Search Results

Search found 38535 results on 1542 pages for 'sql memory'.

Page 41/1542 | < Previous Page | 37 38 39 40 41 42 43 44 45 46 47 48  | Next Page >

  • SQL 2008 Memory Usage

    - by Danilo Brambilla
    I have a SQL Server 2008 (ver 10.0.1600) running on a Windows Server 2008 R2 Enterprise server with 8 GB of physical ram. If I open Task Manager I can see on 'Physical Memory' section of 'Performance' tab that only 340 MB are Available of 8191 Total, but I can't see any process using such amount of memory. Please note SQL Server is memory limited to 6GB (Maximum Server Memory = 6000). If I open Sysinternals Process Explorer, I can see sqlsrvr.exe process has: Private Bytes: 227.000 K Working Set: 140.000 K Virtual Size: 8.762.000 K What does this means? Is there any way to free up this memory for other process? Why Virtual Size figure as allocated memory? I thought that Virtual Size was 'reserved memory' only.

    Read the article

  • SQL Query for generating matrix like output querying related table in SQL Server

    - by Nagesh
    I have three tables: Product ProductID ProductName 1 Cycle 2 Scooter 3 Car Customer CustomerID CustomerName 101 Ronald 102 Michelle 103 Armstrong 104 Schmidt 105 Peterson Transactions TID ProductID CustomerID TranDate Amount 10001 1 101 01-Jan-11 25000.00 10002 2 101 02-Jan-11 98547.52 10003 1 102 03-Feb-11 15000.00 10004 3 102 07-Jan-11 36571.85 10005 2 105 09-Feb-11 82658.23 10006 2 104 10-Feb-11 54000.25 10007 3 103 20-Feb-11 80115.50 10008 3 104 22-Feb-11 45000.65 I have written a query to group the transactions like this: SELECT P.ProductName AS Product, C.CustName AS Customer, SUM(T.Amount) AS Amount FROM Transactions AS T INNER JOIN Product AS P ON T.ProductID = P.ProductID INNER JOIN Customer AS C ON T.CustomerID = C.CustomerID WHERE T.TranDate BETWEEN '2011-01-01' AND '2011-03-31' GROUP BY P.ProductName, C.CustName ORDER BY P.ProductName which gives the result like this: Product Customer Amount Car Armstrong 80115.50 Car Michelle 36571.85 Car Schmidt 45000.65 Cycle Michelle 15000.00 Cycle Ronald 25000.00 Scooter Peterson 82658.23 Scooter Ronald 98547.52 Scooter Schmidt 54000.25 I need result of query in MATRIX form like this: Customer |------------ Amounts --------------- Name |Car Cycle Scooter Totals Armstrong 80115.50 0.00 0.00 80115.50 Michelle 36571.85 15000.00 0.00 51571.85 Ronald 0.00 25000.00 98547.52 123547.52 Peterson 0.00 0.00 82658.23 82658.23 Schmidt 45000.65 0.00 54000.25 99000.90 Please help me to acheive the above result in SQL Server 2005. Using mulitple views or even temporory tables is fine for me.

    Read the article

  • Idera SQL Doctor 3.0 and MS SQL Changes

    New features worth mentioning in SQL doctor 3.0 begin with a new server dashboard that not only gives a comprehensive overview of a SQL Server instance's current health, but also several key details to help database administrators. Some of the details include recommendations on how to optimize server configuration, how to fix certain security issues, and how to get rid of performance bottlenecks. The latest version of SQL doctor also supplies users with key server information. The status of system parameters known to affect SQL Server performance, such as processes, disk partitions, cache, m...

    Read the article

  • Too much memory consumed during TFS automated build

    - by Bernard Chen
    We're running TFS 2010 Standard Edition, and we've set up an automated build to run whenever someone checks in code. We run through all of the automated tests (built with MSTest) as part of the build. We've configured the build to run the tests as a 64-bit process, but the QTAgent.exe that runs the tests grows in memory while the tests are running. It is currently reaching 8GB for the ~650 tests we have, and the process has slowed significantly when we went from 450 tests to 650 tests. When we run all of the tests in the local development environment, memory seems to be freed at least with each TestClass and never exceeds a certain level. The process of running all tests has not increased significantly in the local development environment. Is there a way to configure the build service to free up memory with each Test or each TestClass? With the way things are currently running, the build process gets very slow when we start to run out of memory on the machine. Edit: I found the MSTest invocation in the build log and ran it manually and saw the same behavior of runaway memory. I removed the /publish, /publishbuild, /teamproject, /platform, and /flavor parameters from the invocation of MSTest, in case the test runner was holding onto results until the end, but the behavior didn't change. I ran the same command line on a dev box, separate from the build server, and the memory freed up frequently. It seems there must be something wrong/different about the build server that is causing it to behave different, but I'm stumped where to look. I've looked at qtagent.exe.config, mstest.exe.config, versions of both executables. What else might affect this?

