Search Results

Search found 13403 results on 537 pages for 'epm performance tuning'.

Page 252/537 | < Previous Page | 248 249 250 251 252 253 254 255 256 257 258 259  | Next Page >

  • Intel Parallel XE : Intel publie un eBook gratuit pour découvrir son outil d'optimisation des développements parallèles

    Intel Parallel XE : Intel publie un eBook gratuit Pour découvrir son outil d'optimisation des développements parallèles Les solutions Intel Parallel développées, comme leur nm l'indique, par Intel sont un ensemble d'outils qui permettent une meilleure optimisation des développements parallèles pour tirer partie des architectures multi-coeurs. « La nouvelle gamme d'outils Intel Parallel XE permet aux équipes de développement de délivrer le code en temps et en heure avec le niveau de performance le plus élevé eet le minimum de défauts du cluster au desktop jusqu'au périphériques », explique Intel. Pour aider les développeurs dans la découverte et la prise en main d'Int...

    Read the article

  • SSIS Prehistory video

    - by jamiet
    I’m currently wasting spending my Easter bank holiday putting together my presentation SSIS Dataflow Performance Tuning for the upcoming SQL Bits conference in London and in doing so I’m researching some old material about how the dataflow actually works. Boring as it is I’ve gotten easily sidelined and have chanced upon an old video on Channel 9 entitled Euan Garden - Tour of SQL Server Team (part I). Euan is a former member of the SQL Server team and in this series of videos he walks the halls of the SQL Server building on Microsoft’s Redmond campus talking to some of the various protagonists and in this one he happens upon the SQL Server Integration Services team. The video was shot in 2004 so this is a fascinating (to me anyway) glimpse into the development of SSIS from before it was ever shipped and if you’re a geek like me you’ll really enjoy this behind-the-scenes look into how and why the product was architected. The video is also notable for the presence of the cameraman – none other than the now-rather-more-famous-than-he-was-then Robert Scoble. See it at http://channel9.msdn.com/posts/TheChannel9Team/Euan-Garden-Tour-of-SQL-Server-Team-part-I/ Enjoy! @Jamiet Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

  • Python Forgiveness vs. Permission and Duck Typing

    - by darkfeline
    In Python, I often hear that it is better to "beg forgiveness" (exception catching) instead of "ask permission" (type/condition checking). In regards to enforcing duck typing in Python, is this try: x = foo.bar except AttributeError: pass else: do(x) better or worse than if hasattr(foo, "bar"): do(foo.bar) else: pass in terms of performance, readability, "pythonic", or some other important factor?

    Read the article

  • SAP va étendre le champ d'application de sa technologie « In-Memory » et réduire la complexité des environnements IT

    SAP va étendre le champ d'application de sa technologie « In-Memory » Et réduire la complexité des environnements IT Le « In-Memory » consiste à mettre en cache les données traitées par les applications plutôt que, par exemple, de faire des appels à un serveur. Chez SAP, la technologie est baptisée HANA (High-Performance Analytic Appliance). Elle peut, d'après les benchmarks de l'éditeur Allemand, aller jusqu'à diviser par 1200 le temps de traitement de certains scénarios. SAP a intégré HANA à la dernière version de

    Read the article

  • A new version of Oracle Enterprise Manager Ops Center Doctor (OCDoctor ) Utility released

    - by Anand Akela
    In February,  we posted a blog of Oracle Enterprise Manager Ops Center Doctor aka OCDoctor Utility. This utility assists in various stages of the Ops Center deployment and can be a real life saver. It is updated on a regular basis with additional knowledge (similar to an antivirus subscription) to help you identify and resolve known issues or suggest ways to improve performance.A new version ( Version 4.00 ) of the OCDoctor is now available . This new version adds full support for recently announced Oracle Enterprise Manager Ops Center 12c including prerequisites checks, troubleshoot tests, log collection, tuning and product metadata updates. In addition, it adds several bug fixes and enhancements to OCDoctor Utility.To download OCDoctor for new installations:https://updates.oracle.com/OCDoctor/OCDoctor-latest.zipFor existing installations, simply run:# /var/opt/sun/xvm/OCDoctor/OCDoctor.sh --updateTip : If you have Oracle Enterprise Manager Ops Center12c EC installed, your OCDoctor will automatically update overnight. Join Oracle Launch Webcast : Total Cloud Control for Systems on April 12th at 9 AM PST to learn more about  Oracle Enterprise Manager Ops Center 12c from Oracle Senior Vice President John Fowler, Oracle Vice President of Systems Management Steve Wilson and a panel of Oracle executive. Stay connected with  Oracle Enterprise Manager   :  Twitter | Facebook | YouTube | Linkedin | Newsletter

