Sundry i5/OS V6R1 and System i Enhancements
February 18, 2008 Timothy Prickett Morgan
Our coverage of the forthcoming i5/OS V6R1 operating system continues this week with a hodge-podge of smaller enhancements and features that were rolled up into the January 29 announcements. IBM has made a bunch of tweaks to DB2 for i5/OS, logical partitioning, performance measurement, peripheral support, and other areas that will probably not individually compel customers to upgrade, but add their collective weight to the stack of new features and it could break the camel’s checkbook open to do some spending in 2008.
I know we all mostly still call the integrated relational database inside OS/400 and i5/OS DB2/400, but IBM is trying to get the DB2 for i5/OS brand to stick, and with i5/OS V6R1, the database is being given the full name of DB2 for i5/OS V6R1, in fact. And IBM has woven a lot of new goodies into this release of the database management system. SQL Call Level Interface (CLI) data access has been updated with the support of CLI wide-character APIs; functions related to these APIs are found in applications that use Unicode or double-byte characters. The database also now supports a new numeric data type called decimal float, which has the attributes of both decimal and floating point and which makes use of the new decimal math units inside the Power6 processor. (IBM is the first chip maker to add decimal math units to a chip.)
The .NET data provider for DB2 for i5/OS is also updated to Microsoft‘s ADO.NET 2.0 specification, and the software can now also support distributed transactions and multi-row inserts into the database. IBM has also provided tighter integration with Microsoft’s Visual Studio.NET development tools, including support for Visual Studio database interfaces and wizards, including Server Explorer. Java-style access to DB2 for i5/OS V6R1 has also been tweaked with support for enhancements made to JDBC Version 4.0. On the SQL front, IBM has made improvements to the SQL Query Engine optimizer that first made its debut in OS/400 V5R2 several years ago and that sits side-by-side in the database with the so-called Classic Query Engine, or CQE in IBM Rochester parlance. With each subsequent release of the DB2 for i5/OS database, IBM allows a larger and larger number of SQL functions that were created for the CQE optimizer to run through the new SQE optimizer. Moreover–and this is critically important for IBM if it wants more applications to be ported to i5/OS–SQL has been updated to have the automatic hidden timestamp columns, the full outer join, the Select from Insert, and the Values on From capabilities that are in the Windows-Unix-Linux variant of DB2.
On the logical partitioning front, IBM is making dynamic partitions better square up with the per-core pricing model customers have with their software vendors. Power6 machines running i5/OS V5R4M5 or V6R1 will have a new capability called shared processor pools, which puts a core count cap on a group of partitions instead of on a single partition. With i5/OS V5R4, customers can put a cap on the size of each partition in a machine, but those partitions are relatively static and not necessarily flexible enough to copy with the ups and downs of running a business. Instead of capping each micropartition, IBM is now allowing customers to group a set of partitions and put a cap on that and also at the same time allowing micropartitions to be uncapped.
Here is how this new features is useful. Let’s say you have an eight-core System i 570 with four partitions–one running AIX for Web infrastructure and three running i5/OS–one each for the Americas, Europe, and Asia/Pacific regions where your company does business. With i5/OS, you would give each i5/OS partition two processors, assuming the workloads were more or less equal and necessitated by software licensing terms for application software to be pinned down. But that would leave excess capacity in some partitions during the day–capacity that could be used by other business units. So with shared processor pools, you would take the three logical partitions and group them as a pool, setting a limit of a maximum of six cores for all three. You would pay for six cores worth of i5/OS and application software licenses, just like before, but if the European partition needed four or five cores, that partition could expand and the others could shrink and still balance out to six cores. As the Earth spun around on its axis, the partitions would change to reflect the changing workloads–all without violating software licensing terms. Without this capability, each partition running i5/OS and the application stack would have to have enough processing capacity to deal with its own peaks, which defeats the purpose of virtualization. The pooling also means that, in the case of this System i 570, you can demonstrate to the application vendor that no i5/OS application can grab any processing capacity in the AIX partition, which means you only have to pay for six engines instead of eight.
Shared processor pools are not limited to i5/OS V6R1, but can also be used with V5R4M5, AIX 5.3 and 6.1, and various Linuxes.
The i5/OS V6R1 operating system also includes a new freebie tool developed by IBM Rochester called Performance Data Investigator, which is a graphical front end for the i5/OS Collection Services tool. IBM has added a performance group in the new Navigator for i5/OS management tool (see New Web Console Debuts with i5/OS V6R1 for more on that) and this allows the data culled from a Collection Services run to be displayed visually in charts and organized into tables. IBM says that Collection Services files from OS/400 and i5/OS releases that predate V6R1 can be converted to the V6R1 format and then viewed in the Performance Data Integrator tool, too, which means customers can do before and after analysis of the performance of workloads as they complete their upgrades.
On January 29, IBM also put out a statement of direction that it would be doing a firmware upgrade on the Power6-based System i 570 server (that’s machine number 9406-MMA) that would allow additional nodes to be added to an SMP cluster without having to take the running node or nodes in an existing 570 box down and rebooting it. IBM is also working to allow a crashed or repaired node in a machine to be reintegrated into a cluster without having to restart the machine as well. Once these extra nodes or repaired nodes are brought online, the service processor and system console will see them and operating systems and applications will be able to expand out onto them over top of the PowerVM hypervisor.
Finally, in case you missed it, IBM is now supporting attachment of its entry DS4700 and DS4800 disk arrays through the Virtual I/O Server feature that was ported into i5/OS V6R1 from AIX. VIOS is a means of running device drivers for peripherals in a dedicated partition, which means that an operation system doesn’t need to have drivers written for it specifically any more. Operating systems talk to a VIOS partition inside of a single physical machine, and VIOS talks to the peripherals. VIOS is probably an extremely cut down version of AIX, in fact, although it could be Linux. (Wouldn’t that be funny?)