RE: Jackrabbit 2.10 vs Oak 1.2.7
Hi Marcel, I uploaded all the source to github along with a summary spreadsheet. I would appreciate any time you have to review. https://github.com/Domenic-Ansys/Jackrabbit2-Oak-Tests As you stated the move is a non goal, but in comparison to Jackrabbit 2 I am also finding in my tests that create, update, and copy are all faster in Jackrabbit 2 (10k nodes). Any input would be appreciated... Also, will MySql will not be listed as "Experimental" at some point? Thanks, Domenic -Original Message- From: Marcel Reutegger [mailto:mreut...@adobe.com] Sent: Thursday, March 31, 2016 6:14 AM To: oak-dev@jackrabbit.apache.org Subject: Re: Jackrabbit 2.10 vs Oak 1.2.7 Hi Domenic, On 30/03/16 14:34, "Domenic DiTano" wrote: >"In contrast to Jackrabbit 2, a move of a large subtree is an expensive >operation in Oak" >So should I avoid doing a move of a large number of items using Oak? >If we are using Oak then should we avoid operations with a large number >of items in general? In general it is fine to have a large change set with Oak. With Oak you can even have change sets that do not fit into the heap. > As a FYI - there are other benefits for us to move to Oak, but our >application uses executes JCR operations with a large number of items >quite often. I am worried about the performance. > >The move method is pretty simple - should I be doing it differently? > >public static long moveNodes(Session session, Node node, String >newNodeName) >throws Exception{ > long start = System.currentTimeMillis(); > session.move(node.getPath(), "/"+newNodeName); > session.save(); > long end = System.currentTimeMillis(); > return end-start; >} No, this is fine. As mentioned earlier, with Oak a move operation is not cheap and is basically implemented as copy to new location and delete at the old location. A cheap move operation was considered a non-goal when Oak was designed: https://wiki.apache.org/jackrabbit/Goals%20and%20non%20goals%20for%20Jackr a bbit%203 Regards Marcel
RE: Jackrabbit 2.10 vs Oak 1.2.7
Hi Marcel, Thanks for your input, it is really appreciated - yes I can put all the code in Github (I will need a day or two). I did forget to mention that I ran the tests with the following settings: -Xms1024m -Xmx2048m -Doak.queryLimitInMemory=50 -Doak.queryLimitReads=10 -Dupdate.limit=25 -Doak.fastQuerySize=true As for the data, there is one issue in the second Jackrabbit 2 run there were 100 files uploaded as opposed to 1000. No I did not mix up the other results, I ran these tests about 10 times and these results were pretty consistent. I ran them on my local laptop, so I would assume that you would get better results with a dedicated machine. "In contrast to Jackrabbit 2, a move of a large subtree is an expensive operation in Oak" So should I avoid doing a move of a large number of items using Oak? If we are using Oak then should we avoid operations with a large number of items in general? As a FYI - there are other benefits for us to move to Oak, but our application uses executes JCR operations with a large number of items quite often. I am worried about the performance. The move method is pretty simple - should I be doing it differently? public static long moveNodes(Session session, Node node, String newNodeName) throws Exception{ long start = System.currentTimeMillis(); session.move(node.getPath(), "/"+newNodeName); session.save(); long end = System.currentTimeMillis(); return end-start; } Thanks, Domenic -Original Message- From: Marcel Reutegger [mailto:mreut...