    Read the article

  • Automatically Kill/Restart Process(es) When Memory is Critically Low

    - by nemesisfixx
    I have a Debian Wheezy VPS box where am running a couple of Django apps in production. Ideally, would have tried addressed my current memory footprint issues by optimizing the apps, adding more RAM or augmenting with Swap. But the problem is that I doubt there's much memory optimization I'd milk from optimizing the Django apps (the stack being open-source and robust), and adding RAM is a cost constraint for me (this is a remote VPS), also, the host doesn't offer options to use Swap! So, in the meantime (as I wait to secure more resources to afford more RAM), I wish to mitigate the scenarios where the server runs out memory so that I just have to request a VPS restart (as in, at that point, I can't even SSH into the box!). So, what I would love in a solution is the ability to detect when a process (or generally, total system memory usage) exceeds a certain critical amount (for now, example the FREE RAM falls to say 10%) - which I've noticed occurs after the VPS's been up for long, and when also traffic is suddenly much to some of the heavy apps (most are just staging apps anyway). So, I wish to be able to kill/restart the offending process(es) - most likely Apache. Which solution when done manually in these situations has restored sane memory usage levels - a hint that possibly one or more of the Django apps has a memory leak? In brief: Monitor overall system RAM usage When FREE RAM falls below a given critical threshold (say below 10%), kill/restart the offending process(es) - or simpler, if we assume from my current log analysis (using linux-dash) that Apache is often the offender, then kill/restart it. Rinse and repeat...

    Read the article

  • Scripting a database copy from MS Sql 2005 to 2008 without detach/backup/RDP

    - by James Santiago
    My goal is to move a single SQL 2005 database to a seperate 2008 server. The issue is my level of access to both servers. On each I can only access the database and nothing else. I cant create a backup file or detach the database because I don't have access to the file system or to create a proxy. I've tried using the generate script function of sql 2005 management studio express to restore the schema but receive command not supported errors when attempting to execute the sql on the new database. Similarly I tried using EMS SQL Manager 2005 Lite to script a backup of the schema and data but ran into similar problems. How do I go about acomplishing this? I can't seem to find any solutions outside of using the detach and backup functions.

    Read the article

  • Reconciling vmware memory vs windows memory usage

    - by RyanW
    I have a Windows 2008 R2 64 bit virtual machine on ESXi 4.1 host. The host reports that the virtual machine is actively using less than 1 GB of memory. But, in Windows it's reporting the machine is using 7 GB of memory, even though the total of the processes listed in task manager is less than 1 GB. The machine is rather unresponsive and I'm concerned this is impacting other applications (server's purpose is to run ASP.NET state server process, which has been having trouble and led me to spot the memory question). I just noticed High memory usage Windows Server 2008r2 on VMware and will be looking through those documents more, but what is causing this?

    Read the article

  • Event ID 17890 (A significant part... paged out.) with SQL Server 2008

    - by Godeke
    I have a machine that has SQL Server 2008 Standard installed. Periodically (about once an hour) I am getting Event ID 17890 several times in a row. An example: 6:28:54 "A significant part of sql server process memory has been paged out. This may result in a performance degradation. Duration: 0 seconds. Working set (KB): 10652, committed (KB): 628428, memory utilization: 1%%. 6:34:27 "A significant part of sql server process memory has been paged out. This may result in a performance degradation. Duration: 332 seconds. Working set (KB): 169780, committed (KB): 546124, memory utilization: 31%%." 6:38:55 "A significant part of sql server process memory has been paged out. This may result in a performance degradation. Duration: 600 seconds. Working set (KB): 245068, committed (KB): 546124, memory utilization: 44%%." This pattern repeated at 7:26 - 7:37, 8:26 - 8:36, 9:24 - 9:35 and so with the same increasing working set and memory utilization pattern. I don't have any (known) background tasks running at this time. Backups run at 2:00 This subsided from 11:00 at night until it resumed at 4:00 in the morning and has been continuing the intermittent 10 minute glitch periods. As this server has plenty of RAM (the commit charge has peaked at 2,871,564 of 4,194,012 physical) I disabled the paging files after reading several items I dug up searching Google and not finding any of them changing the situation. This pattern I am documented is after removing the paging files, so I'm not even sure where we are paging the SQL process could be going. I also changed the SQL process memory to have a minimum of 500MB and a maximum of 2GB of RAM (as this is a light duty database server serving only a small workgroup). Has anyone encountered this? Prior to disabling the page files this error would cause 5 minutes of disk thrashing that disabled access to the databases, files, IIS webs and so on. Since disabling the page files it just logs strange things, but I'm not seeing a performance drop at least. Any suggestions would be welcome.

