Edward J. Yoon wrote: >> Is access to *one* of the 20 million different SQLite files getting >> progressively slower? How big is that specific SQLite file? Is that >> the one that is "huge"? I use SQLite over an NAS (at times), and never >> experience any noticeable slowdown. Is access to his NAS itself slow, >> perhaps not just via SQLite but just over the regular filesystem? >> > > Each NAS_000 ~ N storages have approximately 300,000 files, the > average size of file is few MB (not over GB). The broker servers (with > SQLite library) are on the NAS and The front-end web servers (more > than 200 servers) communicate with living broker servers after request > location from location addressing system. There are high frequency > read/write/delete operations. > > The number of files/storages/clients keep increasing little by little. > How is the NAS attached to the network? Is it a USB attached drive to a server?
Many network bandwidth issues aren't caused by the DB it's self but rather the network architecture. A simple rule of thumb is that the bandwidth is only as large as the largest bandwidth of an appliance/component that the data travels through. Some NAS appliances are notoriously slow due to the restrictions on bandwidth within that appliance. And you won't notice it until you get to a specific point, which maybe the number of concurrent users or the amount of date transfered or a combination of both. If you have the NAS attached to a server you could setup a routine that would copy the desired db file to the server (or another server on the network) while the user is using that db file. That would give you more of a distributed file system architecture than just trying to serve everything off of the NAS and would in turn take pressure off of the NAS appliance. I would be curious to know at what point this became an issue? Other factors would be the type/make/model of the NAS and how it's setup on the network? Are you using hubs or switches? Have you had a traffic monitor on the network? Is this network MS based or Unix, do you have a PDC, and how is the NAS authenticating users? Are the connections mapped drives? What I would suggest doing is stepping back to the point of the performance degradation and work from there. It maybe that you have just reached the outer limits of that particular NAS appliance. Personally, I do not use NAS for DB's that have concurrent users using the DB's but rather use NAS for archiving and occasional user storage of non-current data for the reasons I stated above. If NAS is the only solution I have, I'll plan to expand the NAS with another appliance when I notice any degradation of performance occurring and then do load balancing to balance the load between the appliances. From what I have gathered here, I think I'd be safe to assume that this isn't a SQLite issue but rather a Network/NAS issue. Hope this helps, Carl ---- Carl Lindgren C. R. Lindgren Consulting / Business on the Desktop > /Edward > > On Wed, Jan 7, 2009 at 9:57 PM, P Kishor <punk.k...@gmail.com> wrote: > >> On 1/7/09, Thomas Briggs <t...@briggs.cx> wrote: >> >>> I actually thought the original question was perfectly clear. I >>> thought the proposed solution (included in the original post) was >>> perfectly logical too. So what's all the fuss? >>> >> The confusion, at least for me, arose from the following sentence in the OP >> -- >> >> "I'm using SQLite, all data (very huge and 20 million files) " >> >> and the response to request for clarification of the above. >> >> - we know he is using SQLite >> >> - we know "it" is all data (although, I am not sure what else could >> SQLite be used for other than "data") >> >> - we know "it" is very huge >> >> - we know there are 20 million *files* involved >> >> No matter how I put together the above four pieces of information, I >> can't grok it. >> >> Is access to *one* of the 20 million different SQLite files getting >> progressively slower? How big is that specific SQLite file? Is that >> the one that is "huge"? I use SQLite over an NAS (at times), and never >> experience any noticeable slowdown. Is access to his NAS itself slow, >> perhaps not just via SQLite but just over the regular filesystem? >> >> So there... no fuss, just a desire to understand better what exactly >> is the problem. >> >> >>> On Wed, Jan 7, 2009 at 7:28 AM, P Kishor <punk.k...