D
dbeavon
Guest
You are right that I prefer to treat PASOE as a middle tier for business logic. Many of the features of web servers (load balancing and fault tolerance) work their best in the middle tier. Moreover, hardware is provisioned VERY differently for web servers vs. database servers. (IE. You build a big, expensive server to run the database and it is configured with lots of good disk, RAM, etc. Its failover model is quite different as well - based on failover clustering of the disk resource). The PASOE web servers, on the other hand, should typically live on very inexpensive hardware. There should be redundant copies for failover and load balancing (via Citrix netscaler or whatever). Unlike the database server, they would often be provisioned by way of docker containers. They would have virtually no disk. Unlike the database server, the disk would be inexpensive and expendable. Also, the installation of third-party software is quite different between the two. The PASOE web servers are typically loaded with LOTS of third-party cruft that is a dependency of the given application (plugins for FTP, email, excel, custom services, custom utilities, etc). You do NOT want that type of a mess living on your database server. Nor do you want any inefficient or runaway ABL processes (_progres, _proapsv, _mproapsv with custom code) to compete for the RAM, CPU and network of the database server processes (ie. don't interfere with _mprosrv, _sqlsrv, and similar). I agree with you that historically Progress ABL was NOT designed very well for running in the middle tier. At times it seems like it has more in common with a "data manipulation language" that would otherwise run within the memory of the database (ie. more in common with T-SQL than with a general-purpose language like C#.Net). In that vein we have often written "classic" appserver programs just to serve as a substitute for a better query processing engine. (Ie. a replacement for an inadequate SQL.) But I think the "new" PASOE/tomcat product is forcing Progress to prioritize their "client-server" connectivity and start improving it. Sysadmins, especially ones that aren't familiar with the evolution of OpenEdge, are probably very hesitant to be running tomcat and a database engine on the same server. Remember that the most highly advertised feature of OE12 are related to improvements in "client-server" connectivity. (IE. 300% improvement due to multithreading and server-side joins.)
Continue reading...
Continue reading...