Question about ASE 15.7 kernel mode – process vs threaded
Hi there,
This question is related to upgrading Sybase ASE Server
from version 15.5 to version 15.7 on Sun Solaris SPARC Operating System
(Solaris 10). For the upgrade, I need to choose between threaded kernel and
process kernel mode.
1> Is the process kernel mode in ASE 15.7 identical to
process kernel in ASE 15.5? Are there any changes to process kernel mode
between version 15.5 and 15.7? With all the other configurations on the Solaris
host being same (Number of CPUs, RAM, Storage, Sybase ASE Configuration
parameter values like total memory, engines), will the performance on ASE 15.7
in process kernel mode be same as on ASE 15.5?
If there are differences in performance, what are they? What should I be monitoring on ASE 15.7 to see
if there is performance issue on the ASE Server?
2> For the threaded kernel mode, are there any performance
overheads like more CPU usage? We are not planning to change the hardware
configuration of the Solaris host during the upgrade. What values should I be monitoring at the OS
level OR Sybase ASE level to make sure that there are no performance issues? Is
there any particular parameter that needs to be monitored closely when we
change to threaded kernel mode?
Are there any Sybase Configuration parameters that I need
to tune for the New threaded kernel mode? If so, do you have any performance
tuning tips (white papers, technical documents) for threaded kernel mode?
3> What is SAP Sybase recommendation on which Kernel mode I
should be using? I researched the SAP Knowledge Base articles for
recommendations on which Kernel mode to use and I could not find a definitive answer.
Thanks,
Raj
Actually with Solaris on SPARC, there is a very key consideration - you should be on ASE 15.7 sp50 or sp100 or higher AND have applied the kernel jumbo patch (Oracle BugID 16054425) from Solaris that fixes the issue in which they prohibited OS threads from issuing KAIO calls. Then in ASE, you need to configure the 'solaris async i/o mode' parameter.
The changes between threaded and process mode kernels should be transparent to your application. Generally speaking, SAP recommends 'threaded' kernel mode - and specifically for SAP applications only certifies 'threaded' kernel mode. For custom applications, you *could* use either - although, depending on how many engines you run and what chipset you are on, you might see better performance with one vs. the other - typically threaded kernel mode is better, although there are exceptions.
Generally, the main difference is that in process kernel mode, each engine did disk and network IO polling. Due to OS security restrictions, this often meant that a engine had to process the disk IO in which it submitted...or for networking, your SPID had to run on your network engine for any network IO. In threaded mode, there is a single parent process with engine tasks (threads) and IO tasks (threads). Consequently, for customers used to running in the 10's of engines, we suggest a starting point of reducing the number of engines by 2 to give the IO tasks CPU time.
One reason why threaded mode is often better is due to OS scheduling. With hardware threaded cores, most OS dispatch schedulers do a better job of spreading the threads across the cores before utilizing the threads when a process is multi-threaded with POSIX threads - then a multi-process/kernel threaded implementation. At one customer, we proved that we consistently got the best performance with threaded kernel and the only way we could match it with process kernel was to use OS cpu affinity to bind the ASE engine processes to individual cores.
However, one consideration is that older SPARC Niagara chips (T1-T4) were notoriously bad at network handling for some reason. As a result, one of the recommendations for ASE on SPARC Niagara chips is to run the same number of network engines as the number of cores ASE is using. One solution to that is in Solaris 11 (I think) where Solaris finally started being able to handle network interrupts better - I also think the T5/T6 are much better chips - but I have no customer experience data to go on there.
Similar Messages
-
What will i do? ITune screenshot is as follows >
There is a problem with your iPhone. Please visit the Service Answer Center to find answers to all your questions about service options, warranty and other processes in your country. To find your nearest Apple Store, click here.Do what it said to do.
"Please visit the Service Answer Center to find answers to all your questions about service options, warranty and other processes in your country. " -
There is a problem with your iPhone.
