tag:blogger.com,1999:blog-14455177.comments2024-03-07T18:57:25.977+01:00Mikael RonstromMikael Ronstromhttp://www.blogger.com/profile/07134215866292829917noreply@blogger.comBlogger397125tag:blogger.com,1999:blog-14455177.post-59233031855880905902022-09-20T12:57:00.984+02:002022-09-20T12:57:00.984+02:00Agreed.. SSD has completely changed the storage la...Agreed.. SSD has completely changed the storage landscape. Shah Wajahathttp://smarthomeguider.com/noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-85799357408635558662021-11-22T20:00:06.491+01:002021-11-22T20:00:06.491+01:00Hi Jonas,
Yep, Mauritz started out with your patch...Hi Jonas,<br />Yep, Mauritz started out with your patch serie, but he reversed the order, so the final<br />step will be to handle the meta data stuff. This will arrive soon in RonDB 21.10.2 which<br />should be the final step for our journey towards automatic memory management.<br /><br />Partial LCPs is probably resembling your ideas on Local LCP, we still maintain the<br />LCP checkpoint protocol but mostly ignore the signals sent between the nodes.<br />All the information about how to restore the checkpoints is available locally.<br /><br />Quite a challenge to drop the entire LCP protocol with regard to upgrade.Mikael Ronstromhttps://www.blogger.com/profile/07134215866292829917noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-45432824153087807172021-09-25T09:15:52.412+02:002021-09-25T09:15:52.412+02:00Better late than never?Better late than never?Jonas Orelandhttps://www.blogger.com/profile/02551781456999098405noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-11600821542249793802021-08-27T09:37:44.623+02:002021-08-27T09:37:44.623+02:00Added a link to docker-compose.yml in blog post. T...Added a link to docker-compose.yml in blog post. Thx for the comment.Mikael Ronstromhttps://www.blogger.com/profile/07134215866292829917noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-7813264691028793982021-08-12T20:06:41.023+02:002021-08-12T20:06:41.023+02:00Link to the docker-compose file is dead: https://g...Link to the docker-compose file is dead: https://github.com/logicalclocks/rondb-docker/rondb/21.04/docker-compose.ymljbenguiranoreply@blogger.comtag:blogger.com,1999:blog-14455177.post-48922805356358792892021-08-12T12:50:12.095+02:002021-08-12T12:50:12.095+02:00Thanks a lot, it's super useful!Thanks a lot, it's super useful!jbenguiranoreply@blogger.comtag:blogger.com,1999:blog-14455177.post-43726689367432779602021-05-19T18:28:03.237+02:002021-05-19T18:28:03.237+02:00An FYI - there is a CPU regression for large inlis...An FYI - there is a CPU regression for large inlists that started in 8.0.22. I spotted it with the sysbench query you used to get these numbers.<br />https://bugs.mysql.com/bug.php?id=102037Mark Callaghanhttps://www.blogger.com/profile/09590445221922043181noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-27691260056291392102021-03-16T18:07:03.685+01:002021-03-16T18:07:03.685+01:00Wrote a reply to those questions in a separate blo...Wrote a reply to those questions in a separate blog entry:<br />http://mikaelronstrom.blogspot.com/2021/03/designing-thread-pipeline-for-optimal.htmlAnonymoushttps://www.blogger.com/profile/09227135928663197651noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-85456688023636023592021-03-11T10:25:30.854+01:002021-03-11T10:25:30.854+01:00ScyllaDB also does something similar with executio...ScyllaDB also does something similar with execution stages for better instruction cache https://www.scylladb.com/2017/07/06/scyllas-approach-improve-performance-cpu-bound-workloads/. Is this similar to what you're implementing for instruction cache on separaring the threads ?dorianhttps://www.blogger.com/profile/10233168174184518302noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-35486060989378967532021-03-11T07:28:11.403+01:002021-03-11T07:28:11.403+01:00I do not really get function separation of threads...I do not really get function separation of threads. To execute a query, your data will now go through a pipeline of threads, instead of being executed into one. This means, you need to pass data between 4 threads, until the query is finished. It could be achieved via queueing between threads and wakeups , or maybe spinning to avoid more context switches.<br /><br />But I do not seem to see what's in that trick. My understanding is that, your query will execute the same number, or more CPU instructions, and since now there is a pipeline of several threads, it will suffer either more spinning, or more cache misses, or more context switches due to this tossing data between threads.wladhttps://www.blogger.com/profile/14657227220070201326noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-79388570322949363462020-11-19T10:20:42.916+01:002020-11-19T10:20:42.916+01:00Good post. SSD has completely changed the storage ...Good post. SSD has completely changed the storage landscape. The loading time is so quick. Once you use it, there's no going back. They're becoming more and more relevant as time passes. BackBenchers Digital Agencyhttps://backbenchersdigitalagency.com/noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-54597461539470919182020-10-06T19:38:58.746+02:002020-10-06T19:38:58.746+02:00We flush at least up to the latest epoch, but also...We flush at least up to the latest epoch, but also any pages that are ready<br />for writing. Even if a log record survives, it will only be restored if its<br />epoch is restored. So the restore will always restore a consistent checkpoint.Mikael Ronstromhttps://www.blogger.com/profile/07134215866292829917noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-76538400111498850682020-10-06T19:36:54.225+02:002020-10-06T19:36:54.225+02:00The information on IDX_BLD is not an error, it is ...The information on IDX_BLD is not an error, it is an informational<br />message. The problem with the non-starting ndb_mgmd is harder to<br />figure out without looking at the logs.Mikael Ronstromhttps://www.blogger.com/profile/07134215866292829917noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-80404300956397661152020-09-30T13:32:07.264+02:002020-09-30T13:32:07.264+02:00Hello Mikael,
Thank you for this blog.