    Read the article

  • Validate a string in a table in SQL Server - CLR function or T-SQL

    - by Ashish Gupta
    I need to check If a column value (string) in SQL server table starts with a small letter and can only contain '_', '-', numbers and alphabets. I know I can use a SQL server CLR function for that. However, I am trying to implement that validation using a scalar UDF and could make very little here...I can use 'NOT LIKE', but I am not sure how to make sure I validate the string irrespective of the order of characters or in other words write a pattern in SQL for this. Am I better off using a SQL CLR function? Any help will be appreciated.. Thanks in advance Thank you everyone for their comments. This morning, I chose to go CLR function way. For the purpose of what I was trying to achieve, I created one CLR function which does the validation of an input string and have that called from a SQL UDF and It works well. Just to measure the performance of t-SQL UDF using SQL CLR function vs t- SQL UDF, I created a SQL CLR function which will just check if the input string contains only small letters, it should return true else false and have that called from a UDF (IsLowerCaseCLR). After that I also created a regular t-SQL UDF(IsLowerCaseTSQL) which does the same thing using the 'NOT LIKE'. Then I created a table (Person) with columns Name(varchar) and IsValid(bit) columns and populate that with names to test. Data :- 1000 records with 'Ashish' as value for Name column 1000 records with 'ashish' as value for Name column then I ran the following :- UPDATE Person Set IsValid=1 WHERE dbo.IsLowerCaseTSQL (Name) Above updated 1000 records (with Isvalid=1) and took less than a second. I deleted all the data in the table and repopulated the same with same data. Then updated the same table using Sql CLR UDF (with Isvalid=1) and this took 3 seconds! If update happens for 5000 records, regular UDF takes 0 seconds compared to CLR UDF which takes 16 seconds! I am very less knowledgeable on t-SQL regular expression or I could have tested my actual more complex validation criteria. But I just wanted to know, even I could have written that, would that have been faster than the SQL CLR function considering the example above. Are we using SQL CLR because we can implement we can implement lot richer logic which would have been difficult otherwise If we write in regular SQL. Sorry for this long post. I just want to know from the experts. Please feel free to ask if you could not understand anything here. Thank you again for your time.

    Read the article

  • Validate a string in a table in SQL Server - CLR function or T-SQL (Question updated)

    - by Ashish Gupta
    I need to check If a column value (string) in SQL server table starts with a small letter and can only contain '_', '-', numbers and alphabets. I know I can use a SQL server CLR function for that. However, I am trying to implement that validation using a scalar UDF and could make very little here...I can use 'NOT LIKE', but I am not sure how to make sure I validate the string irrespective of the order of characters or in other words write a pattern in SQL for this. Am I better off using a SQL CLR function? Any help will be appreciated.. Thanks in advance Thank you everyone for their comments. This morning, I chose to go CLR function way. For the purpose of what I was trying to achieve, I created one CLR function which does the validation of an input string and have that called from a SQL UDF and It works well. Just to measure the performance of t-SQL UDF using SQL CLR function vs t- SQL UDF, I created a SQL CLR function which will just check if the input string contains only small letters, it should return true else false and have that called from a UDF (IsLowerCaseCLR). After that I also created a regular t-SQL UDF(IsLowerCaseTSQL) which does the same thing using the 'NOT LIKE'. Then I created a table (Person) with columns Name(varchar) and IsValid(bit) columns and populate that with names to test. Data :- 1000 records with 'Ashish' as value for Name column 1000 records with 'ashish' as value for Name column then I ran the following :- UPDATE Person Set IsValid=1 WHERE dbo.IsLowerCaseTSQL (Name) Above updated 1000 records (with Isvalid=1) and took less than a second. I deleted all the data in the table and repopulated the same with same data. Then updated the same table using Sql CLR UDF (with Isvalid=1) and this took 3 seconds! If update happens for 5000 records, regular UDF takes 0 seconds compared to CLR UDF which takes 16 seconds! I am very less knowledgeable on t-SQL regular expression or I could have tested my actual more complex validation criteria. But I just wanted to know, even I could have written that, would that have been faster than the SQL CLR function considering the example above. Are we using SQL CLR because we can implement we can implement lot richer logic which would have been difficult otherwise If we write in regular SQL. Sorry for this long post. I just want to know from the experts. Please feel free to ask if you could not understand anything here. Thank you again for your time.

    Read the article

  • How to join dynamic sql statement in variable with normal statement

    - by Oliver
    I have a quite complicated query which will by built up dynamically and is saved in a variable. As second part i have another normal query and i'd like to make an inner join between these both. To make it a little more easier here is a little example to illustrate my problem. For this little example i used the AdventureWorks database. Some query built up dynamically (Yes, i know here is nothing dynamic here, cause it's just an example.) DECLARE @query AS varchar(max) ; set @query = ' select HumanResources.Employee.EmployeeID ,HumanResources.Employee.LoginID ,HumanResources.Employee.Title ,HumanResources.EmployeeAddress.AddressID from HumanResources.Employee inner join HumanResources.EmployeeAddress on HumanResources.Employee.EmployeeID = HumanResources.EmployeeAddress.EmployeeID ;'; EXEC (@query); The normal query i have select Person.Address.AddressID ,Person.Address.City from Person.Address Maybe what i'd like to have but doesn't work select @query.* ,Addresses.City from @query as Employees inner join ( select Person.Address.AddressID ,Person.Address.City from Person.Address ) as Addresses on Employees.AddressID = Addresses.AddressID

