System iNetwork

I recently participated in a couple iSociety chats with System i managers where I asked whether IBM would continue reporting System i revenue separate from other system revenue. That particular question was ignored, while my other questions were answered. I took that as a hint that at some point in the not-too-distant future, IBM might stop reporting revenue on a platform by platform basis as they did in the past, and these blade servers may be one reason for that.

As Chris mentioned, the traditional revenue model doesn't fit. Actually, even when IBM sells a System i server with an integrated System x server, do they divide the revenue between the two divisions? If so, then how? According to the disk partitions for the two platforms (though how would IBM even know that)?

In the case of blade chassis with say 14 bays like the BladeCenter H that Chris referred to in the announcement, is it really meaningful to classify the hardware revenue between i,p, and x servers that may be housed there? What if a Power blade began with AIX or Linux being installed, then later was switched to I5/OS? Blade Centers are often attached to shared disk arrays. If part of the array is used for I5/OS would that represent I5 revenue? And how would IBM know that? The traditional server revenue model just doesn't hold up.

Bottom line, it appears to me that opening up blade center chassis to I5/OS is the thing that will force IBM to begin pricing I5/OS separately from the hardware that it runs on. If I5/OS is separately priced, does that mean the revenue will be separately reported? Well, it will be interesting to watch.

"...But the main challenge comes back to revenue. IBM has two choices -- squeeze System i customers for all the revenue IBM can get . . . and hope they remain loyal to IBM technologies forever, or bite the bullet, take a major hit on the existing revenue model . . . and change the game forever. ..."

Chris, you are doing a great job. The faction in IBM responsible for the gearing down and over pricing of the i5 have to be retired. i5/OS can support thousands of users. The System p can run the work of those many users. Each user is a solid revenue stream - i5/OS, web query, rational, websphere, db2 extensions, client access ...

You have to worry that the IBMers responsible for 2+ yrs of falling i5 sales ( Jarman?, Herring?, Shearer? ) are still in place and will prevent i5/OS from being successful on the JS22 blade.

-Steve

I don't understand anything about this, no matter how many times someone says Blades.

It's hilarious. It's like people selling this, whatever it is, think that there are customers out here salivating over the prospect of "leadership" this and that, mentioned in every other sentence of IBM's announcement of that which I don't understand, whatever it is.

Roughly, IBM would price systems on workload capability, which I think is basically what they do. Although $10k was mentioned as a starter, as I recall from a previous attempt to figure out what a 515 would cost, I posted that it sounded like $10k was a starter price for that, so I don't understand what's fundamentally different here. (Other than the starter price appears to be worthless to even mention.)

I did a search on the product name and some analyst was comparing a $50k Blades (with AIX no less) to an $100k HP system, but HP had more peripherals, not comparable, yada yada, but point is he didn't even see it as suitable for a small business. Not entirely sure how it got from $10k to $50k but I'm sure it involves some processor "entitlements" IBM talks about.

All I know is, if I don't have a clue what we're talking about, there can't be very many customers out there that do either.

rd

I'm with Ralph. Have I been living in a box? (No pun intended...or is it?)

My question again is ... when I'm sitting in front of a bunch of executives and I want to tell them how great the System i really is, and if I say "It runs on a Blade Server", they will have to know more about blades than I do because I wouldn't know what the heck I just said.

Can someone explain in real language what we're talking about?

I agree with Ralph - much of this is meaningless. I run a software house, and explaining System i pricing compared to other platforms was, and is increasingly a nightmare.

All I can see in favour of blade technology is the 'green' perspective (important I think), but the rest is indecipherable rhetoric. Sorry.

AWA

We are currently sizing a 4 processor P6 570 to replace our two 825s.

Basically we will need the extra horsepower from the middle of next year on. The reason we are looking for this now, instead of waiting, is because after dec. 8 the 825s are no longer upgradeble and we will lose approx. 350k (of 1.3M for the 570) :-(

Reading this really makes me wonder if it is wise what we are doing...

Greetings, rlc

[*Note From Chris: Hi Richard! With this kind of decision on the line, you've got to talk to a knowledgeable hardware reseller or IBM, but I'm doubting that i5/OS blade would make sense for you right now -- especially since they are months away from delivery. Plus, it's highly unlikely that IBM is going to offer i5/OS blades that will drastically undercut their existing offerings, even though the AIX blades seem so cheap. Eventually, yeah, but not in the next few months. Plus, if you don't already have a BladeCenter or blade-savvy expertise on hand, moving from an integrated and easy-to-manage System i to a BladeCenter, which will not be as easy to manage as a dedicated System i, will likely wipe out any processor-based savings you might see in the near term. Plus, how's your storage strategy fit with a BladeCenter? With many existing iSeries-System i shops, it's simple now, and it's just not as simple with blades. There are tons of unanswered questions and a lack of available i5/OS + BladeCenter expertise in the market.

