Trade off between multiple small optimised aggregates - few large aggregate

Hi SDN community
I am asking a purely theoretical question to the sdn.sap.com community if any projects have done verifiable tests between the trade off in performance for a small number of cubes with large unoptimised aggregates, compared to that of multiple cubes with small optimised aggregates.
Will queries run faster acrosss more cubes with more aggregates that are smaller, as opposed to less cubes with larger aggregates.
We are currently trying to improve performance, and wish to get feedback on whether to change design to cater for ongoing performance issues.
Thank you for your assistance.
Simon

Hi Ravi,
Thank you for your reply.
The reason why we need to consider smaller aggregates is such:
- We have 2 value types Budget, and Forecasts in the same cube.
Because of this, we cannot restrict the aggregates to 2 smaller sized aggregates to allow performance gain.
- If we sepearate the data to the Forecasts Cube,
We have potential to create 12 Version specific aggregates
- If we create Fiscal Year Aggregates in addition, we split the size of the aggregates
Now although the roll up times will be longer, we have very targetted small aggregates so our reports should run faster.
we have consistent performance problems so i am proposing the above last major performance tuning that can be thought of, but will the performance be worth the expenditure.
Thank you.
Simon

Similar Messages

Trade-off between the one-arm and two-arm WAE designs

We are configuring a WAE (model 512) for a branch office and I was wondering if someone could please tell me the trade-off between the one-arm and two-arm WAE designs..
thanks..
greg..

if you are using WCCP then the WAE becomes the client withing the servcie groups 61, 62. In order to accelerate both vlans then apply the ip redirect 61 in on the client vlan ineterfaces to the one interface.
If inline, you can use both 2 port groups for each client interface or trunk all to a single inetrface and configure which vlans you would like to accelerate.
Now in terms of of using both GE inetrfaces, I would have to check. A topology diagram would help

Some thoughts on trade-offs between usability and user-friendliness

I use Awesome as a window manager over LXDE. Recently, a friend of mine tried to use my laptop. The experience frustrated him somewhat--he managed to start the browser through the menu shortcut, but it opened out of sight under the "web" tag. Once he realized where it had gone (tags are clickable in Awesome), he had already managed to start several instances of it. Naturally, he tried to close them--first looking for the close window button which is not there, since I dont use title bars, and then trying alt+F4 to no avail (the default key command for closing a window in Awesome is super+shift+c). In desperation, he finally launched the terminal and used xkill to get rid of the redundant Chromium sessions.
The whole thing got me thinking. I find my system very usable--as I guess most of us Linux nerds do, as our systems are do-it-yourself projects to such a large degree. However, it is not exactly user-friendly. At the same time, I can navigate my confused friend's system quite well since he uses Gnome--a system that adheres more to the common desktop metaphor of e.g. Windows or OS X. However, I doubt I would be able to get around very well in some of the more minimalist *box setups so popular among Arch users (though, admittedly, it is probably easier to get the hang of a mouse-centered desktop than one that focuses on keyboard shortcuts). What I find interesting is that there is clearly a trade-off. If I was to build a system for more people than just me, I would probably go with Gnome, but I would not find it as usable as my current system.
Which brings me to my questions. Do others here share that experience? How do you go about managing it--different default sessions for different accounts, or a compromise that is more user friendly but less usable to you personally? Or do you force people to relearn and adapt to your preferred way of doing things (which might sound worse than it is--after all, if my way makes more sense once people adjust to it, then why not)?
Last edited by caligo (2010-03-12 09:53:01)