    Read the article

  • Don’t Miss The Top Exastack ISV Headlines – Week Of May 26

    - by Roxana Babiciu
    Calypso Technology announced that Calypso version 14 has achieved Oracle Exadata Optimized status through OPN. In simulations of data-intensive straight through-processing tasks, Calypso achieved performance gains of up to 500% using Exadata hardware – Read more Infosys achieves Oracle SuperCluster Optimized status with Finacle, a core banking solution. Finacle can process 6x the volume of transactions currently processed by the entire US banking system – Read more

    Read the article

  • SQL SERVER – Copy Column Headers from Resultset – SQL in Sixty Seconds #026 – Video

    - by pinaldave
    SQL Server Management Studio returns results in Grid View, Text View and to the file. When we copy results from Grid View to Excel there is a common complaint that the column  header displayed in resultset is not copied to the Excel. I often spend time in performance tuning databases and I run many DMV’s in SSMS to get a quick view of the server. In my case it is almost certain that I need all the time column headers when I copy my data to excel or any other place. SQL Server Management Studio have two different ways to do this. Method 1: Ad-hoc When result is rendered you can right click on the resultset and click on Copy Header. This will copy the headers along with the resultset. Additionally, you can use the shortcut key CTRL+SHIFT+C for coping column headers along with the resultset. Method 2: Option Setting at SSMS level This is SSMS level settings and I kept this option always selected as I often need the column headers when I select the resultset. Go Tools >> Options >> Query Results >> SQL Server >> Results to Grid >> Check the Box “Include column header when copying or saving the results.” Both of the methods are discussed in following SQL in Sixty Seconds Video. Here is the code used in the video. Related Tips in SQL in Sixty Seconds: Copy Column Headers in Query Analyzers in Result Set Getting Columns Headers without Result Data – SET FMTONLY ON If we like your idea we promise to share with you educational material. Reference: Pinal Dave (http://blog.sqlauthority.com) Filed under: Database, Pinal Dave, PostADay, SQL, SQL Authority, SQL in Sixty Seconds, SQL Query, SQL Scripts, SQL Server, SQL Server Management Studio, SQL Tips and Tricks, T SQL, Technology, Video