As I see it, i5/OS blades offer some interesting options for new applications and workloads for existing and new customers . . . but not a rip-and-replace option for existing customers. Too much risk, not enough information, and zero pricing details are available. ]

The current pricing model of the system i productline is indeed hilarious, and in fact, it is wacko on 2 fronts.

Only looking at raw computing power (the mips, the disk i/o per second, whatever), the pricing of raw system i power versus (w)intel gear is roughly 10:1 for the same raw power.

And the second front where the system i pricing is wacko, is to look at hardware and applications. The system i doesn't support current common pheriperals (like an usb connection, do they support a DVD writer yet?) and in addition, the system i o/s is totally green screen (hehe, doesn't sound too bad, greeeeeen screen, must appeal to the environmental sensitive people).

If ibm decides to descend from it's high cloud in the sky and slash its prices to more reasonably levels for the masses, it sure is gonna raise some bad blood with current customers.

Hi Chris,

I agree with you. I am sure it's nothing for short notice. But ofcourse it makes you think.

Richard

I believe the green screen is not relevant in this discussion. There's no 5250 unless one specifically orders it and pays extra (the interactive tax to attempt to force people to WebSphere).

I determined that from looking at the specs of the 515. As far as I've read here, system administration has been replaced with Navigator and that boot up console program, etc.

Unless 5250 is still required to administer the i5, any mention of green screen is entirely irrelevant.

rd

I attended a user group meeting where Frank Soltis was the speaker. When someone in the audience complained about the price if iSeries disks, Frank Soltis simply suggested ordering from the pServer division. The point being that the disks were identical, and only the pricing was different.

The significance of these blade announcements is that for the first time in its history, the i5 could essentially be a plug-in component of a larger system, which potentially included a combination of AIX, Linux, and Windows servers, and network storage (disk drives) grouped together in one chassis.

This announcement is a logical step in IBM's trend toward server homogenization. As the i5 loses its distinctiveness when compared to other servers, the i5 price will come down. Whether that's good news or bad news remains to be seen.

We're seeing a trend away from the traditional business model where all the software, hardware, developer tools, and anything else you'd ever need would be bundled together in one package. If the trend continues, then at some point in the future, most of the hardware will be commodity components, and most the software will be sold separately, rather than bundled.

Traditional i5 users hate having more choices (added complexity) from IBM. But that's what we're seeing.

To cut a long story short on the value debate, I bought a couple of second hand servers from a System i shop I had an assignment for (actually the IT manager wanted to get rid of these machines, so I was being nice).

I bought a Dell Poweredge 2900 Tower (2 Intel Xeon processors, Quad Core, 2GB, 3x73GB disk 15K, came out in 2006) at $950.00 (new value $2,500.00).

I bought a System I model 250, (50 CPW, 4x9GB, year 2003) at $50.00 (new value around $25,000.00).

These prices looked like a very fair deal to me.

ugeerts, I cannot understand you.
How can one compare residual value of goods so differently aged?

"Frank Soltis simply suggested ordering from the pServer division. The point being that the disks were identical, and only the pricing was different"

This statement from Frank, which I assume is made in good faith, is however entirely false.

I once had the happy occasion myself in 2002 to order from IBM 2 disks, one for an IBM Xserver (Intel x86) and one for an AS/400, both were high capacity drives -at least back then- of 36GB.
When the disks arrived, I noticed they looked identical except for a difference in the carrying trays, but the label in the disk itself said "Ibm Ultrastar 36GB" and every following line was completely identical. The AS/400 disk was priced around a shameless $3,000 while the Xserver disk was priced roughly $300.
It was not my money, but you bet it sure felt like a rip off!

When I asked a service engineer about this, he said the AS/400 disk could operate in an Xserver or Pserver, but not vice versa. Why? IBM deliberatly altered the microcode on the AS/400 disk, and actually yes, the last line on the label was a different micro code version. When I asked if there was any economic justification for the 10times price difference (like a 10x longer life span maybe?), he said, nope, the disks are mechanically completely identical.

So translate this to: another tax hike of the Rochester guys.

PS: Could somebody confirm these price differences are still true today? I assume we are then talking about high capacity drives of 146GB at 15000 rpm.