Please visit the Service Answer Center to find answers to all your questions about service options, warranty and other processes in your country.
To find your nearest Apple Store, click here.
Please help me i can aktivate my phone imei 012423006333181.There is a problem with your iPhone.
Please visit the Service Answer Center to find answers to all your questions about service options, warranty and other processes in your country.
To find your nearest Apple Store
no imei also -
Basic questions about connection sharing and radio mode
I just got TC and have a couple of basic questions.
At home I have a D-Link ADSL modem/router (no wifi) plugged into the wall, and would like TC to provide wifi. I've connected the WAN port to some LAN port of the router and everything works fine. The intented usage of the setup is wifi-only. Would you recommend TC to do DHCP/NAT? Or is Bridge mode better? In the latter case I don't fully understand who's routing the wifi network (though it actually works).
Second question is about radio mode. My wife has an iBook G4 that is not capable of talking 801.11n, so I selected the n mode compatible with b/g. In that mode my black MacBook will still be able to talk 801.11n with TC? The iBook has very low usage, it is fine that computer uses 801.11g for TC, but I'd like the MacBook to go full-speed. (If that's not possible I'd ponder purchasing an 801.11n adaptor of some sort for the iBook.)Would you recommend TC to do DHCP/NAT? Or is Bridge mode better?
It doesn't matter.
In the latter case I don't fully understand who's routing the wifi network (though it actually works).
The D-Link modem/router.
In that mode my black MacBook will still be able to talk 801.11n with TC?
Yes. It will only suffer lower bandwidth if the iBook G4 is actively connected. -
Question about import in archive log mode
Hello.
I am a developer, I have ordered to write a script that makes the import of a schema of a database (release 9.2.0.7). That import will be done once a day. I have seen that in my development environment the import creates 54 archivers files (10M aprox. each), that means more that half a Gb a day, it seems too much to me.
I cannot see why all those archivers can be useful. Would a good way of proceeding the following?
1. Forcing an archiver just before the import (I do not know how to do that) so that a backup could be done to the state before the import.
2. Disabling archive log mode during the import and enabling it just after the import (I do not know how to do that).
3. Forcing a new archiver just after the import (I do not know how to do that).
Thanks in advance.540M is not that much.
One would question why you need to import every day,
there must be better ways to do this.
The three steps you propose look like an awful
scenario, as it would require the database to switch
from archivelog to noarchivelog and vice versa.
This would require the database to close twice a
day.
The scenario is also incomplete as one would need to
take a backup after the import
If you can convince your DBA to close the database
twice a day, he should write the script, which he can
easily derive from the docs.
But likely he will visit Billy Verreynne to borrow
his lead pipe, and rightly so ;)
Sybrand Bakker
Senior Oracle DBAThanks for the answer.
A few things:
- Sorry for my ignorance, I have no experience in database backups, I do not understand why I need to backup just after the import.
- That database is not critical, it is just for a team who will test on that database several applications, so the database will only need to be open during the office schedule.
- I have no dba. The client has a dba for several databases but I have no contact with him/her nor my boss has. -
A little question about compiling my own kernel
Hii, How can I know please, wich moudle is loaded in my current kernel and being used?
for example, if I have now the RAID xxxx , and I don't use it, so when compiling my own Kernel, I don't need it.
How can I know wich ones arn't needed?Ok Lsmod I know, is there any thing else?
-
Question about stacked objects using blend modes
Say there are three stacked objects. The two topmost have a blend mode.
In the area where all 3 intersect, the bottom object interacts with the transparency. Is there a way to limit how many levels the transparency affects? If you could limit it to two levels, then in the area where all three intersect the bottom object wouldn't participate in the blending. But in the areas where it overlaps only one of the topmost objects, it would participate.
