Feeds

back to article WTF is... scale-in?

I don't get it when IBM talks about "scale-in" for its new PureSystems converged server platforms. And I bet you don't either, and got the same chuckle at it that the rest of us hardware motorheads here out on the Intertubes did. Scalability has been an issue as long as there have been systems, and back in the day when a system …

COMMENTS

This topic is closed for new posts.
Bronze badge

sounds like a mainframe in a box, to me

How many 1U, 1/2 height, dual CPU servers could you fit in a AS/400 chassis? 30? 60?

My guess is that there'll be an on/off switch on the front, and a couple of 10Gig connections on the back, along with an IBM maintenance port so when you need more power, they log in and enable another internal blade..... I wonder how long it will be before all of z/OS runs on blades the size of a paperback book?

0
0
Anonymous Coward

Re: sounds like a mainframe in a box, to me

"I wonder how long it will be before all of z/OS runs on blades the size of a paperback book?"

Sexagesimo-quarto or folio sized paperback as I'm sure they already have the latter in size.

Though IBM do like to call there processing cards books - even in Z/OS land.

0
0

Brad Day is to blame

First there was scale up...buy big expensive boxes

Then there was scale out....buy lots of inexpensive boxes and puts lots of expensive cluster software on them

Finally there is scale-in...buy big boxes again but take all of those little apps and create hundreds of virtual machines.

I first heard the expression when I was talking to Brad Day in 2005. Formally of Giga then Forrester now doing his own thing.

I thought it was a cool way to describe virtualization on big boxes.

tudalu

0
0
Holmes

Assuming it's not a rhetorical question...

...and that you're not just taking the piss out of IBM for what sounds like a backward step...

From the blurb, "scale-in" sounds like scale-up, but allowing consolidation, possibly including virtualisation. Scale down the number of boxes, the footprint, the unused CPU & memory, the management overhead - It actually makes sense!

2
1
Silver badge

Sound like

A mix of appliances and hyper scale servers to me.

Craming more compute power into lower data Center volumes.

1
0
WTF?

Marketing bruhaha

These guys create new buzzword, it's their job...

2
0
Silver badge
Joke

That's the first time...

I see El Reg use a, hmm, "attention seeking picture" in both the overview as the article itself.

Yet now I also know why they usually don't do this, could anyone explain the article to me again, I kinda lost my attention up there? ;-)

2
0
Coat

Re: That's the first time...

That's enough navel-gazing for you!

Also, they got that picture from wikipedia (or it's an amazing coincidence) - http://en.wikipedia.org/wiki/Navel

1
0
tpm
(Written by Reg staff)

Re: That's the first time...

I was just amusing myself on a late Friday afternoon because a half dozen people jokingly asked me what scale-in was.

I know what IBM was trying to say, but it was a silly way to say it.

0
0

Pretty straightforward

By "scale-in" they mean combining network, compute and storage in the same frame. Taking all of the stuff which was externalized in past generations of IT and bringing it all back into the frame. Similar to scale-up, but add the network and storage to the "up" as well as compute.

0
0
Stop

My Understanding

From what I can read on the interwebs and from my Knowledge/experience using IBM systems, what they aim is "cloud computing" and "virtualization".

Apparently Amazon is eating into their business or at least they fear that. So their "pureScale" approach means that you essentially buy a DB/2 and server appliance. All the failover, workload increase/decrease and storage management issues are supposed to be handled by the appliance.

So if you need more query processing capacity, an additional set of CPU cores will automagically be assigned to that purpose. I guess they will also automagically remove that resource when possible.

Their whole pitch revolves around the (pseudo-)distributed capabilities of DB/2. The database engine is effectively supposed to facilitate communication between the elements of the appliance. They claim that application developers don't have to worry about scaling the whole system because of all that.

The problem I can see with all IBM products is that they are highly complex and always have some clunky mainframe-style aspect. DB/2 is basically a mainframe product and you need the friendly IBM customer engineer to run it. IBM software group is too busy creating fancy features for their big-iron customers to even create a working ODBC installer. The customer engineer (which is often on-site for days a week with large customers) will actually create a working installer.

These antics are in my opinion the very reason for the existence of Microsoft and Oracle. IBM still doesn't know how to make a product which works w/o a customer engineer holding the user's hand.

1
1
Silver badge
Pirate

Re: My Understanding

Hope the "expandable" DB2 isn't like the Oracle version, where you have to pay for licences for all the CPUs available, whether they are active or not.

1
0
Flame

Re: My Understanding

You were doing alright until the last two paragraphs.

"The problem I can see with all IBM products is that they are highly complex"

Rubbish. Just a throwaway insult to address your generalization.

"DB/2 is basically a mainframe product and you need the friendly IBM customer engineer to run it."

Bollocks. I ran dB systems from IBM for 16 years without needing a CE at all. I'm not the only one either.

"The customer engineer (which is often on-site for days a week with large customers) will actually create a working installer."

That may be so these days but it wasn't in the past (before the work went East and I had to change direction) and, quite frankly, says more about the quality of the people that are available than about the product - experience talking here!

"These antics are in my opinion the very reason for the existence of Microsoft and Oracle. IBM still doesn't know how to make a product which works w/o a customer engineer holding the user's hand."

Rubbish and bollocks. You may not be able to do stuff but some of us can.

(Irritated? Who, me? Surely not!)

2
0
Stop

Re: My Understanding

The Customer Engineer who helped me with the DB/2 ODBC driver acknowledged that they had lots of issues with Win7 / 64 bit and he himself created an installer. That one worked like a breeze. The CE is in the corporation for two of five days and I know very well why.