I personally welcome this addition. I think really in the long term it makes i5/OS even more viable.

There are those of us who love our System i in its present form, however, we must look to what the market is doing and how System i fits. I think the message is that we're talking about i5/OS here. System i isn't running on a blade, i5/OS is.

i5/OS gives me the environment I need to do what I need to do.

I would align two buzzwords for executives.

Blades are "Green".

I have read both your articles concerning the iBlades and I am amazed at the issues that no one has mentioned.

The entry price list price for a model 515 is $ 7,995. This compares to your JS22 Express configuration with AIX for $10,363. That Express form factor has 1 drive whereas the 515 has 2 drives. Of course, the 515 is a 1-way POWER5, but IBM has stated publicly that the low-end of the System i will be refreshed to POWER6 in the first half of 2008.

What everyone has missed is that a BladeCenter requires EXTERNAL DASD. The disks in the blade itself are merely for load source, NOT operations. Therefore, you will need to drag along an IBM DS43xx, DS47xx, DS68xx or DS8xxx. These DSxxxx boxes are essentially POWER5 boxes running AIX, i.e., they are System p storage boxes . . . a SAN.

Most System i customers do not manage storage. The system does all of the storage management. However, storage management expertise is a requirement with ANY BladeCenter. Someone will need training in this SAN.

Storage management on the System i is concerned with the number of ARMS not the capacity of the drive. Anyone knows from AS/400 days that given the choice between having a system with one 288 Gbs drive or 4 - 72 Gbs drives, that the latter (4 drives) is much preferred because of how System I manages storage. When you attach a SAN, the philosophy is to go to larger drives with fewer arms. This can wreck havoc with performance.

Today, when the System i is attached to external storage, be it IBM or EMC, the internals of the system treats this as if they are directly attached 9337 drives. Remember the 9337 drives? They were discontinued many, many, years ago.

Customers concerned with performance know that internal (now known as integrated DASD) both outperforms and is LESS expensive than external storage. That is why EMC and IBM have had such low market share in the System I marketplace selling SANs.

The iBlade is positioned for customers that want DISCRETE systems. These are customers who have not bought into the idea of virtualization of resources. Fully 70% of System i customers with 550s and above, use virtualization (LPAR) technology. The iBlades will not do such since i5/OS does not own the resources. I5/OS is a client to the Blade Center. This undoubtedly impact the idea of single-level store, part of the value proposition for this product.

IBM should be pressured to bring down System I prices. The user-based pricing models for the 515 and 525 do not make sense and IBM should let the number of users be set by the marketplace and the application. System I disk prices need to come down too.

Ervin,

I think you've made a very logical presentation . I've often wondered how iBlades could fit with the System i market. I'm especially intrigued by the point about virtualization and System i LPAR. It will be interesting to see how a lot of this plays out once IBM brings the product to market--until then it's all pretty much conjecture.

On another point you say, "IBM has stated publicly that the low-end of the System i will be refreshed to POWER6 in the first half of 2008." Can you (or anyone else) point to anywhere that documents this? I've heard it several times from various sources, but I've never heard anything actually from IBM. It would be nice to have this confirmed.

IBM's price conundrum looks a little different when the idea of "Storage Management" is added. The IBM Ultrastar 36GB was priced, in Ugeerts example, at $3,000 while the X86 storage that looked exactly alike was $300.

What wasn't mentioned is that the $300 disk was part of a system requiring a Server Applications Specialist while the $3,000 was for a self-managing system. This is just an example of IBM sharing with itself the Human Resource savings with the System i.

-John

Michael,

The plans that IBM has for moving POWER6 processors into the existing product line have been articulated by numerous IBMers. I recall one users group where Mark Shearer, the General Manager of the iSeries brand outlined these plans. In other cases this message has been presented by IBM Business partners, IBM Business Unit Executives, and IBM Strategic Workloads Specialists.

Some of IBM’s plans get encapsulated formally in Statements of Direction, while others are carried informally by IBM Executives and sales or technical personnel. Customers wanting to know IBM’s directions can ask IBM or their BP to deliver a trends and directions briefing.

Of course you can always direct these questions to IBM executives like Mark Shearer. His email address is shearerm@us.ibm.com.

John deCoville wrote "IBM's price conundrum looks a little different when the idea of "Storage Management" is added...."

John, as far I recall for this particular disk issue, they were both installed to add capacity and were delivered as CIF's (customer installable features). As I recall, installation on both as/400 and xserver was painless and the servers didn't even have to come down.