lolilolicon wrote:
mythus wrote:
While I understand your point, I have to fundamentally disagree with your subject. Having memory issues does not make someone brainless, and I overtly object to such. It would be just as bad as me saying people who think user friendly is for brainless creatures with no memory are elitist pigs *shrugs*.
Case in point, myself. I have memory issues. I didn't always have memory issues, having used to have a very sharp memory. But hey, getting hit by semi-trucks and having your head go through a windshield does some nasty things to your brain functions. Does me now having problems with remembering certain things all the sudden make me brainless? I sure don't think so, being that I am still able to think, process equations and goals, as well as teach myself new things and relearn forgotten things on a daily basis. It is just that most stuff I have to write down or print out and have in a huge binder in front of me at all times now. Having memory issues simply caused me to have to adjust to how I do things, not make me into a brainless creature. In fact it was after my accident that I came to try and use linux, and while I do have a certain need of the mouse at times, I also have my printed out shortcuts here at my disposal.
The lesson here, be careful of adding insults to posts when trying to make a point. Without that insult I could have easily agreed with your point.
I'm really sorry if it came out like that, I didn't mean that. My point was not at all about memory you know. I forget things too, and it happens often, and I'd curse myself if it were really bad.
I respect men like you. You managed to learn linux (and it was arch! ++), and resolve your issues, e.g. even you do forget sometimes your shortcuts, you still have your memory of them on the paper, plus backups of it. This is nothing like "brainless creators". What I meant by that was more about the lazy people who never know what they have.
I brought my mood to somewhere extreme, because I was feeling again the reasons why I switched to linux. It was the moment I decided I had been a lazy pig who had hated his computer but never had done anything about it.
s/it's for the brainless creatures who've got no memory/it's for the button lovers ;P/
Sorry mythus, I wish you all the best.
Thank you for your apology.
As I said earlier, I do agree with most of your points, it was just that one glaring sentence at the beginning which took on a role as the subject of your post, which I disagreed with. I do agree that the modern day idea of user-friendliness is for lazy people who don't wish to learn how to use a computer and simply want their computer to know what he/she wants to do and do it without any real input from them. Having to use all ten fingers versus one to two fingers is where that all boils down to. Just imagine having to actually sit up and make full use of both of your hands to use your computer instead of lean back in a recliner and rest your hand on a small plastic device, barely having to flinch your wrist and move your index finger. That is really where IMO the "User Friendly" systems of today are targeted. In reality, they aren't user friendly at all, but lazy friendly. Truly, I am still waiting for them to invent a mind reading device or a device that monitors your line of vision so that if you say.. look at the upper right side of a window it will close it for you, or if you open your eyes real big, it will full screen the window for you. *sarcasm intended*
For a system to be user friendly it has to be completely usable to it's principal user with little or no complications. The user should be able to do his or her work and other computer related activities without confusion and/or delay. Unfortunately, no two users are alike so the method that fits them best wouldn't work for everyone. However what does seem to work for the majority isn't necessarily a user friendly system, but a lazy friendly system that is familiar after generations of being presented the same UI.
It is also because of the lazy-friendly needs of a O/S so that it can be accepted by the largest amount of consumers possible (after all, it is all about money) that advancements and changes to the UI are sure to never come. At least from a corporation.... So you will always be faced with having to decide if you want your computer lazy-friendly and familiar for your friends/family, or user friendly for yourself.
BTW- I do not think that the mouse is a bad tool. It is a highly useful tool. It just should not be treated as a keyboard.

More smaller disks or fewer larger ones?

Hi:
We're in the middle of procuring disks for a new Oracle server. We plan to stripe disks to randomly distribute I/O load across disk heads. There is some debate on whether or not we should use fewer large disks vs more smaller ones. For example, the datafiles for one schema will consume 432G. In terms of performance, would it be better to stripe 6-72G disks or 12-36G disks? Are there any other pros/cons for one scenario vs the other?
Thanks in Advance !

Generally, the more heads you have working the better so more smaller disks would be the way to go. Additionally, with smaller drives, recovery from drive failure would be faster. Of course the major drawback to this approach is that you will need bays for the drives and if you were looking at Server attached storage, you may not have enough drive bays available.
Another consideration would be the storage technology you are considering; for example, the HP EVA SANs require a minimum of 8 drives for a drive set so in your example you would need a mimimum of 8-72G disks whether you needed the capacity or not.