    Read the article

  • Digital Storage for Airline Entertainment

    - by Bill Evjen
    by Thomas Coughlin Common flash memory cards The most common flash memory products currently in use are SD cards and derivative products (e.g. mini and micro-SD cards) Some compact flash used for professional applications (such as DSLR cameras) Evolution of leading flash formats Standardization –> market expansion Market expansion –> volume iNAND –> focus is on enabling embedded X3 iSSD –> ideal for thin form factor devices Flash memory applications Phones are the #1 user of flash memory Flash memory is used as embedded and removable storage in many mobile applications Flash memory is being used in computers as USB sticks and SSDs Possible use of flash memory in computer combined with HDDs (hybrid HDDs and paired or dual storage computers) It can be a removable card or an embedded card These devices can only handle a specific number of writes Flash memory reads considerably quicker than hard drives Hybrid and dual storage in computers SSDs can provide fast performance but they are expensive HDDs can provide cheap storage but they are relatively slow Combining some flash memory with a HDD can provide costs close to those of HDDs and performance close to flash memory Seagate Momentus XT hybrid HDD Various dual storage offerings putting flash memory with HDDs Other common flash memory devices USB sticks All forms and colors Used for moving files around Some sold with content on them (Sony Movies on USB sticks) Solid State Drives (SSDs) Floating Gate Flash Memory Cell When a bit is programmed, electrons are stored upon the floating gate This has the effect of offsetting the charge on the control gate of the transistor If there is no charge upon the floating gate, then the control gate’s charge determines whether or not a current flows through the channel A strong charge on the control gate assumes that no current flows. A weak charge will allow a strong current to flow through. Similar to HDDs, flash memory must provide: Bit error correction Bad block management NAND and NOR memories are treated differently when it comes to managing wear In many NOR-based systems no management is used at all, since the NOR is simply used to store code, and data is stored in other devices. In this case, it would take a near-infinite amount of time for wear to become an issue since the only time the chip would see an erase/write cycle is when the code in the system is being upgraded, which rarely if ever happens over the life of a typical system. NAND is usually found in very different application than is NOR Flash memory wears out This is expected to get worse over time Retention: Disappearing data Bits fade away Retention decreases with increasing read/writes Bits may change when adjacent bits are read Time and traffic are concerns Controllers typically groom read disturb errors Like DRAM refresh Increases erase/write frequency Application characteristics Music – reads high / writes very low Video – r high / writes very low Internet Cache – r high / writes low On airplanes Many consumers now have their own content viewing devices – do they need the airlines? Is there a way to offer more to consumers, especially with their own viewers Additional special content tie into airplane network access to electrical power, internet Should there be fixed embedded or removable storage for on-board airline entertainment? Is there a way to leverage personal and airline viewers and content in new and entertaining ways?

    Read the article

  • BizTalk 2009 - BizTalk Server Best Practice Analyser

    - by StuartBrierley
    The BizTalk Server Best Practices Analyser  allows you to carry out a configuration level verification of your BizTalk installation, evaluating the deployed configuration but not modifying or tuning anything that it finds. The Best Practices Analyser uses "reading and reporting" to gather data from different sources, such as: Windows Management Instrumentation (WMI) classes SQL Server databases Registry entries When I first ran the analyser I got a number of errors, if you get any errors these should all be acted upon to resolve them, you should then run the scan again and see if any thing else is reported that needs acting upon. As you can see in the image above, the initial issue that jumped out to me was that the SQL Server Agent was not started. The reasons for this was absent mindedness - this run was against my development PC and I don't have SQL/BizTalk actively running unless I am using them.  Starting the agent service and running the scan again gave me the following results: This resolved most of the issues for me, but next major issue to look at was that there was no tracking host running.  You can also see that I was still getting an error with two of the SQL jobs.  The problem here was that I had not yet configured these two SQL jobs.  Configuring the backup and purge jobs and then starting the tracking host before running the scan again gave: This had cleared all the critical issues, but I did stil have a number of warnings.  For example on this report I was warned that the BizTalk Message box is hosted on the BizTalk Server.  While this is known to be less than ideal, it is as I expected on my development environment where I have installed Visual Studio, SQL and BizTalk on my laptop and I was happy to ignore this and other similar warnings. In your case you should take a look at any warnings you receive and decide what you want to do about each of them in turn.

    Read the article

  • Understanding Column Properties for a SQL Server Table

    Designing a table can be a little complicated if you don’t have the correct knowledge of data types, relationships, and even column properties. In this tip, Brady Upton goes over the column properties and provides examples. "It really helped us isolate where we were experiencing a bottleneck"- John Q Martin, SQL Server DBA. Get started with SQL Monitor today to solve tricky performance problems - download a free trial

    Read the article

  • Invitation: Oracle EMEA Analytics & Data Integration Partner Forum, 12th November 2012, London (UK)

