Release of InterSystems Caché 2015.2 DBMS
InterSystems Corporation at the Global Summit 2015 made a socialist commitment to announce the policy of frequent releases - the release of new versions at least 4 times a year. In fulfillment of this vow, on July 8, the release of the DBMS InterSystems Caché 2015.2 was released 3 months after the release of Caché 2015.1 . Video about the release.
In Caché 2015.2, performance and scalability were improved, support for Russian and Ukrainian in iKnow appeared, support for CORS in REST was implemented.
Optimization of the global caching algorithm in systems with a large number of cores due to the reduction of internal locks with competitive access.
Dynamic addressing of global vectors. In this release, the performance of global references (global references) in processes that constantly use more than 32 different globals is significantly improved. This was made possible by increasing the number of global vectors for each process. Each vector of globals contains information about the global to which the process has recently accessed. This information is used to optimize future access to the same global. Previously, 32 globals were stored for each process. Now this number can dynamically increase up to 1000 vectors for the process. More details .
Bitslide indexes are now used for aggregate functions Sum, Average, Count and if WHERE is present in the query. This should automatically increase the performance of some queries in your applications.
')
Parallel queries in SQL now work not only for numeric fields. In the previous version, the optimizer allowed parallel execution only if the requested fields were numeric. Now this restriction is removed. Details about % Parallel.
Increased scalability of ECP solutions
The improved optimization of the global cash cache, mentioned above, also applies to the ECP. Solutions with ECP in the new version now automatically become more productive and increased scaling capabilities.
CORS support in REST
Since this release in the base class % CSP.REST , support for CORS has appeared, which allows you to provide this functionality in your RESTful Caché solutions out of the box. Caché now expects OPTIONS header in requests, also in the class % CSP.REST appeared HandleCORSRequest (pUrl) callback, which will allow to implement the functionality required for the CORS application.
Support for Russian and Ukrainian languages ​​in iKnow officially announced with this release of Caché. This great news will allow developers in Russia and Ukraine to fully utilize InterSystems iKnow technology for Text Mining solutions. About examples of solutions with iKnow in Russian there are already publications on Habré ( one , two , three ).
Support for complex queries in iFind using brackets and AND OR NOT operators. Those. now you can make a request
"(long strings) AND (memory OR disk *) AND NOT (error message)"
which will find all the records containing the phrase "long lines", the word "memory" or the word beginning with "disk" and not containing the phrase "error message".
Other iFind and iKnow improvements :
Cube Versions
This innovation will make the restructuring of cubes more invisible to users of DeepSee solutions. Rebuilding cubes can take a significant amount of time, which can result in idle cube users. Now, when rebuilding, the cube is not deleted, but a new version is created, while users continue to work with the previous version of the cube. As soon as the new cube is built, users will be switched to it, and the old cube will be deleted. This will obviously increase the disk space requirements for the DeepSee solution, but will seriously reduce the time for switching users to the updated cube.
List groups in DeepSee
Previously, if a user wanted to add an arbitrary list (Listing) to DeepSee (the list is an SQL query to the fact table), you had to make a change to the cube itself, which is not always convenient. Starting from 2015.2, these lists can be changed without changing the cube: now lists can be created, edited and deleted for a group of lists (Listing Group), which can be applied to a specific DeepSee cube and does not require its modification and compilation. This is a significant convenience in the application and management of queries to the fact tables of DeepSee cubes.
Improved mirroring technology
The automatic file server is now faster in some cases due to an improved algorithm for detecting the failure of the main server or disabling its network.
It also simplifies the setting of filer.
Added support for Node.js v0.12.
Installing Caché and Ensemble not from root or admin becomes possible starting with this release.
Support for the Islamic calendar (Hijri) in Caché. The Islamic calendar is also supported in DeepSee, which allows you to create dimensions and levels based on it for analyzing fact tables.
Support for two-factor authentication based on temporary one-time passwords (implemented by RFC6238 ).
QR code support. Now, for the generation and use of QR codes in their applications, a special system class has appeared. This functionality is implemented by methods of class % SYS.QRCode .
The full text of the release can be found here .
In Caché 2015.2, performance and scalability were improved, support for Russian and Ukrainian in iKnow appeared, support for CORS in REST was implemented.