Looking to buy new MacBook Pro for editing with Premiere/After Effects, but wondering about trade-off between Processor Speed and Graphics Card

I'm a professional video editor (using Premiere and After Effects) looking to buy a new MacBook Pro and am deciding between two models. The slightly older model has a 2.8GHZ i7 (3rd generation) Quad Core processor with a 1GB SDRAM of NVIDIA GE FORCE GT 650M Graphics Card. Then newer model has a 2.3GHZ i7 (4th generation) Quad Core preocessor with 2GB SDRAM of NVIDIA GE FORCE 750M/Intel Iris Pro Graphics Card.
Which makes the most difference (processor speed vs. graphics card) with editing with Premiere and After Effects?
Any help/guidance would be greatly appreciated.
Thanks!
mike

Poikkeus wrote:
1. Your MBP will be somewhat slower than your iMac, as reflected in the general speed; desktop Macs have more RAM and storage.
You recon? If he get's the 17", he would have up to 8x more RAM, 4 x more GPU,, and a bit faster CPU;.
2. Be aware of the advantages and disadvantages of extra RAM. Loading up the slot will make juggling multiple applications easier, like Photoshop, VLC, and Safari. However, more than 4gigs of RAM will make loading your MBP on startup twice as slow - at least a minute, probably longer. That's why a MBP user with extra RAM should sleep their machine nearly always when not in use, rather than powering off.
I did not know this, I just upgraded from 4gb to 8gb the other day. Have not noticed it being slower, but I don't often shut it down. It's nice to not even have to bother with ifreemem.
3. Additional storage and RAM will maximize the basic capabilities of your MBP, but you won't be able to make a 2.3ghz machine any faster than it already is.
SSD
4. I still feel that your iMac will be faster than your prospective MBP. The only way to dramatically increase the speed would be the installation of a SSD drive (like the lauded OWC series). But they're not cheap.
I don't want to rain on your parade, but want you to get a more realistic idea of your performance.
I chose a macbook pro, 17" of cause. I use it for gaming. Yes a iMac is better for gaming. But, it's nice to be able to move around. Set up a man cave in the lounge 1 week, or in the bedroom, the next. But you fork out a lot more dosh for that luxury. And yes, not as much power as Poikkeus has said.

Pixel/compression trade off

Picture qulaity appears to be a trade off between the number of pixels you have & the amount of compression that is applied when saving a JPEG .
Is there any gide or rule of thumb that you can provide that will tell which is best for your needs? For instance will a 10MP picture with high commpression be than an 8MP picture with low compression? What about 6MB etc? Is there a chart?
Any comments will be apprecieated.

The Old Fart wrote:
Picture qulaity appears to be a trade off between the number of pixels you have & the amount of compression that is applied when saving a JPEG .
Is there any gide or rule of thumb that you can provide that will tell which is best for your needs? For instance will a 10MP picture with high commpression be than an 8MP picture with low compression? What about 6MB etc? Is there a chart?
Any comments will be apprecieated.
I'll add a few personal thoughts to the advice already given.
- First, nearly all digital cameras use Bayer pattern sensors : there are theoretical reasons you can downsize to about 70% without losing significant detail...
- In many cases, you know the kind of output you want for a particular purpose : printing 4' x 6' or A4 format, displaying on high res displays (1900 x 1080 px) or saving for web (800 x 600 pix for instance). Then the first thing you can do before thinking about jpeg compression is to downsize to the adequate printing size. It's generally good enough not to output to more than 300 dpi or downsizing to the pixel dimension needed for display.
- There are several downisizing method : you might prefer bicubic or bicubic sharper. Some prefer more advanced method (Lanczos, available in Faststone Photo Resizer)
Now for jpeg compression :
- The level of detail depends very much on the noise level : you are losing details with noise anyway, so the denoising will be a compromise between detail and noise, and that will be important for the jpeg file size.
- The final size of an image depends highly on the level of detail of the scene and on the noise as stated above. For the same pixel dimensions, your file size may vary from 50% to 100% depending on detail and colors. If your purpose is to save with important compression, 'save for web' is your friend. If you only want to print without visible quality loss, use higher compression settings : for instance try the difference between 12 and 10 : you might gain 50% size without visible quality loss.