    Read the article

  • ANTS Memory Profiler 7.0 Review

    - by Michael B. McLaughlin
    (This is my first review as a part of the GeeksWithBlogs.net Influencers program. It’s a program in which I (and the others who have been selected for it) get the opportunity to check out new products and services and write reviews about them. We don’t get paid for this, but we do generally get to keep a copy of the software or retain an account for some period of time on the service that we review. In this case I received a copy of Red Gate Software’s ANTS Memory Profiler 7.0, which was released in January. I don’t have any upgrade rights nor is my review guided, restrained, influenced, or otherwise controlled by Red Gate or anyone else. But I do get to keep the software license. I will always be clear about what I received whenever I do a review – I leave it up to you to decide whether you believe I can be objective. I believe I can be. If I used something and really didn’t like it, keeping a copy of it wouldn’t be worth anything to me. In that case though, I would simply uninstall/deactivate/whatever the software or service and tell the company what I didn’t like about it so they could (hopefully) make it better in the future. I don’t think it’d be polite to write up a terrible review, nor do I think it would be a particularly good use of my time. There are people who get paid for a living to review things, so I leave it to them to tell you what they think is bad and why. I’ll only spend my time telling you about things I think are good.) Overview of Common .NET Memory Problems When coming to land of managed memory from the wilds of unmanaged code, it’s easy to say to one’s self, “Wow! Now I never have to worry about memory problems again!” But this simply isn’t true. Managed code environments, such as .NET, make many, many things easier. You will never have to worry about memory corruption due to a bad pointer, for example (unless you’re working with unsafe code, of course). But managed code has its own set of memory concerns. For example, failing to unsubscribe from events when you are done with them leaves the publisher of an event with a reference to the subscriber. If you eliminate all your own references to the subscriber, then that memory is effectively lost since the GC won’t delete it because of the publishing object’s reference. When the publishing object itself becomes subject to garbage collection then you’ll get that memory back finally, but that could take a very long time depending of the life of the publisher. Another common source of resource leaks is failing to properly release unmanaged resources. When writing a class that contains members that hold unmanaged resources (e.g. any of the Stream-derived classes, IsolatedStorageFile, most classes ending in “Reader” or “Writer”), you should always implement IDisposable, making sure to use a properly written Dispose method. And when you are using an instance of a class that implements IDisposable, you should always make sure to use a 'using' statement in order to ensure that the object’s unmanaged resources are disposed of properly. (A ‘using’ statement is a nicer, cleaner looking, and easier to use version of a try-finally block. The compiler actually translates it as though it were a try-finally block. Note that Code Analysis warning 2202 (CA2202) will often be triggered by nested using blocks. A properly written dispose method ensures that it only runs once such that calling dispose multiple times should not be a problem. Nonetheless, CA2202 exists and if you want to avoid triggering it then you should write your code such that only the innermost IDisposable object uses a ‘using’ statement, with any outer code making use of appropriate try-finally blocks instead). Then, of course, there are situations where you are operating in a memory-constrained environment or else you want to limit or even eliminate allocations within a certain part of your program (e.g. within the main game loop of an XNA game) in order to avoid having the GC run. On the Xbox 360 and Windows Phone 7, for example, for every 1 MB of heap allocations you make, the GC runs; the added time of a GC collection can cause a game to drop frames or run slowly thereby making it look bad. Eliminating allocations (or else minimizing them and calling an explicit Collect at an appropriate time) is a common way of avoiding this (the other way is to simplify your heap so that the GC’s latency is low enough not to cause performance issues). ANTS Memory Profiler 7.0 When the opportunity to review Red Gate’s recently released ANTS Memory Profiler 7.0 arose, I jumped at it. In order to review it, I was given a free copy (which does not include upgrade rights for future versions) which I am allowed to keep. For those of you who are familiar with ANTS Memory Profiler, you can find a list of new features and enhancements here. If you are an experienced .NET developer who is familiar with .NET memory management issues, ANTS Memory Profiler is great. More importantly still, if you are new to .NET development or you have no experience or limited experience with memory profiling, ANTS Memory Profiler is awesome. From the very beginning, it guides you through the process of memory profiling. If you’re experienced and just want dive in however, it doesn’t get in your way. The help items GAHSFLASHDAJLDJA are well designed and located right next to the UI controls so that they are easy to find without being intrusive. When you first launch it, it presents you with a “Getting Started” screen that contains links to “Memory profiling video tutorials”, “Strategies for memory profiling”, and the “ANTS Memory Profiler forum”. I’m normally the kind of person who looks at a screen like that only to find the “Don’t show this again” checkbox. Since I was doing a review, though, I decided I should examine them. I was pleasantly surprised. The overview video clocks in at three minutes and fifty seconds. It begins by showing you how to get started profiling an application. It explains that profiling is done by taking memory snapshots periodically while your program is running and then comparing them. ANTS Memory Profiler (I’m just going to call it “ANTS MP” from here) analyzes these snapshots in the background while your application is running. It briefly mentions a new feature in Version 7, a new API that give you the ability to trigger snapshots from within your application’s source code (more about this below). You can also, and this is the more common way you would do it, take a memory snapshot at any time from within the ANTS MP window by clicking the “Take Memory Snapshot” button in the upper right corner. The overview video goes on to demonstrate a basic profiling session on an application that pulls information from a database and displays it. It shows how to switch which snapshots you are comparing, explains the different sections of the Summary view and what they are showing, and proceeds to show you how to investigate memory problems using the “Instance Categorizer” to track the path from an object (or