The System i Strategy for New Customers

[MCWebsite.Staff]

** This thread discusses the article: The System i Strategy for New Customers **
** This thread discusses the Content article: The System i Strategy for New Customers **
0

[charles400]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
Have the brilliant minds of IBM Marketeing ever used a search engine? Good God. When I Google something for work, I have to enter AS400 one one search, then modify it again with iSeries to get different results. STOP WITH THE NAME GAME ALREADY! It makes IBM look foolish.

[pcatlin]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
The real challenge is to get the ISV's and even IBM'ers to step up to the System i. When I sit in executive level sales presentations and hear the answer to "what is your preferred platform?", even IBM'ers selling Lotus Notes won't answer with System i first... _Perception_ is that if System i is supported, it comes after all the other platforms the vendor supports, and if there's an issue, it won't get fixed on System i first (or even second...)

[R.Khoury]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
Charles: I agree with you on the new pain the IBM name change makes for searching. Tho' with Google, you can use one search instead of two. For example, if I'm searching for forms software for the AS400 (still my favorite name), I'll enter this: forms software (as400|iseries) The items within the parentheses separated by a pipe are treated as OR along with the other search terms. Use it all the time when searching for stuff. HTH

[David Abramowitz]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
The good news is that Only IBMers and people who purchased the box specifically to run Linux refer to it as an iseries. The rest of the world knows it and loves it as an AS400. The problem searching is that you often have to specify AS400 or AS/400 or AS-400 or "AS 400". Dave

[spincode@bigpond.com]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
As told by Zend at Common the PHP implementation is a PASE port not a "native" implementation. Other vendors are also using the term "native" for quick & dirty PASE ports as well. This is not what I nor others consider native code.

[R.Daugherty]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
At least Zend could do a PASE port, which is more than IBM can or is willing to do for most of its code. My initial reading of PASE literature several years ago said that AIX, Unix, was required to compile the code to run in PASE, a Unix environment in OS/400. That I considered beyond pathetic. But I agree with your point as well. Reading the stuff about PHP gives me a sense of deja vu from the giddiest days of the dot com bubble. It's as if people's knees go weak when someone says web page. My primary interface with my web host is PHP. My web site runs on it. I use it in my home page to dynamically show the latest posts on my board. All fine and good. When I researched using a web server to serve my Double Deck Pinochle game that I had rewritten in Java, it would have been far easier to access PHP execution on the host than accessing a JVM, so I had to at least consider rewriting again in PHP because it's easier to access. But what a tradeoff. People go on about PHP as if they don't have a clue what it is and how it works, but because it emits HTML it's like it's magic or something. With us coming from business systems with such concepts as a session environment with programs running that provide secure, high performance business transaction processing you'd think that not having any of that would matter, or at least people would even know they wouldn't have it. But they don't care. Because it puts out a web page. I'll just stick with RPG to generate 5250 or HTML, thank you. When someone shows me something on the level and scale of what we do in RPG but written in PHP, then maybe I'll get a little giddy too. rd

[R.Daugherty]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
I need to clarify that serving my Double Deck Pinochle program written in Java does not mean downloading a standalone game program to a desktop to run but rather serving game screen information to one or more players in each game, that is as in serving 5250 or HTML screens, not a file server. Also, I'm aware of the various wrapper bits in PHP (as well as Python and Perl) used by major league sites such as Yahoo and Google. If someone develops world class custom code server systems as they did but for OS/400 and uses Zend's PHP compile caching technologies as part of the web page wrapper interface, or offers client accessed SOAP services through PHP as they do, then those would be good uses of PHP on the AS/400. rd

[T.Stockwell]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
I specifically asked the technical leads at COMMON during a briefing by Zend. The implementation of Zend core WILL be native. What's out right now is not. Tom Stockwell

[nandelin]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
Considering the amount of work it would take to convert just a handful of the PHP libraries to *PGM and *SRVPGM objects - I have my doubts. I still bet the next ZEND release of PHP will run under PASE. Do the technical leads know the meaning of native? Actually, since PASE runs in an OS/400 partition, as opposed to say a linux or AIX partion, and compared to running PHP on an integrated xSeries server, I suppose you could call that native ;-)

[R.Daugherty]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
I would think that if it's running in OS/400 executing on the AS/400 side of the CPU one would think it'd be called native. From what I understood, it was the memory pointer handling of the AS/400 CPU that made PASE ports of C code not taking that into consideration that was so difficult. On the other hand the PASE environment is different somehow, it's a hosted environment. I thought it was just the POSIX API's souped up to handle all the necessary AIX API's. I don't know, whatever I thought I understood in the past probably has changed by now anyway. But given IBM's attention to PHP, I would think they will provide the same close to OS/400 hooks they provide for things like the JVM and I assume Apache, etc., for high performance. The JVM was under the interface and then pulled back out but still close, that kind of super native, right? Although *PGM native would be just fine. rd

[nandelin]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
But given IBM's attention to PHP, I would think they will provide the same close to OS/400 hooks they provide for things like the JVM and I assume Apache, etc., for high performance. With respect to Java, are the "close hooks" you refer to the Host Servers that Java applications communicate through to access the database and call ILE programs and procedures? When IBM began compiling Java code to machine code, I think folks mistakenly assumed that the JVM and the native virtual machine shared pointers to the same address space to improve performance. But Java programs communicating through Host Servers to access native resources are evidence that the two runtime environments are separate and self contained. I think PASE offers something of a third virtual machine, running on top of the native virtual machine, whereas the traditional JVM runs along side the native virtual machine. IIRC the new 32 bit JVM in V5R4 runs under PASE, which further separates the java runtime environment from the native runtime environment (used by ILE programs). PHP is a virtual machine too, providing an interpretive runtime environment for scripts. If PHP runs under PASE, it will probably use the same communication interfaces used by the new 32 bit JVM. In contrast, the Apache based server IS an ILE based service.