My iphone 4 will not unlock when I press the home button, I have to alternate between the home and off button multiple times before my home scree will appear

My iphone 4 will not unlock when I press the home button, I have to alternate between the home and off button multiple times before my home scree will appear....please help!!!???

Do a reset
Reset: Hold down the Sleep/Wake button and the Home button at the same time for at least ten seconds, until the Apple logo appears. Note: You will not lose any data

Trade-offs of different Immutability designs

Hi,
I'm wondering what the trade-offs are for the following immutability designs:
1) Object composition: a mutable class wraps an immutable class without any sort of inheritance. Example: String vs StringBuilder.
2) Single interface with optional operations. Example: List throws UnsupportedOperationException when someone tries modifying an immutable list.
3) Object inheritance: a mutable class extends an immutable class. Example: MutableString extends String by adding mutation methods to it. I could not find an example of this in the JDK.
4) Are there other possible designs you're familiar with?
Thank you,
Gili

Why would you need to do this? Why couldn't the mutable interface inherit from the immutable one?I was originally talking about classes where if the super is immutable then you quite literally cannot make it muttable (like String).
But to answer the direction of the inheritance tree problem (mutable->immutable or vice versa), if Mutable extended Immutable it's then fundamentally not an Immutable breaking the "is a" rule, you have this problem with both directions of inheritance. More reason to steer clear of it I think.
Deciding what modifiability to return/accept is an even more tricky business with both inherited route and with distinct classes, for example imagine the .iterator() method you would have to duplicate the methods to: mutableIterator() and immutableIterator(), this would plague your API designs and make it a royal pain to use.
I meant unmodifiable class, not immutable.The discussion works for immutable and unmodifiables, though I think that converting between (im)mutables becomes even hairier (should be avoided) as users are lead to believe the internal class will never change, whereas unmodifiables allow the internal class data to change just not via API calls.
I should explain that my "interest" here is the best way to introduce "turning modifiability on and off" (which is only a slight deviation to the original post). In my case, I've decided it's a must have feature.
Why would the caller ever need to know whether a Collection is mutable or not?Where we've experienced this issue we've been designing an API that's not just a collection, more like heavyweight resources, where:
-in certain cases we don't trust the user to give them access to the modifiable
-where they can only receive/read from the resource and they should not alter it
-to let them register interest in the resource before it's construction is finalised - ensuring they don't interrupt the finalisation process.
-the API is about the same size as the collections API, which I think is bordering on the (to) largish size (if it's a small API I'd consider distinct classes).
The decision was made to give them the ability to discover whether the resource was unmodifiable for two reasons:
-code could be written (annoying that it's not "have to be" I agree) which would then be guaranteed not to fail unexpectedly later on (assuming compliance).
-not having it meant that users are subtly encouraged to to write large chunks of code within an �ber try/catch to handle possible failures, which has it's obvious disadvantages.
I also believe there should be some method of discovering whether the resource/collection is unmodifiable. Say you want fail fast behaviour for example, you have to call some modifiable method and deal with the resulting exception if it failed. And (slightly worse) try and undo it if it did work - which in some cases is impossible.
In most cases (e.g. collections being used) this isn't required, as generally the collection's modifiability will stay the same for an apps lifetime and good testing (or the first time the error is discovered :) will out any mistakes which can be fixed once and for good.
I think that Josh got it right when designing the collections framework, I can't see a better way of doing it given the size of the resulting API and ignoring the "not being able to discern modifiability" problem. Adding anything other than a tagging interface or simple checking method would have resulted in a seriously bloated API.
I think the reason they didn't do anything about trying to discern modifiability was that, as collections don't change their modifiability (accepting composition), they probably decided, as you said, there's no benefit to adding the ability to check.
I followed the link btw, it's interesting to get other peoples take on these issues, I constantly worry if I'm going down the right route. I am trying to get a few of our projects made open source so I can get other peoples feedback/input for that very reason.
Wowsers, I also didn't realise I had this much to say about it, sorry.