@gmail.com> wrote: >>> > On 1/6/09, Edward J. Yoon <edwardy...@apache.org> wrote: >>> >> Thanks, >>> >> >>> >> In more detail, SQLite used for user-based applications (20 million is >>> >> the size of app-users). and MySQL used for user location (file path on >>> >> NAS) addressing. >>> > >>> > Edward, >>> > >>> > At least I still don't understand why you have 20 million databases. >>> > My suspicion is that something is getting lost in the translation >>> > above, and neither you nor anyone on the list is benefitting from it. >>> > Could you please make a little more effort at explaining what exactly >>> > is your problem -- it well might be an "xy problem." >>> > >>> > If you really do have 20 million SQLite databases on a NAS, and you >>> > don't care about changing anything about the situation except for >>> > improving the speed of access from that NAS, well, since you will >>> > likely be accessing only one db at a time, perhaps you could copy that >>> > specific db to a local drive before opening it. >>> > >>> > In any case, something tells me that you will get better mileage if >>> > you construct a good question for the list with enough background >>> > detail. >>> > >>> > >>> >> >>> >> >>> >> On Wed, Jan 7, 2009 at 1:31 PM, P Kishor <punk.k...@gmail.com> wrote: >>> >> > On 1/6/09, Edward J. Yoon <edwardy...@apache.org> wrote: >>> >> >> > Do you have 20 million sqlite databases? >>> >> >> >>> >> >> >>> >> >> Yes. >>> >> > >>> >> > Since all these databases are just files, you should stuff them into >>> a >>> >> > Postgres database, then write an application that extracts the >>> >> > specific row from the pg database with 20 mil rows giving you your >>> >> > specific SQLite database on which you can do your final db work. >>> >> > >>> >> > Seriously, you need to rethink 20 mil databases as they defeat the >>> >> > very purpose of having a database. >>> >> > >>> >> > >>> >> >> >>> >> >> >>> >> >> On Wed, Jan 7, 2009 at 12:36 PM, Jim Dodgen <j...@dodgen.us> wrote: >>> >> >> > I think the question was about the structure of your data >>> >> >> > >>> >> >> > a sqlite database is a file and can contain many tables. tables >>> can contain >>> >> >> > many rows. >>> >> >> > >>> >> >> > Do you have 20 million sqlite databases? >>> >> >> > >>> >> >> > This information can help people formulate an answer. >>> >> >> > >>> >> >> > On Tue, Jan 6, 2009 at 6:14 PM, Edward J. Yoon >>> <edwardy...@apache.org>wrote: >>> >> >> > >>> >> >> >> Thanks for your reply. >>> >> >> >> >>> >> >> >> > That's a lot of files. Or did you mean rows? >>> >> >> >> > Are you sure? There can be many other reasons. >>> >> >> >> >>> >> >> >> There is a lot of files. So, I don't know exactly why at this >>> time, >>> >> >> >> But thought network latency can´t be denied. >>> >> >> >> >>> >> >> >> /Edward >>> >> >> >> >>> >> >> >> On Wed, Jan 7, 2009 at 4:07 AM, Kees Nuyt <k.n...@zonnet.nl> >>> wrote: >>> >> >> >> > On Tue, 6 Jan 2009 11:23:29 +0900, "Edward J. Yoon" >>> >> >> >> > <edwardy...