As far as storage management, I'm not sure what you mean by that. In the Xserver (x86) world, about 60% of databases run on Oracle, and Oracle nowadays manages it's own storage fully automatically (all disks act as one storage container) and the tuning it does is based on a user defined policy. In the distant past, several Oracle DB administrators may have been needed, but today, this is no longer the case.

As for the comment about the latest Oracle DB not requiring several DBAs. That is a very deceptive statement. DB2/400 has never required DBAs but I know of not a single Oracle account where at least one DBA is not present. Even Oracle does not make such a claim!

Oracle on the System i only runs under an AIX or Linux LPAR. Of what relevance is Oracle to the typical System i customer? Besides, Oracle disk management is relegated to the Oracle environment. A survey done a few years ago pointed out that most customers have at least 2 ERP systems. What about the other applications? What about the operating system? I think that Oracle has enough to do to manage itself and at $40K per core customers were expecting more.

The Oracle DBA function still exists although mainly reduced to one, not several, employees as I indicated in my post.

The times that an Oracle DBA had to administer disks, storage, tablespaces are completely gone since that is all automated once configuration is set up. In fact, an Oracle DBA mainly becomes an advisory function for developers or users on how to structure new tables in the datamodel or how to optimize their queries, so hardly a fulltime job anymore.

But this discussion originated from the determination that System i disks, compared to x86 disks, for the same capacity and model, were 10 times pricier at least back in 2002.

"For today, I just did a quick check on Google. An HP model 146 GB 15K (the fastest disk on earth) sells at $450. The comparable System i disk for the same capacity and speed sells around $1,800, so we still have a 4x price difference without any real justification, except that IBM has a monopoly in the System i space and keeps on using it . . . well, can you blame them?"

Comment for the guys comparing disk drive prices -

You're comparing oranges to orangutans.

While the physical hardware of a drive built for a PC might be similar to a drive from an AS/400 - one difference is yes, in the microcode and in the format of the drive itself.

PC drives are formatted to 512 bytes per sector, last I knew AS/400 drives are formatted 520 bytes per sector. This allows the AS/400 to manage the vast amounts of storage available in the AS/400 platform vs. the limitations of other platforms.

I'm sure that there are other microcode differences as well, including predictive failure analysis, that I have yet to find on competing platforms.

In the 'olden' days, one other price point that users often fail to see is the fact that AS/400 hardware came with 24x7 installation by a trained technician as well as 24x7 maintenace during the warranty period. While much AS/400 hardware is now customer install, it still comes with 24x7 maintainance during the warranty period.

If you want an orangutans, go get one. I'm sure you'll be happy, until you understand what you're giving up.

I just did a similar comparison on Google, and I found that the HP model 146 GB 15K (the fastest disk on earth) may sell for $450, but when you include the hot-swap tray, the price is actually $1030. (search for AD206A)

So the prices are much closer than what what was implied - and you're getting more for the AS/400 drive than just a PC hard drive.

The price of System i disks has to come down to the levels of commodity servers -- that is a long outstanding request from the midrange community.

This whole discussion about disk prices is just a part of the overall system picture; the price of a System i blade has to become comparable to equal performance and capacity sized other server platforms.

The law of supply and demand says declining demand for your product while stubbornly maintaining premium prices levels gives you ever declining sales. After 3 years, IBM is beginning to understand this.

There is no question that System i disks are overpriced relative to the x86 market; however, given how the system manages storage I would expect there to be a premium. The question is how high a premium. Over the years I've seen 30 percent as being the one least objectionable.

System i customers should be careful of what they wish for. Over the next couple of years integrated disk may become a thing of the past. IBM is currently funding two separate disk development groups: the System i group (in Rochester) and the Storage Division (in Tucson). Duplication is costly and the Rochester operation is likely to go away. Plans call for external disk to become the standard disk architecture for the System i.

[Note from Chris: Ervin . . . "plans call for external disk to become the standard disk architecture for the System i" . . . do you happen to have a source for that? An IBMer said this? Is it in a document? Or did you hear it elsewhere?]

Chris, the strategy for external DASD on the System i was first told to me by an IBMer in the Rochester Briefing Center. After hearing this, I was told of this issue being raised by Mark Shearer in a LUG meeting (within the last few months). I in turn called another IBMer who is a well-known authority on System i DASD, and he also verified that that is the strategy (but he had some skepticism as to whether internal DASD could go away).

The issues for IBM have to do with costs of duplicate organizations, and for customers, the issues are performance and cost. I'm told that performance will be at par or better than internal next year. Currently, internal outperforms external. And the cost of external is higher today. That should change too.