The corp has greater 300K employees and my dept is full of CS graduates. We don't have these issues with MS or MySQL products, but they exist with IBM stuff. In another company we had a mainframe and the CE from the Böblingen lab was a regular visitor in our offices. He only came for the mainframe. (AIX was much more MS-like) We also had the issue that the mainframe could not correctly divide integers ! The lab didn't believe it until we proved it to them; so they sent a microcode update.

So I stick to the statement that IBM product typically need a CE on customer premises. Maybe AIX is the exception from that rule.

0
2

Re: My Understanding

DB2 pureScale is hard capped, so you can "black out" cores that are in the server, but not running DB2 at the moment. When you need to increase capacity, you can turn them on and license a new core. Similar to the way System i has worked.

I think you can use hard caps with Oracle DB, but they have all kinds of hypervisor and OS exclusions. You can do it with Oracle Linux and OVM on x86, but you cannot do it with RHEL and VMware... or really anything on x86 other than OEL and OVM. I think they still support hard caps on AIX too. They obviously support it on Solaris and Sparc. Although the Oracle equivalent of DB2 pureScale would include RAC, which is a really strange licensing structure. I am not sure if RAC is an all or nothing.

0
0

Re: My Understanding

No way, a mainframe that can't correctly divide integers. That has to be an application level FUBAR. If it were a widespread microcode issue, no one in the world would be able to receive a correct bank statement.

Mainframes only require an IBM CE if your staff is made up of people that don't understand the mainframe. Most of the bank, insurers, governments, etc run their mission critical transactional systems on mainframe with a fraction of a fraction of the people they have dedicated to patching and tweaking Wintel. AS/400s (aka System i) run for years without being cycled off.... If anything people use an IBM CE because they so infrequently need to do anything to the mainframe that no one internally develops the skills, whereas they are constantly messing around with Wintel and Linux/Unix, so people have the skills.

0
0
Flame

Challenge

Just try to install the ODBC driver and create a working data source from the "official" installer on Win7 / 64bit. I bet you can't.

0
0
Anonymous Coward

Re: Challenge

(for DB/2 9.7)

0
0
Flame

More Complaints

As I can sense the IBM loyalists here, I also challenge them to change a single detail of an ODBC data source in the TCP settings. Assume you have a typo in the host name.

At least using the GUI I have to delete the data source completely, as the TCP settings are "write-once, see-never again". Of course, that's just annoying, but the sum of annoying stuff really pisses people off.

0
0
Anonymous Coward

Re: More Complaints

just change it to half-duplex change the name then change back to full duplex

0
0
Anonymous Coward

Re: More Complaints

Thanks for that tip, I'll try it. Totally obvious to try that :-) I must be a cretin.

0
0

Geek speak is good for sales

If you can't dazzle them with brilliance, baffle them with B.S.

0
0
Silver badge
Happy

Re: Geek speak is good for sales

I once made the mistake of creating my own BS marketting term ("in-frame computing", leaning heavily on the Egenera concept of a completely virtualised system), then telling the reps from IBM, hp and Sun that it was the creation of one of a mysetery other vendor and asking them to explain how their alternatives compared. I thought this would get me a few week's peace and quiet before they twigged. What I thought was a harmless bit of fun turned into a nightmare as IBM demanded the opportunity to present FUD on how "in-frame computing" was a dead-end, how their product strategy was superior, etc, which was pretty impressive considering I'd made the thing up! It was late 2008, when CISCO were starting to plan their UCS campaign, and I guess the IBM guys thought we were being targeted by CISCO.

0
1
Anonymous Coward

Re: Geek speak is good for sales

Be careful how you taught the clueless...

0
0
Boffin

Lock-in

It's like lock-in, but you feel cleverer.

1
0

Re: Lock-in

How is it like a lock-in? You have options up and down the stack and can basically run anything you would run on plain jane x86 boxes. The "patterns"/automation analytics do not require any proprietary version of the software (e.g. VMware for PureSystems), so you can decide to migrate the whole deal just as easily as you could migrate from any other platform.

0
0
Anonymous Coward

Strange...

Is it me, or is it suspicious any controversial article here never gets less than a "so-so" rating?

I wonder if anyone is boosting their own article ratings. ;)

0
0
Anonymous Coward

Re: Strange...

i found the belly button "orgasmic"

0
0
Trollface

Since they are making up new terms...

What about scale-off? You turn off the servers and the application doesn't run at all... I bet all the issues would disappear, it is impossible to have downtime if you don't even have uptime!!

0
0
tpm
(Written by Reg staff)

Re: Since they are making up new terms...

Perfect!

0
0

Re: Since they are making up new terms...

Patent Pending - Dell Corporation

0
0
Boffin

Explanation of Scale-In

Scale-in relates to elasticity. It's a concept that's been around a few years, and it's actually quite relevant to IBM's new systems, although not at all in the way their marketing folks have spun it.

http://blogs.technet.com/b/scvmm/archive/2011/05/18/scvmm-2012-an-explanation-of-scale-in-and-scale-out-for-a-service.aspx

http://blog.codingoutloud.com/2010/07/27/three-types-of-scaling-in-the-cloud-scale-up-scale-out-and-now-scale-side-by-side-with-juxtaposition-scaling/

http://publib.boulder.ibm.com/infocenter/wxsinfo/v7r0/topic/com.ibm.websphere.extremescale.over.doc/cxsoverview.html

Tinyurl for IBM link:

http://tinyurl.com/6o8xkt3

And there's a very nice picture here (Fig. 5):

http://www.cs.uwaterloo.ca/~ashraf/pubs/smdb12voltdb.pdf

0
0
This topic is closed for new posts.