    - by rituchhibber
    Oracle PartnerNetwork | Account | Feedback INVITATIONORACLE EMEA ANALYTICS & DATA INTEGRATION PARTNER FORUM MONDAY 12TH NOVEMBER, 2012 IN LONDON (UK) Dear partner Come to hear the latest news from Oracle OpenWorld about Oracle BI & Data Integration, and propel your business growth as an Oracle partner. This event should appeal to BI or Data Integration specialised partners, Executives, Sales, Pre-sales and Solution architects: with a choice of participation in the plenary day and then a set of special interest (technical) sessions. The follow on breakout sessions from the 13th November provide deeper dives and technical training for those of you who wish to stay for more detailed and hands-on workshops.Keynote: Andrew Sutherland, SVP Oracle Technology. Data Integration can bring great value to your customers by moving data to transform their business experiences in Oracle pan-EMEA Data Integration business development and opportunities for partners. Hot agenda items will include: The Fusion Middleware Stack: Engineered to work together A complete Analytics and Data Integration Solution Architecture: Big Data and Little Data combined In-Memory Analytics for Extreme Insight Latest Product Development roadmap for Data Integration and Analytics Venue: Oracles London CITY Moorgate OfficesDuring this event you can learn about partner success stories, participate in an array of break-out sessions, exchange information with other partners and enjoy a vibrant panel discussion. Places are limited, Register your seat today! To register to this event CLICK HERE Note: Registration for the conference and the deeper dives and technical training is free of charge to OPN member Partners, but you will be responsible for your own travel and hotel expenses. Event Schedule November 12th:Day 1 Main Plenary Session : Full day, starting 10.30 am.Oracle Hosted Dinner in the Evening November 13th:onwards Architecture Masterclass : IM Reference Architecture – Big Data and Little Data combined(1 day) BI-Apps Bootcamp(4-days) Oracle Data Integrator and Oracle Enterprise Data Quality workshop(1-day) Golden Gate Workshop(1-day) For further information and detail download the Agenda (pdf) or contact Michael Hallett at [email protected] look forward to seeing you in there. Best regards, Mike HallettAlliances and Channels DirectorBI & EPM Oracle EMEAM.No: +44 7831 276 989 [email protected] Duncan HarveyBusiness Development Directorfor Data IntegrationM.No: +420 608 283 [email protected] Milomir VojvodicBusiness Development Manager for Data IntegrationM.No: +420 608 283 [email protected] Copyright © 2012, Oracle and/or its affiliates. All rights reserved. Contact PBC | Legal Notices and Terms of Use | Privacy

    Read the article

  • JDeveloper and ADF at UKOUG

    - by Grant Ronald
    This year, Oracle ADF and JDeveloper has a big showing at the UKOUG (about 22 hours worth!!)- Europe's largest Oracle User Group.  There are three days packed with awesome ADF content delivered by some of the leading lights in ADF Developement including Duncan Mills, Frank Nimphius, Shay Shmeltzer, Susan Duncan, Lucas Jellema, Steven Davelaar, Sten Vesterli (and I'll be there as well!). Please make sure you refer to the official agenda for timings but an outline is here (if you think there are any sessions I have missed let me know and I will add them) Monday 10:00 - 10:45 - Deepdive into logical and physical data modeling with JDeveloper 10:00 - 12:15 - Debugging ADF Applications 12:15 - 13:15 - Learn ADF Task Flows in 60 Minutes 14:30 - 15:15 - ADF's Hidden Gem - the Groovy scripting language in Oracle ADF 15:25 - 16:10 - ADF Patterns for Forms Conversions 16:35 - 17:35 - Dummies Guide to Oracle ADF 16:35 - 17:35 - ADF Security Overview - Strategies and Best Practices 17:45 - 18:30 - A Methodology for Enterprise Applications with Oracle ADF Tuesday 09:00 - 10:00 - Real World Performance Tuning for Oracle ADF 11:15 - 12:15 - Keynote: Modern Development, Mobility and Rich Internet Applications 11:15 - 12:15 - Migration to Fusion Middleware 11g: Real world cases of Forms, ADF and Identity Management upgrades 14:40 - 15:20 - What's new in JDeveloper 11gR2 14:40 - 15:20 - Development Tools Roundtable 15:35 - 16:20 - ALM in Jdeveloper is exciting! 16:40 - 17:40 - Moving Oracle Forms to Oracle ADF: Case Studies Wednesday 09:00 - 10:00 - Building a Multi-Tasking ADF Application with Dynamic Regions and Dynamic Tabs 10:10 - 10:55 - Building Highly Reusable ADF Taskflows 12:30 - 13:30 - Design Patterns, Customization and Extensibility of Fusion Applications 14:25 - 15:10 - Continuous Integration with Hudson: What a year! 14:00 - 17:00 - Wednesday Wizardry with Fusion Middleware - Live application development demonstration with ADF, SOA Suite 15:20 - 16:05 - Adding Mobile and Web 2.0 UIs to Existing Applications - The Fusion Way  16:15 - 17:00 - Leveraging ADF for Building Complex Custom Applications

    Read the article

  • Technical differences between square and hexagon for a grid?