I know this can be achieved by creating another object, or using the Pathfinder, I'd just like to know if there is a simpler method to use.The attached PDF explains this better. Page 1 - the yellow and magenta are multiplied on top. But the cyan in not participating in the area where all 3 overlap. But in the areas with only 2 levels, it does participate, so there are the green and blue overlaps.
Since the cyan is negated from the one area, the result is red – yellow and magenta multiplying
The problem is, the only way to get this is to punch a hole in the cyan. So when you move the yellow and magenta around, the red doesn't follow (see page 2)
It would be nice if the blend modes included some sort of level limit. Default would be unlimited, but you could change it to 2 or whatever number you want. In this case if the level was 2, no hole in the cyan would be required.
As far as I know Adobe has not added to transparency capabilities for a long time. You have the 16 bend modes, and opacities. And knock out group and isolate blending, which are good features. But maybe it's time to expand on this and add more points of control. -
Question about iPod 4th Gen Diagnostic Mode
I am having a problem with my iPod. 4th Generation, 40GB Model A1059.
I am interested in what should be the expected results in two of the Diagnostic modes.
HDD R/W; and
HDD SCAN
In the Scan test mode, I get a "Pass" message and when I press "play" I go back to the diagnostic mode menu. So I guess this means I passed the HD Scan.
In the R/W test mode, I get a message that reads
HDD Test
38 GB
But pressing 'play' nothing happens, and it seems (feels and sounds)like the hard drive is still spinning.
So is the test still underway, or am I finished with the test when I see the "38 GB" message on the display?
I did a search of this site and the internet and really can't figure out what the anticipated result should be. It just won't let me search under "HDD R/W" It seems that the / is some kind of boolean conector that the search engine reads as something other than a ASCII character.
For anyone who may be interested in my problem, here it is:
It used to run forever with the Apple Logo. I put it away and came back to it recently hoping the problem would have magically gone away. Well it hasn't. Now the HDD just 'clicks' It goes something like 1,2,3,"click"... 1,2,3"click" until the file folder with an exclamation pint comes up.
Interestingly enough, the 1,2,3,"click"... 1,2,3"click" happens in the HD Scan Mode Test, but when I run the HDD R/W mode test the drive seems to spin freely without the 'click'
A1059 Windows XP
A1059 Windows XPTry:
- iOS: Not responding or does not turn on
- Also try DFU mode after try recovery mode
How to put iPod touch / iPhone into DFU mode « Karthik's scribblings
- If not successful and you can't fully turn the iOS device fully off, let the battery fully drain. After charging for an least an hour try the above again.
- Try on another computer
- If still not successful that usually indicates a hardware problem and an appointment at the Genius Bar of an Apple store is in order.
Apple Retail Store - Genius Bar -
Help
Hieu Apple wrote:
Help
We'd love to. But you haven't told us anything about your issue. This website/forum does not support HPML (Hyper-Psychic Mindreading Logic). -
Question about using uuid in kernel boot argument
Have my root partition on sda3 with uuid-> 81222792-bd9f-47d6-8d28-b4b59d93cbac
Grub is able to boot if I use either:
root=/dev/sda3
or
root=/dev/disk/by-uuid/81222792-bd9f-47d6-8d28-b4b59d93cbac
But not by using:
root=UUID=81222792-bd9f-47d6-8d28-b4b59d93cbac
Do I need to do some steps if I want to use root=UUID=xxx syntax in the root partition of the grub/grub2 bootloader?Found this link on Google http://en.gentoo-wiki.com/wiki/Initramfs. In short it tells that you need some support in the initrd image. Since both kernel26.img and kernel26-fallback.img failed to boot using root=UUID=xxxx syntax, then I guess that support is not included by default on our images.
Since working/hacking the initrd/mkinitcio is atm too much for me to handle, I hope someone with more experience can give me some light in the issue. -
Please help me
You do not say what your problem is...