@adobe.com] Sent: Wednesday, March 30, 2016 4:42 AM To: oak-dev@jackrabbit.apache.org Subject: Re: Jackrabbit 2.10 vs Oak 1.2.7 Hi, On 29/03/16 14:55, "Domenic DiTano" wrote: >Sending the data again, I hope this is makes it clearer. I do not mind >sharing the source, assuming you just want the code that does the >creating, deleting etc of nodes (attached) How I created the Document >stores is in the previous email, but if you want I can send that also. yes, I'm just interested in the test code. Can you please make it available, e.g. over github? some comments on the results: In contrast to Jackrabbit 2, a move of a large subtree is an expensive operation in Oak. With Jackrabbit 2, both the content update as well as the index update is rather cheap when a subtree is moved. With Oak, the cost depends on the number of items you move. Some of the results for Jackrabbit 2 with 10k nodes are better than with just 1k. Did you mix up numbers? As mentioned before you potentially get a speedup with Oak when you tweak the update.limit for large change sets with 10k nodes. Regards Marcel >All milliseconds... > >Oak: >Create 1000 (Mysql,PostGress,Mongo) 3444,2483,8497 Query 1000 >(Mysql,PostGress,Mongo) 2,19,2 Upload 100 files (Mysql,PostGress,Mongo) >1455,1130,845 Move 1000 (Mysql,PostGress,Mongo) 96349,2404,14428 Copied >1000 (Mysql,PostGress,Mongo) 2246,556,4432 Delete 1000 >(Mysql,PostGress,Mongo) 92923,1523,7667 Update 1000 >(Mysql,PostGress,Mongo) 48647,1055,4640 Read 1000 >(Mysql,PostGress,Mongo) 98,111,142 > > >Jackrabbit 2: >Create 1000 (Mysql) 3022 >Query 1000 (Mysql) 143 >Upload 100 files (Mysql) 1105 >Move 1000 (Mysql) 16 >Copied 1000 (Mysql) 764 >Delete 1000 (Mysql) 1481 >Update 1000 (Mysql) 1139 >Read 1000 (Mysql) 12 > > >Oak: >Create 1 (Mysql,PostGress,Mongo) 31250,16475,342192 Query 1 >(Mysql,PostGress,Mongo) 4,16,2 Upload 100 files (Mysql,PostGress,Mongo) >1146,605,753 Move 1 (Mysql,PostGress,Mongo) 741474,30339,406259 >Copied 1 (Mysql,PostGress,Mongo) 20755,7615,43670 Delete 1 >(Mysql,PostGress,Mongo) 728737,24461,43670 Update 1 >(Mysql,PostGress,Mongo) 374387,12453,41053 Read 1 >(Mysql,PostGress,Mongo) 2216,2989,968 > > >Jackrabbit 2: >Create 1 (Mysql) 8507 >Query 1 (Mysql) 94 >Upload 100 files (Mysql) 744 >Move 1 (Mysql) 14 >Copied 1 (Mysql) 489 >Delete 1 (Mysql) 824 >Update 1 (Mysql) 987 >Read 1 (Mysql) 8 > > >On Tue, Mar 29, 2016 at 8:28 AM, Marcel Reutegger <mreut...@adobe.com> >wrote: > >Hi Domenic, > >the number of test cases do not match the results you provided. i.e. >the column headers do not match data columns. can you please clarify >how the results map to the test cases? > >also, do you mind sharing the test code? I'd like to better understand >what the tests do. > >Regards > Marcel > >On 29/03/16 14:04, "Domenic DiTano" wrote: > >>Sorry those images did not come through, posting the email again with >>the raw data: >> >>I work with web application that has Jackrabbit 2.10 embedded and we >>wanted to try upgrading to Oak. Our current configuration that we use >>for Jackrabbit 2.10 is the FileDataStore a
RE: Jackrabbit 2.10 vs Oak 1.2.7
't make it through to the mailing list. Can you please post raw numbers or a link to the graphs? Without access to more data, my guess is that Oak on DocumentNodeStore is slower with the bigger changes set because it internally creates a branch to stage changes when it reaches a given threshold. This introduces more traffic to the backend storage when save() is called, because previously written data is retrieved again from the backend. Jackrabbit 2.10 on the other hand keeps the entire changes in memory until save() is called. You can increase the threshold for the DocumentNodeStore with a system property: -Dupdate.limit=10 The default is 10'000. Regards Marcel On 29/03/16 04:19, "Domenic DiTano" wrote: >Hello, > >I work with web application that has Jackrabbit 2.10 embedded and we >wanted to try upgrading to Oak. Our current configuration that we use >for Jackrabbit >2.10 is the FileDataStore along with MySql for the Persistence DataStore. > We wrote some test cases to measure the performance of JackRabbit >2.1.0 vs latest Oak 1.2. In the case of JackRabbit 2.10, we used what >our current application configuration FileDataStore along with MySql. >In the case of Oak we tried many configurations but the one we settled >on was a DocumentNodeStore with a FileDataStore backend.We tried all 3 >RDB options (Mongo, PostGress, MySql). >All Test cases used the same code which standard >JCR 2.0 code. The test cases did the following: > >. >create 1000 & 10,000 nodes >. >move 1000 & 10,000 nodes >. >copy 1000 & 10,000 nodes >. >delete 1000 & 10,000 nodes >. >upload 100 files >. >read 1 property on 1000 & 10,000 nodes >. >update 1 property on 1000 & 10,000 nodes > > >The results were as follows (all results in milliseconds): > >Oak tests ran with the creation, move, copy, delete, update, and read >of >1000 nodes: > > > >Postgress seems to perform well overall. > >In the case of Jackrabbit 2.10 (tests ran with the creation, move, >copy, delete, update, and read of 1000 nodes): > > > >Jackrabbit 2.10 performs slightly better than Oak. > >The next set of tests were ran with Oak with the creation, move, copy, >delete, update, and read of 1 nodes: > > > >Postgress once again performed ok. Mongo and MySql did not do well >around Moves, deletes, and updates. Querying did well also as indexes >were created. > >In the case of Jackrabbit 2.10 (tests ran with the creation, move, >copy, delete, update, and read of 1 nodes): > > > >Jackrabbit 2.10 performed much >better than Oak in general. > >Based on the results I have a few questions/comments: > >. >Are these fair comparisons between Jackrabbit and Oak? In our >application it is very possible to create 1-10,000 nodes in a user >session. >. >Should I have assumed Oak would outperform Jackrabbit 2.10? >. >I understand MySql is experimental but Mongo is not I would assume >Mongo would perform as well if not better than Postgress >. >The performance bottlenecks seem to be at the JDBC level for MySql. I >made some configuration changes which helped performance but the >changes would make MySql fail any ACID tests. > >Just a few notes: > >The same JCR code was used for creating, moving, deleting etc any nodes. >The JCR code was used for all the tests. The tests were all run on the >same machine > >Used DocumentMK Builder for all DataStores: > >Mongo: >DocumentNodeStore storeD = new >DocumentMK.Builder().setPersistentCache("D:\\ekm-oak\\Mongo,size=1024,b >ina >ry=0").setMongoDB(db).setBlobStore(new >DataStoreBlobStore(fds)).getNodeStore(); > >MySql: > RDBOptions options = new >RDBOptions().tablePrefix(prefix).dropTablesOnClose(false); >DocumentNodeStore storeD = new >DocumentMK.Builder().setBlobStore(new >DataStoreBlobStore(fds)).setClusterId(1).memoryCacheSize(64 * 1024 * >1024). > >setPersistentCache("D:\\ekm-oak\\MySql,size=1024,binary=0").setRDBConne >cti on(RDBDataSourceFactory.forJdbcUrl(url, userName, password), >options).getNodeStore(); >PostGres: >RDBOptions options = new >RDBOptions().tablePrefix(prefix).dropTablesOnClose(false); >DocumentNodeStore storeD = new >DocumentMK.Builder().setAsyncDelay(0).setBlobStore(new >DataStoreBlobStore(fds)).setClusterId(1).memoryCacheSize(64 * 1024 * >1024). > >setPersistentCache("D:\\ekm-oak\\postGress,size=1024,binary=0").setRDBC >onn ection(RDBDataSourceFactory.forJdbcUrl(url, userName, password), >options).getNodeStore(); > >The repository was created the same for all three: >Repository repository = new Jcr(new Oak(storeD)).with(new >LuceneIndexEditorProvider()).with(configureSearch()).createRepository() >; > >Any input is welcomeŠ. > >Thanks, >Domenic > >
