: {int2 ident of matching server this fingerprint is stored on}
:* this line is inserted by the matching redirector, the data is only interesting for the main server's cron job. Done only if store=1.
: {int result count}|{int compare count}|{int time taken in ms}
:* this line will be suppressed by the matching redirector which processes it to decide where to store a new fingerprint (load balancing)

The reply starts with
: 301 UNMATCHED LOCAL
if the ofid being queried is stored on that specific matching server
:* the matching redirector will filter 201 replies, a normal client will never see them.

==== Submitting a new foosic fingerprint ====
Used by:
* matching redirector, access is restricted

Client:
 STORE ofid={int4 ostfile id}&foosic={str ascii hex representation of fingerprint}

Server Reply:
* Fingerprint was not yet in DB
: 210 STORED

* Fingerprint was already in DB
: 310 ALREADY STORED

==== Submitting a new foosic fingerprint ====
Used by:
* main server's cron job, access is restricted
* matching redirector (forwarded), access is restricted

Client:
 DELETE ofid={int4 ostfile id}[&{int2 ident of matching server this fingerprint is stored on}]

Server Reply:
* Fingerprint was in DB
: 220 DELETED

* Fingerprint was not in DB
: 320 NOT FOUND

==== Query the current server load/utilization ====
Used by:
* matching redirector, access is restricted

Client:
 LOADSTAT

Server Reply:
: 299 LOAD STAT
: {int2 load factor}|{int4 number of fingerprints in db}|{int2 system load}
:* load factor: a simply multiplicative constant which is used to distinguish between fast and slow server hardware. This can i.e. be used to store twice as many fingerprints on one server compared to others. The fingerprint count is converted according to the following formula prior to comparison/selection of least used server.
::* relative fingerprint number/load = number of fingerprints in db * (load factor / 100)

[[Category:Development]]