Query performance - A single large document VS multiple small documents

Hi all,
What are the performance trade offs when using a single large document VS multiple small documents ?
I want to store xml snippets with similar structure in a container. Is there any benefit while querying, if I use a single large document to store all these snippets against adding each snippet as a different document. Would it degrade the performance when adding an xml snippet each time by modifying an existing document?
How could we decide whether to use a single large document VS multiple small documents?
Thanks,
Anoop

Hello Anoop,
In case you wanted to get a comparison between the storage types for containers, wholedoc and node, let us know.
What are the performance trade offs when using a
single large document VS multiple small documents ?Depends on what is more important to you, performance when creating the container, and inserting the document(s) or performance when retrieving data.
For querying the best option is to go with smaller documents, as node indexes would help in improving query performance.
For inserting initial data, you can construct your large document composed of smaller xml snippets and insert the document as a whole.
If you further want to modify this document changing its structure implies performance penalties; more indicated is to store the xml snippets as documents.
Overall, I see no point in using a large document that will hold all of your xml snippets, so I strongly recommend going with multiple smaller documents.
Regards,
Andrei Costache
Oracle Support Services

Firm or Trade-Off Zone Indicator not set

Hi
I have created the scheduling agreement and run the MRP by this the schedule lines are generated but the system has not set Firm or Trade-Off Zone Indicator how do I go head? IS there any customization/master data required to set the this indicator?
Please suggest.
Regards,
Prashant.

This depends a bit on how you communicate to your vendors.
Assuming EDI, then the message to the vendor will contain a few date fields
A. ABFDE End of production go-ahead
Which is the 'end' of the FIRM ZONE
B. ABMDE End of material go-ahead
Which is the 'end' of the TRADEOFF ZONE.
Upon receipt the vendors system will recognize these dates.
Assuming printout, you need to work in SAP-Script/Smartform to detail the effect of these date on your print.
Either by also mentioning these in the item header (remember each item has an unique FZ/ToZ setting). Or by using special markers on the date lines, based on these dates.
Regards
JP

Sharing SharedObjects between multiple .swfs

Relatively simple problem; I have a project that I think may be more manageable if I break it up into multiple small .swf files rather than having a gargantuan mega .swf.
Now, I know that I could use a root .swf and load data from the others, but I think that it'll be a lot easier to debug and update if I'm using separate self-contained .swfs, so this is the direction I'm currently investigating.
However, the .swfs won't really work unless I can have user-data persist between them, so I'm wondering if I can load a SharedObject saved by another .swf file? Are there any important caveats that I need to be aware of?
In particular, can anyone give me sample code that would allow me to do this with both an online and offline copy of the files; i.e - if I run the same set of .swfs from haravikk.com, and locally, I want them to still be able to share data with one-another. In this case though I won't need to worry about sharing data between a local copy and haravikk.com, just so long as the files themselves can pass data within whatever domain they're loaded it should be fine.

To do this I still need to make sure that all .swfs are retrieving/flushing their SharedObjects to the same domain though, right?
I'm curious, is there a good way to work out what the right domain is dynamically?
For example; say I have "Menu.swf", and "About.swf", both store in the same folder "MySite", so the files are located at "MySite/Menu.swf" and "MySite/About.swf" respectively. To avoid tying the files to a particular domain-name (so it can still work offline) I'd like to grab the current location of the .swf, and remove the file-name so that I have a domain containing just the enclosing folder.
Is there a best way to do this? Basically so I can have my .swf files using a consistent domain for storing/saving, while still supporting relocating the files, so they will work offline if I copy them to my local machine for example.

