Purpose of Kernel 2.6.x ? Unionfs and Aufs ?
Submitted by atang1 on Wed, 06/18/2008 - 02:49.
Since 2002, kernel 2.6.x started to be add-on to kernel 2.4.x. RedHat subcontractor thought it work take 2 years to adapt RHEL to kernel 2.6.x. Well, it is now 2008.
The purpose of kernel 2.6.x was simple. to have local priority of database(on regional servers) in drams; and updated/upgraded ASAP. This was assigned to unionfs and later aufs.
Since then, fs was imagined to be used to update/upgrade apps(not the data in the database). So, the scripts of the unionfs and aufs are basically useless.
Take Drupal, a standalone software. If you have ten regional servers, the priority is to distribute the webpages from the main server or distributed local servers to each other. unionfs and aufs can not do the tasks.
Take Mysql with html or xml webpages, how can unionfs or aufs do the update or upgrade of all the servers from each other?
Take open Office Suite, how can you update or upgrade workgroup peer to peer data storage of all open office data?
Kernel 2.6.x will take another 5 years before it will accomplish the task it was planned for; but first get the identification of the servers in the network, that needed the update/upgrade in unionfs and aufs? Unionfs and aufs should have the same data storage in each server ASAP, then distributed from the drams in each server to users. In other words, search engines search uniofs or aufs for webpages.
Open sourced community did not understand what the purpose of kernel 2.6.x was and planned for the completion of the task using synchronized unionfs or aufs among all the regional servers on the network?
Kernel 2.6.x udev, unionfs and aufs ? Realtime matrix switching
Kernel 2.6.x was invented for priority in distributed data service. But, first udev has to take care of ambiguous drivers*firmware) and USB hotplug of any devices. Udev is in two parts, boot to install drivers, and hotplug to install devices at anytime after booting. Developers had to make sure that removing hotplugs must reset all variables for the next hotplug. It took six years since kernel 2.6.x began, some developers are just beginning to understand the importance of resetting variables; after a task is finished, so another task can begin.
The development of unionfs and aufs is still elementary. They are targeting repository only. Instead of many regional database servers to have peer to peer cron jobs done in sequence in bonded channels of data streams. we have 114 data centers, and we use atomic clock to schedule data streams for cron jobs. Each data center has a partition for cron jobs, and each file is a thread. Telephony posix threads handle data rather well.
Until cron jobs are scheduled, unionfs and aufs can be deleted from at least all the Linux desktops. Servers must have schedule modules for data synchronization among all the data centers. Afterall, matrix switching archtecture must be used to synchronize data from all the data centers by cpu cores(millions in parallel processing simultaneously), storage partitions(114) and memory protected thread of posix packets(billions). Mosix workload balancing(simultaneous processing of multicpu cores) is done by extra cpu core to control data transfer overload, failsafe and failover to new resources.
Aufs and Unionfs are used as cache for data synchronization ?
We have found out years ago that large regional data, has to be kept as regional, so the cache in data transfer is kept in their regional partitions. No intermixing of data. And checksum will keep data integrity. Power and speed of data transfer(edgeQam 256 channels) is absolutely important.
The development of aufs and unionfs are in layers of fs. Which is in a way same as partitions. But layers may have to correspondent to the number of regional servers, each with its own data input.
Without knowing how many regional servers you have; leave the aufs and unionfs off your computer. Why bother to even put them in your desktop only operating system.
To know how many layers in your aufs or unionfs, you need a cron job program, which has all the links to your regional servers and the data file names to be transferred to your local server. This achieves the data priority in local servers to keep current all the sequential data from regional servers immediately within certain propagation delay allowed.
This will make kernel 2.6.x achieve its objectives to have very larger enterprise systems current on all the regional data with high speed parallel and simultaneous processing. Cron job programming of course depends on data media, mpeg4, xml, dhtml, etc. and their own database programs.
Do we need to copy from Oracle distributed database(?); no, just write a simple cron job program with as many links as layers to do matrix data transfer to aufs or unionfs in as short a time period as possible.