set of objects) to the GC’s root in order to find what things along the path are holding a reference to it/them. For a set of objects, you can then click on it and get the “Instance List” view. This displays all of the individual objects (including their individual sizes, values, etc.) of that type which share the same path to the GC root. You can then click on one of the objects to generate an “Instance Retention Graph” view. This lets you track directly up to see the reference chain for that individual object. In the overview video, it turned out that there was an event handler which was holding on to a reference, thereby keeping a large number of strings that should have been freed in memory. Lastly the video shows the “Class List” view, which lets you dig in deeply to find problems that might not have been clear when following the previous workflow. Once you have at least one memory snapshot you can begin analyzing. The main interface is in the “Analysis” tab. You can also switch to the “Session Overview” tab, which gives you several bar charts highlighting basic memory data about the snapshots you’ve taken. If you hover over the individual bars (and the individual colors in bars that have more than one), you will see a detailed text description of what the bar is representing visually. The Session Overview is good for a quick summary of memory usage and information about the different heaps. You are going to spend most of your time in the Analysis tab, but it’s good to remember that the Session Overview is there to give you some quick feedback on basic memory usage stats. As described above in the summary of the overview video, there is a certain natural workflow to the Analysis tab. You’ll spin up your application and take some snapshots at various times such as before and after clicking a button to open a window or before and after closing a window. Taking these snapshots lets you examine what is happening with memory. You would normally expect that a lot of memory would be freed up when closing a window or exiting a document. By taking snapshots before and after performing an action like that you can see whether or not the memory is really being freed. If you already know an area that’s giving you trouble, you can run your application just like normal until just before getting to that part and then you can take a few strategic snapshots that should help you pin down the problem. Something the overview didn’t go into is how to use the “Filters” section at the bottom of ANTS MP together with the Class List view in order to narrow things down. The video tutorials page has a nice 3 minute intro video called “How to use the filters”. It’s a nice introduction and covers some of the basics. I’m going to cover a bit more because I think they’re a really neat, really helpful feature. Large programs can bring up thousands of classes. Even simple programs can instantiate far more classes than you might realize. In a basic .NET 4 WPF application for example (and when I say basic, I mean just MainWindow.xaml with a button added to it), the unfiltered Class List view will have in excess of 1000 classes (my simple test app had anywhere from 1066 to 1148 classes depending on which snapshot I was using as the “Current” snapshot). This is amazing in some ways as it shows you how in stark detail just how immensely powerful the WPF framework is. But hunting through 1100 classes isn’t productive, no matter how cool it is that there are that many classes instantiated and doing all sorts of awesome things. Let’s say you wanted to examine just the classes your application contains source code for (in my simple example, that would be the MainWindow and App). Under “Basic Filters”, click on “Classes with source” under “Show only…”. Voilà. Down from 1070 classes in the snapshot I was using as “Current” to 2 classes. If you then click on a class’s name, it will show you (to the right of the class name) two little icon buttons. Hover over them and you will see that you can click one to view the Instance Categorizer for the class and another to view the Instance List for the class. You can also show classes based on which heap they live on. If you chose both a Baseline snapshot and a Current snapshot then you can use the “Comparing snapshots” filters to show only: “New objects”; “Surviving objects”; “Survivors in growing classes”; or “Zombie objects” (if you aren’t sure what one of these means, you can click the helpful “?” in a green circle icon to bring up a popup that explains them and provides context). Remember that your selection(s) under the “Show only…” heading will still apply, so you should update those selections to make sure you are seeing the view you want. There are also links under the “What is my memory problem?” heading that can help you diagnose the problems you are seeing including one for “I don’t know which kind I have” for situations where you know generally that your application has some problems but aren’t sure what the behavior you have been seeing (OutOfMemoryExceptions, continually growing memory usage, larger memory use than expected at certain points in the program). The Basic Filters are not the only filters there are. “Filter by Object Type” gives you the ability to filter by: “Objects that are disposable”; “Objects that are/are not disposed”; “Objects that are/are not GC roots” (GC roots are things like static variables); and “Objects that implement _______”. “Objects that implement” is particularly neat. Once you check the box, you can then add one or more classes and interfaces that an object must implement in order to survive the filtering. Lastly there is “Filter by Reference”, which gives you the option to pare down the list based on whether an object is “Kept in memory exclusively by” a particular item, a class/interface, or a namespace; whether an object is “Referenced by” one or more of those choices; and whether an object is “Never referenced by” one or more of those choices. Remember that filtering is cumulative, so anything you had set in one of the filter sections still remains in effect unless and until you go back and change it. There’s quite a bit more to ANTS MP – it’s a very full featured product – but I think I touched on all of the most significant pieces. You can use it to debug: a .NET executable; an ASP.NET web application (running on IIS); an ASP.NET web application (running on Visual Studio’s built-in web development server); a Silverlight 4 browser application; a Windows service; a COM+ server; and even something called an XBAP (local XAML browser application). You can also attach to a .NET 4 process to profile an application that’s already running. The startup screen also has a large number of “Charting Options” that let you adjust which statistics ANTS MP should collect. The default selection is a good, minimal set. It’s worth your time to browse through the charting options to examine other statistics that may also help you diagnose a particular problem. The more statistics ANTS MP collects, the longer it will take to collect statistics. So just turning everything