Go to Service Answer Center - iPhone and select your country from the drop-down box if that is all you want to do. -
these is the message i get on itunes screen, when i active my iphone 4 which was working fine i just updated IOS to 6.1. could you please help
You do not say what your problem is...
Go to Service Answer Center - iPhone and select your country from the drop-down box if that is all you want to do. -
Stupid questions about keeping important calls on the FX application thread
I'm working on a project that I've been developing for about the last 6 months. I'm using JavaFX as the graphical side, but I'm taking input from a music processing program called Max/MSP, and I'm getting user input from a camera. Both of these things need to run on separate threads from my FX application thread. Up until now, things have been going smoothly. I can manipulate things on screen from my user input, and I have two way communication working between my JavaFX application and Max. Now, however, I need to start actually drawing stuff at runtime. Up until now, pretty much all FX side manipulations of been to just make translations. I now need to create, add, and remove children from Nodes. So I'm of course getting an error saying "not on FX application thread". I understand why it isn't on this thread, and why this is a problem. So I made some changes, so that calls that will make changes to the nodes on the stage, will only note the data, and then make the changes later, on the FX application thread. And then I run into problems, involving my own stupidity...
I have no idea how to get these calls onto the FX application thread. I mean, this should be really easy. I'm obviously missing something. Or maybe I'm just really tired, and have been looking at this for too long, or maybe the fact that it's been 6 months since I wrote this part of the code. But I see my Mainstage class, which extends Application. It has a start function, which sets up all my on screen do-dads (extended from Nodes and such), calls stage.show()... and then seems to end. When I've done things like call translate functions on my nodes from my user input thread, or from my communication from Max thread, things seemed to have worked. But now I'm thinking I might have been handling that wrong from the ground up. It seems likely that I should be having a "main game loop" (it isn't actually a game, but still, same principle) where I can tell the stage to update all the various nodes with the changes that have happened since the last frame, and try to get everything to update in time for the graphics to render at a reliable frame rate. And yet, I'm just not seeing how to do that. I'm sure this should be obvious, but I'm flat out blanking. And the time crunch is starting to descend... Please, if you can point out some standard way of avoiding this hopefully common newb mistake, I would greatly appreciate it!
-cullamCheckout the documentation on "Pulse" in the JavaFX Archictecture overview which may help clarify some of your questions.
http://docs.oracle.com/javafx/2.0/architecture/jfxpub-architecture.htm
In simple terms, as I understand it, the framework will check 60 times a second if anything is dirty and needs repainting. If everything is clean, it will do nothing. If something is dirty, it will repaint it (although the repaint is kind of clever and highly optimized to support region image caching, make use of hardware rendering, etc).
So you can see that the main rendering loop is essentially hidden from you. Instead you construct a SceneGraph, the system checks regularly whether or not the SceneGraph has changed, and if it has it renders the updated SceneGraph.
In most JavaFX applications, you have a UI with controls and add event handlers which respond to control input to update the SceneGraph. In such programs, creating a main loop is unnecessary.
You can set up an animation timeline say at, for example, 30fps (half the pulse rate), then, on every other pulse, nothing will have changed so no redraw will happen. This is only necessary if you are performing an animation or capturing camera input, etc yourself (rather than using a Transition for example).
Actually, if I get my state changes working properly in a runLater call, do I need to also write a main loop? No, you don't need to write a main loop. Depending on the exact application, you may want one, but if you are already getting events generated by your camera input at 30fps, processing each 30fps from the camera in a runLater call would work fine without creating a loop.
But you do need to be careful that you don't send too many runLater calls and eventually overload the JavaFX event queue. For example it makes no sense to make more than 60 runLater calls a second to update the scene, because the system will never update the scene more than 60 times a second anyway.
If you do get a working camera app integrated with JavaFX, please post or blog about it as I am sure many people are interested in such an application. -
Hi I have a question about shooting in Raw with my Canon EOS 6d.
I'm in the process of learning photography and my goal was to start shooting in raw.