I am tryi...Hello Mikael,<br /><br />Thank you for this blog.<br /><br />I am trying the example threadconfig you posted:<br />ThreadConfig="ldm={count=4,cpuset=0-3},tc={count=1,cpuset=4-7},main={count=1,cpuset=4-7},rep={count=1,cpuset=4-7},send={count=1,cpuset=4-7},recv={count=1,cpuset=4-7},io={cpuset=4-7},wd={cpuset=4-7}"<br /><br />when I started ndb_mgmd the following errors are generated:<br />at line 128: IO threads explicitly bound, but IDX_BLD threads not. Binding IDX_BLD to 0,1,2,3,4,5,6,7.<br /><br />I then started the ndb_mgmd with early working config.ini, but one mangement is started correctly and the second management node seems to start correctly, but it isn't. I do a ps -ef | grep ndb_mgmd; I don't see a proccess.<br /><br />I hope you explain the errors means and how to resolve the issue I encounter. I am trying to erase all RPM nd install again. <br /><br />Thank in advanced.<br /><br />Best Regards,<br /><br />Cuong Doan Nguyen<br /> <br /><br /><br /><br /><br /> Cuong Doanhttps://www.blogger.com/profile/13226089564092963729noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-10934421582577790562020-09-30T11:26:50.017+02:002020-09-30T11:26:50.017+02:00Thank you for this blog and Also thank you for bei...Thank you for this blog and Also thank you for being a good Christian.Cuong Doanhttps://www.blogger.com/profile/13226089564092963729noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-41245794471476816952020-07-17T10:59:32.592+02:002020-07-17T10:59:32.592+02:00Hi Mikael,
With this new checkpoint algorithm,the...Hi Mikael,<br /><br />With this new checkpoint algorithm,the log is still flushed to persistent storage every second and in a system-wide failure we can still recover to everything less the last second of data as before, right? If the answer is yes, I am wondering in the flush if you flush everything in the log available at that time or up to the boundary of an epoch. In other words, if there are committed transactions in the last epoch that hasn't completed, do you flush them too or just everything up to second to the last epoch?<br /><br />AlexAlex Ouhttps://www.blogger.com/profile/09780641053339854377noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-49281752066533653372020-02-14T02:48:32.890+01:002020-02-14T02:48:32.890+01:00Some more details: The benchmark is a variant of t...Some more details: The benchmark is a variant of the YCSB benchmark where<br />data is stored in Disk Data Columns in NDB. More details will come later.Mikael Ronstromhttps://www.blogger.com/profile/07134215866292829917noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-55281485354786164522020-02-10T14:45:01.874+01:002020-02-10T14:45:01.874+01:00Really awesome but difficult make the same test in...Really awesome but difficult make the same test in house FARUSAChttps://www.blogger.com/profile/10187762707507605860noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-16678126962622874742019-11-18T15:48:05.823+01:002019-11-18T15:48:05.823+01:00Hi Mikael. It is great to see all the improvments ...Hi Mikael. It is great to see all the improvments coming in NDB.<br /><br />One thing I have a question about is the following from the MYSQL 8.0 release notes:<br /><br />"Two MySQL storage engines currently provide native partitioning support—InnoDB and NDB; of these, only InnoDB is supported in MySQL 8.0. Any attempt to create partitioned tables in MySQL 8.0 using any other storage engine fails. "<br />https://dev.mysql.com/doc/refman/8.0/en/mysql-nutshell.html<br /><br />Is this going to be true for the GA version of MySQL Cluster?Anonymoushttps://www.blogger.com/profile/16721899523543507291noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-41019859418506206492019-09-10T14:18:42.425+02:002019-09-10T14:18:42.425+02:00thanksthanksdireecthttps://direect.denoreply@blogger.comtag:blogger.com,1999:blog-14455177.post-80654435195256557052019-08-14T17:28:16.556+02:002019-08-14T17:28:16.556+02:00Thanks, very good work.
Did you always kept a rep...Thanks, very good work.<br /><br />Did you always kept a replica or did you use replica only for the first 2 nodes test?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-14455177.post-48707200183856769812019-05-10T07:45:45.716+02:002019-05-10T07:45:45.716+02:00Can you please tell me how do I map JSON field in ...Can you please tell me how do I map JSON field in the table to a Object in ClusterJ?<br /><br />Create table table_name (<br />-----<br />-----<br />json JSON,<br />-----)ENGINE=NDBCLUSTER;<br /><br /><br />how do I map json field to a java object?Anonymoushttps://www.blogger.com/profile/14717934785134974896noreply@blogger.comtag:blogger.com,1999:blog-14455177.post-14464283416922098602019-05-02T13:38:18.368+02:002019-05-02T13:38:18.368+02:00I am very grateful you did share your knowledge he...I am very grateful you did share your knowledge here. It is an excellent post<br />Robert Smithhttps://www.clippingsolutions.com/blog.htmlnoreply@blogger.comtag:blogger.com,1999:blog-14455177.post-67252081746265549112019-03-25T10:59:37.444+01:002019-03-25T10:59:37.444+01:00Agree with one of the above comment sounds like Wi...Agree with one of the above comment sounds like Windows Vista's Ready Boost feature.Web Designhttps://linkpak.comnoreply@blogger.comtag:blogger.com,1999:blog-14455177.post-45339373561260810872019-03-20T09:42:29.778+01:002019-03-20T09:42:29.778+01:00Nice PostNice PostSEO serviceshttps://linkpak.com/services/seo-services-in-karachi/noreply@blogger.com