PLL Library Size - Few Larger Libraries Or More Smaller Libraries

Hi
I am interested in people's thoughts on whether it is better to have fewer larger libraries or more smaller libraries. If you do break into smaller pll libraries, should the grouping be that some forms will not use modules in library, ie do not have to attach to all forms. For common modules that all forms require access to, do you achieve anything in having many little libraries, rather than on larger library?
Is it correct that pll libraries are loaded into memory at run time when form is loaded and library is attached?
What are issues to consider here
Thanks

Hi Till,
My honest opinion...do not merge the libraries. Switch between them using iPhoto Library manager and leave it at that.
A 22 gig library is way too big to run efficiently.
Lori

Space between multiple objects

Is there a way to alter the space between multiple objects without changing the objects themselves? You know, kind of like inflating/deflating a balloon with dots on the surface (yes, the dots would get bigger/smaller, but I think you get the idea). I've tried Transform>Move>Distance, and Effect>Distort & Transform>Transform>Move>Horizontal/Vertical, but both methods just moved all selected objects away from the origin point - no change to the space between objects. any ideas?

...you'd think there'd be an easier way...
Are you talking about objects that already have equal spaces between them? If so, use the Distribute Spacing function in the Align palette. You can specify the desired horizontal/vertical distance between an array of objects.
JET

Cross dissolve transition between multiple video tracks?

I'm wanting to cross dissolve between 2 clips, with each clip consisting of 2 or 3 video tracks (one clip is a still image in track v2, sitting above a color matte in v1; the other clip is motion footage in v1, with another cropped motion image in v2 and v3). When I put cross dissolves between the 2 clips (ie 2 dissolves, one for video track v1 and one for v2), the dissolve is "messy", showing the cropped images in v2 and v3.
To make the dissolves "clean" I can render and export the multiple-track clips, but then I have to add handles to allow the dissolve, and then it takes time to resync each clip, which is very annoying!
Is there a technique for making cross dissolves between multiple-track video clips?

Any other suggestions? <</div>
Practice and logical deconstruction.
If your original clip is 10 seconds and your nest is 10 seconds you must trim off 2 seconds of the nest to perform a 2-4 second dissolve. If that pulls too much of your original clip out of the timeline, you must add two seconds to the original clip so the nest is 12 seconds long so you can trim off the 2 seconds requried to perform the dissolve.
An easier way of accomplishing this is to stack the two nests and apply a keyframed opacity ramp to the upper track. This will help you visualize what you're doing wrong and how to add more handles to the upstream clips.
bogiesan

How do I use the time capsule to share itunes music between multiple apple devices? Also, is it possible to control the music on one device using another, and how do you set this up?

How do I use the time capsule to share itunes music between multiple apple devices? Also, is it possible to control the music on one device using another, and how do you set this up?

unless i'm missing something, i think you got mixed up, this is easy google for walk throughs
i'm assuming this is the new 3tb tc AC or 'tower' shape, if so, its wifi will run circles around your at&t device
unplug the at&t box for a minute and plug it back in
factory reset your tc - unplug it, hold down reset and keep holding while you plug it back in - only release reset when amber light flashes in 10-20s
connect the tc to your at&t box via eth in the wan port, wait 1 minute, open airport utility look in 'other wifi devices' to setup the tc
create a new wifi network (give it a different name than your at&t one) and put the tc in bridge mode (it may do this automatically for you, but you should double check) under the 'network' tab
login to your at&t router and disable wifi on it
add new clients to the new wifi network, and point your Macs to the time machine for backups

Trade off between multiple small optimised aggregates - few large aggregate

Similar Messages

Maybe you are looking for