I have Photoshop CS5. When I tried to edit my images in raw I received an error message stating, "The photoshop camera raw plug-in did not recognize the format. If these files are from a camera, you may need to update your camera raw plug in."
In researching the issue I read that to edit in raw you need a camera model requirement of at least 7.3 which only works with CS6. My version of CS5 is 6.0.0.205. Being new to all this I see that my options are to upgrade to CS6 or convert by using DNG converter and paying a monthly fee. Two things I know nothing about and don't know which is would be more beneficial.
I'd appreciate any advise on which route to go and how upgrade and what it may cost. THANKS in advance!
HeatherIn researching the issue I read that to edit in raw you need a camera model requirement of at least 7.3 which only works with CS6.
That is correct. Your camera was first supported by Camera Raw 7.3. Camera Raw 7.3 will not work with CS5. You need CS6 or CC.
Being new to all this I see that my options are to upgrade to CS6 or convert by using DNG converter and paying a monthly fee. Two things I know nothing about and don't know which is would be more beneficial.
I'd appreciate any advise on which route to go and how upgrade and what it may cost.
It all depends on your preferred workflow and your budget.
Using the DNG converter is free. There is no monthly fee. You use the converter to convert all Raw files from the EOS 6D to DNGs then edit the DNGs in CS5. That's an extra step every time - every photo. Some people don't like the extra step. Others don't mind.
Camera raw, DNG | Adobe Photoshop CC
Or you can upgrade to CS6 (non-Cloud) and pay the upgrade fee
Creative Suite 6
Or join the Cloud and pay the monthly fee
Or join the Photoshop Photography Program (US9.99/month) and get PS CC+LR -
Question about sql batch process in java app
hi all
i have few questions about using batch process in the java.sql package. the addBatch method can take sql statements like inserts or updates. can we use a mixture of insert and update then? can we use prepared statement for this? it's just for performance consideration. thanks in advance.hi all
i have few questions about using batch process in the
java.sql package. the addBatch method can take sql
statements like inserts or updates. addBatch() is a method that has no parameters. It doesn't 'add' sql statements.
can we use a
mixture of insert and update then? can we use
prepared statement for this? it's just for
performance consideration. thanks in advance.The point of batching is that you take something that is invariant and then 'add' a variant part.
Thus a single insert has an invariant part (the table and specific columns) and a variant part (the data for each row by column.)
You can use anything that is valid SQL (for jdbc, driver, database) and use it presuming your database allows that particular usage in batching. But that does require some regular pattern - it won't work if your usage is random. Nor will it work if some statements need to be executed only some of the time. Finally also note that transaction processing will often require smaller chunks - you can't insert a million rows in one batch.
Maybe you are looking for
-
Issue of the components to the production
Hi, There is situation to issue the componnets of 500-600 to the production through syatem against single Finish Product, Production person create requisition with the components and sent to stores,store person manually issue the components against t
-
What is the best solution for migrating from Maverick to Yosemite?
What is the best solution for migrating from Maverick to Yosemite? Anyone have suggestions?
-
Cannot Open Project 2010 file in Project 2013
I have a 2010 Project file that will not open in Project 2013. It seems to be something dealing with predecessors and successors. Project 2013 stop responding when I try to open the file, and when I try to copy the information from Project 2010 and p
-
X11 and X11 applications (Eagle, OpenOffice, Gimp) won't open
I am on a macbook, running 10.4.6, and I installed X11 from my Tiger install DVD, just like I did on my G5. X11 won't start, it pops up in the dock, and bounces for a few seconds, and that's it. It goes back down. Help would be greatly appriciated. -
-
ZFBL3N - Internal error in ANALYSE_ACT_FIELDCAT
After copy FBL3N to ZFBL3N, i have this error "Internal error in ANALYSE_ACT_FIELDCAT" during layout modification. Someone could help me to resolve this problem? Thks in advanced !!!