on is probably a bad idea. But the option to selectively add in additional performance counters from the extensive list could be a very helpful thing for your memory profiling as it lets you see additional data that might provide clues about a particular problem that has been bothering you. ANTS MP integrates very nicely with all versions of Visual Studio that support plugins (i.e. all of the non-Express versions). Just note that if you choose “Profile Memory” from the “ANTS” menu that it will launch profiling for whichever project you have set as the Startup project. One quick tip from my experience so far using ANTS MP: if you want to properly understand your memory usage in an application you’ve written, first create an “empty” version of the type of project you are going to profile (a WPF application, an XNA game, etc.) and do a quick profiling session on that so that you know the baseline memory usage of the framework itself. By “empty” I mean just create a new project of that type in Visual Studio then compile it and run it with profiling – don’t do anything special or add in anything (except perhaps for any external libraries you’re planning to use). The first thing I tried ANTS MP out on was a demo XNA project of an editor that I’ve been working on for quite some time that involves a custom extension to XNA’s content pipeline. The first time I ran it and saw the unmanaged memory usage I was convinced I had some horrible bug that was creating extra copies of texture data (the demo project didn’t have a lot of texture data so when I saw a lot of unmanaged memory I instantly figured I was doing something wrong). Then I thought to run an empty project through and when I saw that the amount of unmanaged memory was virtually identical, it dawned on me that the CLR itself sits in unmanaged memory and that (thankfully) there was nothing wrong with my code! Quite a relief. Earlier, when discussing the overview video, I mentioned the API that lets you take snapshots from within your application. I gave it a quick trial and it’s very easy to integrate and make use of and is a really nice addition (especially for projects where you want to know what, if any, allocations there are in a specific, complicated section of code). The only concern I had was that if I hadn’t watched the overview video I might never have known it existed. Even then it took me five minutes of hunting around Red Gate’s website before I found the “Taking snapshots from your code" article that explains what DLL you need to add as a reference and what method of what class you should call in order to take an automatic snapshot (including the helpful warning to wrap it in a try-catch block since, under certain circumstances, it can raise an exception, such as trying to call it more than 5 times in 30 seconds. The difficulty in discovering and then finding information about the automatic snapshots API was one thing I thought could use improvement. Another thing I think would make it even better would be local copies of the webpages it links to. Although I’m generally always connected to the internet, I imagine there are more than a few developers who aren’t or who are behind very restrictive firewalls. For them (and for me, too, if my internet connection happens to be down), it would be nice to have those documents installed locally or to have the option to download an additional “documentation” package that would add local copies. Another thing that I wish could be easier to manage is the Filters area. Finding and setting individual filters is very easy as is understanding what those filter do. And breaking it up into three sections (basic, by object, and by reference) makes sense. But I could easily see myself running a long profiling session and forgetting that I had set some filter a long while earlier in a different filter section and then spending quite a bit of time trying to figure out why some problem that was clearly visible in the data wasn’t showing up in, e.g. the instance list before remembering to check all the filters for that one setting that was only culling a few things from view. Some sort of indicator icon next to the filter section names that appears you have at least one filter set in that area would be a nice visual clue to remind me that “oh yeah, I told it to only show objects on the Gen 2 heap! That’s why I’m not seeing those instances of the SuperMagic class!” Something that would be nice (but that Red Gate cannot really do anything about) would be if this could be used in Windows Phone 7 development. If Microsoft and Red Gate could work together to make this happen (even if just on the WP7 emulator), that would be amazing. Especially given the memory constraints that apps and games running on mobile devices need to work within, a good memory profiler would be a phenomenally helpful tool. If anyone at Microsoft reads this, it’d be really great if you could make something like that happen. Perhaps even a (subsidized) custom version just for WP7 development. (For XNA games, of course, you can create a Windows version of the game and use ANTS MP on the Windows version in order to get a better picture of your memory situation. For Silverlight on WP7, though, there’s quite a bit of educated guess work and WeakReference creation followed by forced collections in order to find the source of a memory problem.) The only other thing I found myself wanting was a “Back” button. Between my Windows Phone 7, Zune, and other things, I’ve grown very used to having a “back stack” that lets me just navigate back to where I came from. The ANTS MP interface is surprisingly easy to use given how much it lets you do, and once you start using it for any amount of time, you learn all of the different areas such that you know where to go. And it does remember the state of the areas you were previously in, of course. So if you go to, e.g., the Instance Retention Graph from the Class List and then return back to the Class List, it will remember which class you had selected and all that other state information. Still, a “Back” button would be a welcome addition to a future release. Bottom Line ANTS Memory Profiler is not an inexpensive tool. But my time is valuable. I can easily see ANTS MP saving me enough time tracking down memory problems to justify it on a cost basis. More importantly to me, knowing what is happening memory-wise in my programs and having the confidence that my code doesn’t have any hidden time bombs in it that will cause it to OOM if I leave it running for longer than I do when I spin it up real quickly for debugging or just to see how a new feature looks and feels is a good feeling. It’s a feeling that I like having and want to continue to have. I got the current version for free in order to review it. Having done so, I’ve now added it to my must-have tools and will gladly lay out the money for the next version when it comes out. It has a 14 day free trial, so if you aren’t sure if it’s right for you or if you think it seems interesting but aren’t really sure if it’s worth shelling out the money for it, give it a try.