Productivity increase
Optimization of the global caching algorithm in systems with a large number of cores due to the reduction of internal locks with competitive access.
Dynamic addressing of global vectors. In this release, the performance of global references (global references) in processes that constantly use more than 32 different globals is significantly improved. This was made possible by increasing the number of global vectors for each process. Each vector of globals contains information about the global to which the process has recently accessed. This information is used to optimize future access to the same global. Previously, 32 globals were stored for each process. Now this number can dynamically increase up to 1000 vectors for the process. More details .
Bitslide indexes are now used for aggregate functions Sum, Average, Count and if WHERE is present in the query. This should automatically increase the performance of some queries in your applications.
')
Parallel queries in SQL now work not only for numeric fields. In the previous version, the optimizer allowed parallel execution only if the requested fields were numeric. Now this restriction is removed. Details about % Parallel.
Increased scalability of ECP solutions
The improved optimization of the global cash cache, mentioned above, also applies to the ECP. Solutions with ECP in the new version now automatically become more productive and increased scaling capabilities.
Web development
CORS support in REST
Since this release in the base class % CSP.REST , support for CORS has appeared, which allows you to provide this functionality in your RESTful Caché solutions out of the box. Caché now expects OPTIONS header in requests, also in the class % CSP.REST appeared HandleCORSRequest (pUrl) callback, which will allow to implement the functionality required for the CORS application.
Unstructured data processing with iKnow and iFind
Support for Russian and Ukrainian languages ​​in iKnow officially announced with this release of Caché. This great news will allow developers in Russia and Ukraine to fully utilize InterSystems iKnow technology for Text Mining solutions. About examples of solutions with iKnow in Russian there are already publications on Habré ( one , two , three ).
Support for complex queries in iFind using brackets and AND OR NOT operators. Those. now you can make a request
"(long strings) AND (memory OR disk *) AND NOT (error message)"
which will find all the records containing the phrase "long lines", the word "memory" or the word beginning with "disk" and not containing the phrase "error message".
Other iFind and iKnow improvements :
- Support for stemming and decomposition in iFind.
- You can use iFind indexes even for those languages ​​that iKnow does not support, assuming that words are separated by spaces from each other.
- Improved algorithm for determining the semantic dominance of entities in the text, which increased performance and requires less disk space for iFind and iKnow solutions.
- Now iKnow engine can transfer its “marker terms” for definitions of “negations” (negation) and the tonality of sentences.
Business Analytics with DeepSee
Cube Versions
This innovation will make the restructuring of cubes more invisible to users of DeepSee solutions. Rebuilding cubes can take a significant amount of time, which can result in idle cube users. Now, when rebuilding, the cube is not deleted, but a new version is created, while users continue to work with the previous version of the cube. As soon as the new cube is built, users will be switched to it, and the old cube will be deleted. This will obviously increase the disk space requirements for the DeepSee solution, but will seriously reduce the time for switching users to the updated cube.
List groups in DeepSee
Previously, if a user wanted to add an arbitrary list (Listing) to DeepSee (the list is an SQL query to the fact table), you had to make a change to the cube itself, which is not always convenient. Starting from 2015.2, these lists can be changed without changing the cube: now lists can be created, edited and deleted for a group of lists (Listing Group), which can be applied to a specific DeepSee cube and does not require its modification and compilation. This is a significant convenience in the application and management of queries to the fact tables of DeepSee cubes.
High availability enhancements
Improved mirroring technology
The automatic file server is now faster in some cases due to an improved algorithm for detecting the failure of the main server or disabling its network.
It also simplifies the setting of filer.
Other improvements
Added support for Node.js v0.12.
Installing Caché and Ensemble not from root or admin becomes possible starting with this release.
Support for the Islamic calendar (Hijri) in Caché. The Islamic calendar is also supported in DeepSee, which allows you to create dimensions and levels based on it for analyzing fact tables.
Support for two-factor authentication based on temporary one-time passwords (implemented by RFC6238 ).
QR code support. Now, for the generation and use of QR codes in their applications, a special system class has appeared. This functionality is implemented by methods of class % SYS.QRCode .
The full text of the release can be found here .
Source: https://habr.com/ru/post/263065/
All Articles