[J.Pluta]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
Nathan, I need to clarify a few of your points. The TIMI, the technology independent machine interface, is what allows RPG programs and other HLLs (and the OS itself!) to be able to move from one set of hardware to another without changes. The TIMI basically separates the software from the hardware, and was crucial in the move from CISC to RISC. The classic JVM actually runs UNDER the TIMI, making Java bytecode more efficient from an OS/400 perspective than even compiled HLLs (this is the same principle which allows SQL to run faster than RPG the longer it stays "down low"), and is why thin JSP architectures like the ones I advocate run so well. However, the base Java bytecode doesn't have support for things like database I/O, and this is where the host servers come into play. The Java toolbox allows access to those host servers to provide basic i5/OS functionality to the JVM. What's really cool about the toolbox is that it can run either on the iSeries or on another machine; on another machine, the host servers are required to enable a non-iSeries JVM to access the iSeries via TCP/IP. Magic happens when the toolbox runs on the iSeries, though. In that mode, the Java toolbox uses a highly optimized JNI interface to many of the basic system functions. And in this case, they ALL use the same 128-byte addressing space of the iSeries, so your point about different address spaces is sort of lost on me. Now comes the new J9 JVM. As you note, this JVM is no longer running in the MI; instead it's designed from the ground up to be very close to the silicon. Because of this, it will be very dependent on the hardware. But since IBM is standardizing on the Power PC line, that's no longer an issue, and so the same JVM used on AIX will be able to be used (with some interface bits) on the iSeries. The J9 JVM will run inside the 32-bit PASE environment. But this is not a bad thing. In fact, this means that the J9 JVM wkill actually be faster -- in many cases, especially startup, SIGNIFICANTLY faster -- than the old classic JVM. The downside is that you are limited to a 32-bit address space right now which means a limit on the heap of about 3.5GB (of course, if you need more then 3.5GB of heap space, you might consider a redesign). A 64-bit JVM is in the works, at which point you'll have to decide whether the increased address space is worth the double size memory costs of every pointer. In any case, there is speculation that the new JVM running in the PASE environment should be equivalent in performance to the AIX version, and that it will be the JVM of choice for SMBs who typically won't need anything more than 32-bit addressing. Joe

[nandelin]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
Joe, your points are helpful, and do clarify a few things. Adding to your points, I'll fall back on diagrams and terms used by Frank Soltis in his book Fortress Rochester. The 64 bit JVM is implemented below the MI (machine interface) in an area defined by SLIC (software licensed internal code). The native virtual machine is also implemented below the MI in an area defined by SLIC, and both ILE and Java compilers generate code that runs below the MI, which Frank Soltis calls Direct Execution Code. My point is that the native virtual machine is separate and distinct from the Java virtual machine, and when Java applications running in the iSeries JVM need resources that are available in the native virtual machine, the requests are passed through a communication interface defined by Host Servers, whereas ILE applications which are running in the same address space use a shared memory interface, which is a lot faster. For example, ILE compilers like the Java compiler don't generate code to access the database directly. The difference is that ILE programs share memory with low level OS routines that perform I/O, so the interface is faster. My understanding is that the new 32 bit JVM has a smaller footprint and will load faster, but there isn't much information available about whether applications running in the new JVM will load and run faster or not. I suspect NOT. The statements coming from IBM about the new JVM remind me of the statements they were saying about the performance of Webfaced applications being comparable to their 5250 counterparts (when run on an appropriately sized server). Nathan.

[J.Pluta]

The System i Strategy for New Customers

** This thread discusses the article: The System i Strategy for New Customers **
I don't think I was clear. The JVM runs in the same 128-bit address space as the rest of the system. Special MI instructions were written to execute Java bytecodes. The reason Java doesn't execute database calls directly is NOT because it's running in a different address space, but that there are no native database instructions for the JVM. In fact, a fun project would be to write special Java bytecode extensions specifically to access database functions. My guess is that that particular enhancement would kick some serious butt. Next, AFAIK when running the native optimizations, there is no communications interface used by the toolbox to connect to the database; it uses direct JNI calls to system database functions. That's one reason why RLA outperforms JDBC quite handily. I'm pretty sure the host servers are only used by toolbox code running on a different machine. And finally, IBM has specifically stated that they believe that the new JVM will definitely start up faster and they believe run times will be equivalent to those of AIX. This comes from Blair Wyman directly. As far as your implications of misinformation I have never heard IBM say that WebFacing would be comparable to 5250 counterparts; certainly George Farr would never say that. Nor do I subscribe to the idea that Bliar Wyman is either being deliberately or accidentally misleading. The former would indicate a lack of ethics, the latter a lack of competence. If you want to call him a liar, that's your business, and while IBM may not be perfect I'd never cite IBM developers as incompetent systems architects. Joe