Jackrabbit 2.10 vs Oak 1.2.7
Hello, I work with web application that has Jackrabbit 2.10 embedded and we wanted to try upgrading to Oak. Our current configuration that we use for Jackrabbit 2.10 is the FileDataStore along with MySql for the Persistence DataStore. We wrote some test cases to measure the performance of JackRabbit 2.1.0 vs latest Oak 1.2. In the case of JackRabbit 2.10, we used what our current application configuration – FileDataStore along with MySql. In the case of Oak we tried many configurations but the one we settled on was a DocumentNodeStore with a FileDataStore backend. We tried all 3 RDB options (Mongo, PostGress, MySql). All Test cases used the same code which standard JCR 2.0 code. The test cases did the following: · create 1000 & 10,000 nodes · move 1000 & 10,000 nodes · copy 1000 & 10,000 nodes · delete 1000 & 10,000 nodes · upload 100 files · read 1 property on 1000 & 10,000 nodes · update 1 property on 1000 & 10,000 nodes *The results were as follows (all results in milliseconds):* Oak tests ran with the creation, move, copy, delete, update, and read of 1000 nodes: Postgress seems to perform well overall. In the case of Jackrabbit 2.10 (tests ran with the creation, move, copy, delete, update, and read of 1000 nodes): Jackrabbit 2.10 performs slightly better than Oak. The next set of tests were ran with Oak with the creation, move, copy, delete, update, and read of 1 nodes: Postgress once again performed ok. Mongo and MySql did not do well around Moves, deletes, and updates. Querying did well also as indexes were created. In the case of Jackrabbit 2.10 (tests ran with the creation, move, copy, delete, update, and read of 1 nodes): Jackrabbit 2.10 performed much better than Oak in general. *Based on the results I have a few questions/comments:* · Are these fair comparisons between Jackrabbit and Oak? In our application it is very possible to create 1-10,000 nodes in a user session. · Should I have assumed Oak would outperform Jackrabbit 2.10? · I understand MySql is experimental but Mongo is not – I would assume Mongo would perform as well if not better than Postgress · The performance bottlenecks seem to be at the JDBC level for MySql. I made some configuration changes which helped performance but the changes would make MySql fail any ACID tests. *Just a few notes:* The same JCR code was used for creating, moving, deleting etc any nodes. The JCR code was used for all the tests. The tests were all run on the same machine Used DocumentMK Builder for all DataStores: Mongo: DocumentNodeStore storeD = new DocumentMK.Builder().setPersistentCache("D:\\ekm-oak\\Mongo,size=1024,binary=0").setMongoDB(db).setBlobStore(new DataStoreBlobStore(fds)).getNodeStore(); MySql: RDBOptions options = new RDBOptions().tablePrefix(prefix).dropTablesOnClose(false); DocumentNodeStore storeD = new DocumentMK.Builder().setBlobStore(new DataStoreBlobStore(fds)).setClusterId(1).memoryCacheSize(64 * 1024 * 1024). setPersistentCache("D:\\ekm-oak\\MySql,size=1024,binary=0").setRDBConnection(RDBDataSourceFactory.forJdbcUrl(url, userName, password), options).getNodeStore(); PostGres: RDBOptions options = new RDBOptions().tablePrefix(prefix).dropTablesOnClose(false); DocumentNodeStore storeD = new DocumentMK.Builder().setAsyncDelay(0).setBlobStore(new DataStoreBlobStore(fds)).setClusterId(1).memoryCacheSize(64 * 1024 * 1024). setPersistentCache("D:\\ekm-oak\\postGress,size=1024,binary=0").setRDBConnection(RDBDataSourceFactory.forJdbcUrl(url, userName, password), options).getNodeStore(); The repository was created the same for all three: Repository repository = new Jcr(new Oak(storeD)).with(new LuceneIndexEditorProvider()).with(configureSearch()).createRepository(); Any input is welcome…. Thanks, Domenic