@apache.org> wrote in General Discussion of >>> >> >> >> > SQLite Database <sqlite-users@sqlite.org>: >>> >> >> >> > >>> >> >> >> >> Hi, I'm newbie in here. >>> >> >> >> >> >>> >> >> >> >> I'm using SQLite, all data (very huge and 20 million files) >>> >> >> >> > >>> >> >> >> > That's a lot of files. Or did you mean rows? >>> >> >> >> > >>> >> >> >> >> stored on NAS storage. Lately my system has been getting >>> >> >> >> >> progressively slower. Network cost seems too large. >>> >> >> >> > >>> >> >> >> > Are you sure? There can be many other reasons. >>> >> >> >> > >>> >> >> >> >> To improve its performance, I'm think about local lock file >>> >> >> >> >> instead of NAS as describe below. >>> >> >> >> >> >>> >> >> >> >> char str[1024] = "/tmp"; >>> >> >> >> >> strcat(str, lockfile); >>> >> >> >> >> sprintf(str, "%s-lock", zFilename); >>> >> >> >> >> >>> >> >> >> >> But, I'm not sure this is good idea. >>> >> >> >> >> I would love to hear your advice!! >>> >> >> >> > >>> >> >> >> > I think that's not the right way to start. >>> >> >> >> > This is what I would do, more or less in >>> >> >> >> > this order: >>> >> >> >> > >>> >> >> >> > 1- Optimize the physical database properties >>> >> >> >> > PRAGMA page_size (read the docss first!) >>> >> >> >> > PRAGMA [default_]cache_size >>> >> >> >> > >>> >> >> >> > 2- Optimize SQL: use transactions >>> >> >> >> > where appropriate. >>> >> >> >> > >>> >> >> >> > 3- Optimize your code. Don't close database >>> >> >> >> > connections if they can be reused. >>> >> >> >> > >>> >> >> >> > 4- Optimize the schema: create indexes that >>> >> >> >> > help, leave out indexes that don't help. >>> >> >> >> > >>> >> >> >> > 5- Investigate the communication to/from NAS. >>> >> >> >> > Do all NIC's train at the highest possible speed? >>> >> >> >> > Some limiting switch or router in between? >>> >> >> >> > Do you allow jumbo frames? >>> >> >> >> > >>> >> >> >> > 6- Consider SAN/fSCSI, direct attached storage. >>> >> >> >> > >>> >> >> >> > 7- Consider changing SQLite code. >>> >> >> >> > >>> >> >> >> > >>> >> >> >> > Without more details on your use case, people will only get >>> >> >> >> > general advice like the above. >>> >> >> >> > >>> >> >> >> >>Thanks. >>> >> >> >> > >>> >> >> >> > Hope this helps. >>> >> >> >> > -- >>> >> >> >> > ( Kees Nuyt >>> >> >> >> > ) >>> >> >> >> > c[_] >>> >> >> >> > _______________________________________________ >>> >> >> >> > sqlite-users mailing list >>> >> >> >> > sqlite-users@sqlite.org >>> >> >> >> > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >>> >> >> >> > >>> >> >> >> >>> >> >> >> >>> >> >> >> >>> >> >> >> -- >>> >> >> >> Best Regards, Edward J. Yoon @ NHN, corp. >>> >> >> >> edwardy...@apache.org >>> >> >> >> http://blog.udanax.org >>> >> >> >> _______________________________________________ >>> >> >> >> sqlite-users mailing list >>> >> >> >> sqlite-users@sqlite.org >>> >> >> >> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >>> >> >> >> >>> >> >> > >>> >> >> > >>> >> >> > >>> >> >> > -- >>> >> >> > Jim Dodgen >>> >> >> > j...@dodgen.us >>> >> >> > _______________________________________________ >>> >> >> > sqlite-users mailing list >>> >> >> > sqlite-users@sqlite.org >>> >> >> > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >>> >> >> > >>> >> >> >>> >> >> >>> >> >> >>> >> >> >>> >> >> -- >>> >> >> >>> >> >> Best Regards, Edward J. Yoon @ NHN, corp. >>> >> >> edwardy...@apache.org >>> >> >> http://blog.udanax.org >>> > _______________________________________________ >>> > sqlite-users mailing list >>> > sqlite-users@sqlite.org >>> > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >>> > >>> _______________________________________________ >>> sqlite-users mailing list >>> sqlite-users@sqlite.org >>> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >>> >>> >> -- >> Puneet Kishor http://www.punkish.org/ >> Nelson Institute for Environmental Studies http://www.nelson.wisc.edu/ >> Open Source Geospatial Foundation (OSGeo) http://www.osgeo.org/ >> _______________________________________________ >> sqlite-users mailing list >> sqlite-users@sqlite.org >> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users >> >> > > > > _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