    Read the article

  • sql 2008 metadata modified date

    - by Kumar
    Is there a way to identify the timestamp when an object(table/view/stored proc...) was modified ? there's a refdate in sysobjects but it's always the same as crdate atleast in my case and i know that alter view/alter table/alter proc commands have been run many times post creation

    Read the article

  • t-sql recursive query

    - by stackoverflowuser
    Based on an existing table I used CTE recursive query to come up with following data. But failing to apply it a level further. Data is as below id name parentid -------------------------- 1 project 0 2 structure 1 3 path_1 2 4 path_2 2 5 path_3 2 6 path_4 3 7 path_5 4 8 path_6 5 I want to recursively form full paths from the above data. Means the recursion will give the following output. FullPaths ------------- Project Project\Structure Project\Structure\Path_1 Project\Structure\Path_2 Project\Structure\Path_3 Project\Structure\Path_1\path_4 Project\Structure\Path_2\path_5 Project\Structure\Path_3\path_6 Thanks

    Read the article

  • Use SQL to filter the results of a stored procedure

    - by Ben McCormack
    I've looked at other questions on Stack Overflow related to this question, but none of them seemed to answer this question clearly. We have a system Stored Procedure called sp_who2 which returns a result set of information for all running processes on the server. I want to filter the data returned by the stored procedure; conceptually, I might do it like so: SELECT * FROM sp_who2 WHERE login='bmccormack' That method, though, doesn't work. What are good practices for achieving the goal of querying the returned data of a stored procedure, preferably without having to look of the code of the original stored procedure and modify it.

    Read the article

  • Stored Procedure To Search the AccessRights given to the Users.

    - by thevan
    Hi, I want to display the Access Rights given to the Users for the particular module. I have Seven Tables such as RoleAccess, Roles, Functions, Module, SubModule, Company and Unit. RoleAccess is the Main Table. The AccessRights given will be stored in the RoleAccess Table only. RoleAccess Table has the following columns such as RoleID, CompanyID, UnitID, FunctionID, ModuleID, SubModuleID, Create, Update, Delete, Read, Approve. Here Create_f, Update_f, Delete_f, Read_f and Approve_f are flags. Company Table has two columns such as CompanyID and CompanyName. Unit Table has three columns such as UnitID, UnitName and CompanyID. Roles Table has four columns such as RoleID, RoleName, CompanyID and UnitID. Module Table has two columns such as ModuleID and ModuleName. SubModule Table has three columns such as ModuleID, SubModuleID, SubModuleName. Functions Table has five columns such as FunctionID, FunctionName, ModuleID and SubModuleID. At First, The RoleAccess Table does not contain any records. So I want to display the ModuleName, SubModuleName, FunctionName, CompanyID, RoleID, UnitID, FunctionID, ModuleID, SubModuleID, Create_f, Update_f, Delete_f, Read_f and Approve_f. If the AccessRights is assigned to the Particular RoleID means the flags in the search results will be 1 else it will be 0. I have witten one stored procedure but it displays the records based on the RoleID stored in the RoleAccess table. But I also want to display the Flags as 0 for the Roles not stored in the RoleAccess Table. I want the Stored Procedure for this. Any one please help me.

    Read the article

  • sql server 2008 takes alot of memory?

    - by Ahmed Said
    I making stress test on my database which is hosted on sqlserver 2008 64bit running on 64bit machine 10 GB of RAM. I have 400 threads each thread query the database for every second but the query time does not take time as the sql profiler says that, but after 18 hours sql takes 7.2 GB RAM and 7.2 on virtual memroy. Does is this normal behavior? and how can I adjust sql to clean up not in use memory?

    Read the article

  • How to Track Number of Changes Occured in a Column? T-SQL - SQL Server

    - by user327387
    For example, i have a column named EmployeeName. Every time a user changes/fix his name, i need to keep a count. If he changes his name twice, then count is 2. Also, i need to store the time of every change employee makes on EmployeeName e.g. if the name essentially is James and time created is 9:00 AM and then employee changes to John on 11:00 AM, i need to preserve this new time and new value as well as the previous one which is James at 9:00 AM. I hope its clear! Thank you guys...Best Answer will be chosen... If this requires a trigger, giving a sketchy pseudo-code will be very much appreciated.

    Read the article

  • The SQL Server Setup Portal

    - by BuckWoody
    One of the tasks that takes a long time for the data professional is setting up SQL Server. No, it isn’t that difficult to slide a DVD in a drive and click “Setup” but the overall process of planning the hardware and software environment, making decisions for high-availability, security and dozens of other choices can make the process more difficult. And then, of course, there are the inevitable issues that arise. Microsoft supports literally hundreds and even thousands of combinations of hardware and software drivers from vendors you’ve never even heard of. Making all of that work together is a small miracle, so things are bound to arise that you need to deal with. So, to help you out, we’ve designed a new “SQL Server Setup Portal”. It’s a one-stop-shop for everything you need to know about planning and setting up SQL Server. As time goes on you’ll see even more content added. There are already whitepapers, videos, and multiple places to search on everything from topic names to error codes. So go check it out – and if you have to do a lot of SQL Server Setups – and especially if you don’t – bookmark it as a favorite! Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

  • How would I design this table in SQL?

    - by RSharma
    I have a parent master table that is generic enough to hold the common information of the children. Since the children were substantially different, we created separate tables for them. So I have something like this: tblMaster -------- MasterID int Name varchar(50) --Common to all children and there are a bunch of fields like this ChildType int -- Type of Child either ChildOne or ChildTwo ChildID int -- need to store ChildOneID or ChildTwoID depending on type of Child, so that i can refer to children tblChild1 -------- ChildOneID int IDENTITY tblChild2 --------- ChildTwoID int IDENTITY Should I have a ChildID in the master that is either ChildOneID or ChildTwoID based on the ChildType column? I have a number of children and I have simplified it for this question. The other way is to add ChildOneID and ChildTwoID as columns in the master, but since i have a number of columns, I will have a lot of null columns EDIT: Any help is appreciated

    Read the article

  • SQL Server 2008 (64-bit)

    - by Grace09
    I have transaction log backup failures randomly. SQL Server Log says 'Memory constraints resulted reduced backup/restore buffer sizes. Proceding with 6 buffers of size 64KB.', and the SQL Server Agent Error Log has quite a few errors like 'Unable to start Job Manager thread for job xxx', '[298] SQLServer Error: 768, Client unable to establish connection [SQLSTATE 08001]', [298] SQLServer Error: 768, SSL Provider: Not enough memory is available to complete this request [SQLSTATE 08001]'. I am really low on the memory? It has total of 32 GB, but I set maximum to 20. Task manager shows it is using 99% of physical memory. Memoryclerk-sqlbufferpool has 32GB for the Virtual Memory Reserved, and 20 GB for the Virtual Memory Committed. From the perfmon, SQLServer:Memory Manager/Total ServerMemory shows 21GB of memory in use, that's what I set the maximum to. I don't where the rest of memory go. Can anyone advice? Thanks in advance.

    Read the article

  • Ballooning Mac OS X kernel_task and Wired memory usage. How to diagnose / fix?

    - by user28930
    I have a very strange issue, which I'm having a hard time diagnosing as to the root cause. I have a Mac Pro (2008, 8-core 2.8 GHz, 8800GT) with 14 GB of RAM (recently upgraded because of this issue!). When I boot my system and log in, vm_stat / top / Activity Monitor will show that kernel_task has about 150 MB allocated, and the machine has about 800 MB of Wired memory being allocated. Even initially, 800 MB seems an awful lot of wired memory to be allocated with no applications running - but, it gets worse. (NB: Wired is locked, unswappable memory) After a very short time, sometimes triggered by something as simple as launching a terminal, kernel_task will balloon to 8-900 MB of Real Mem (RSIZE), and Wired Memory will accelerate to 1.6 GB (implying that all the extra memory requests are for wired RAM in the kernel). If I quit everything (I.E: no running applications, bar an activity monitor or terminal to view top), there is no appreciable reduction in either kernel_task RSIZE, or Wired Memory usage. Going the opposite way, and loading the system with tasks also shows that wired memory does not get reduced - and that importantly it is not reduced in preference to heavy swapping. If I log out and log back in again, it reduces a bit (450 MB kernel_task, 1.28 GB Wired), but not back to the start. I'm not running any wacky kexts - and futhermore, kextstat shows no huge memory allocations there; the largest being com.apple.nvidia.nv50hal at about 4 MB of Memory. The machine feels overall more sluggish when this has happened - unsurprisingly because such a huge amount of RAM has been marked as non-pageable. So I have a few questions: 1) Is there a good way to diagnose what has allocated all of this wired memory? It's often over 2 times the kernel_task size, running no applications. The real memory total doesn't seem to add up - it seems that there is a bunch of RAM that isn't being accounted for anywhere. 2) What is happening to cause the kernel to suddenly require 6 times as much memory?

    Read the article

  • T-SQL Tuesday #15 : Running T-SQL workloads remotely on multiple servers

    - by AaronBertrand
    This month's installment of T-SQL Tuesday is hosted by Pat Wright ( blog | twitter ). Pat says: "So the topic I have chosen for this month is Automation! It can be Automation with T-SQL or with Powershell or a mix of both. Give us your best tips/tricks and ideas for making our lives easier through Automation." In a recent project, we've had a need to run concurrent workloads on as many as 100 instances of SQL Server in a test environment. A goal, obviously, is to accomplish this without having to...(read more)

    Read the article

  • T-SQL Tuesday #15 : Running T-SQL workloads remotely on multiple servers

    - by AaronBertrand
    This month's installment of T-SQL Tuesday is hosted by Pat Wright ( blog | twitter ). Pat says: "So the topic I have chosen for this month is Automation! It can be Automation with T-SQL or with Powershell or a mix of both. Give us your best tips/tricks and ideas for making our lives easier through Automation." In a project we are working on, we've had a need to run concurrent workloads on as many as 100 instances of SQL Server in a test environment. A goal, obviously, is to accomplish this without...(read more)

    Read the article

< Previous Page | 37 38 39 40 41 42 43 44 45 46 47 48  | Next Page >