    - by Marlon Dias
    I'm developing a 2D city-building game and trying to decide on the type of grid. There will be vehicles, so the unit movement is important too. I know there are visual differences for using Squares or Hexagons, what I want know is: What are the issues for programming each type of grid regarding implementation and performance? Is there a tradeoff or specific benefit for using one of them in a game context?

    Read the article

  • Working with Sub-Optimal Disk Configurations (Making the best of what you’ve got)

    - by Jonathan Kehayias
    This is the first post in a what will be a series of posts on working with a sub-optimal disk configuration to squeeze as much performance out of it as possible.  You might ask what a Sub-Optimal Disk Configuration?  In this case it is a Dell Powervault MD3000 with 15 Seagate Barracuda ES.2 SAS 1 TB 7.2K RPM disks (Model Number ST31000640SS).  This equates to just under 14TB of raw storage that can configured into a number of RAID configurations.  In this case, the disk array...(read more)

    Read the article

  • Printer Review: HP LaserJet Pro 1606dn

    Looking for a black-and-white laser printer for your small office or workgroup? HP's $199 entry offers Ethernet, duplex printing, and fast performance -- and can install itself with no CD to insert or driver to download.

    Read the article

  • Printer Review: HP LaserJet Pro 1606dn

    Looking for a black-and-white laser printer for your small office or workgroup? HP's $199 entry offers Ethernet, duplex printing, and fast performance -- and can install itself with no CD to insert or driver to download.

    Read the article

  • Apple met à jour son Mac Mini : un nouveau design, et un démontage plus aisé, mais un prix salé

    Apple vient de dévoilé son nouveau MacMini : Citation: Apple Unveils All New Mac mini CUPERTINO, California?June 15, 2010?Apple® today unveiled a completely redesigned Mac® mini, featuring up to twice the graphics performance, a new HDMI port and a new SD card slot, all in an amazingly compact aluminum enclosure. Mac mini is the world's most energy efficient desktop and starting at $699, is the most affordable way to enjoy Mac OS® X, iLife® or Mac OS X Snow Leopard® Server. ?The sleek, aluminum Mac mini packs gre...

    Read the article

  • Announcing SO-Aware Test Workbench

    - by gsusx
    Yesterday was a big day for Tellago Studios . After a few months hands down working, we announced the release of the SO-Aware Test Workbench tool which brings sophisticated performance testing and test visualization capabilities to theWCF world. This work has been the result of the feedback received by many of our SO-Aware and Tellago customers in terms of how to improve the WCF testing. More importantly, with the SO-Aware Test Workbench we are trying to address what has been one of the biggest challenges...(read more)

    Read the article

  • ASP.NET 4.0- Menu control enhancement.

    - by Jalpesh P. Vadgama
    Till asp.net 3.5 asp.net menu control was rendered through table. And we all know that it is very hard to have CSS applied to table. For a professional look of our website a CSS is must required thing. But in asp.net 4.0 Menu control is table less it will loaded with UL and LI tags which is easier to manage through CSS. Another problem with table is it will create a large html which will increase your asp.net page KB and decrease your performance. While with UL and LI Tags its very easy very short. So You page KB Size will also be down. Let’s take a simple example. Let’s Create a menu control in asp.net with four menu item like following. <asp:Menu ID="myCustomMenu" runat="server" > <Items> <asp:MenuItem Text="Menu1" Value="Menu1"></asp:MenuItem> <asp:MenuItem Text="Menu2" Value="Menu2"></asp:MenuItem> <asp:MenuItem Text="Menu3" Value="Menu3"></asp:MenuItem> <asp:MenuItem Text="Menu4" Value="Menu4"></asp:MenuItem> </Items></asp:Menu> It will render menu in browser like following. Now If we render this menu control with tables then HTML as you can see via view page source like following.   Now If in asp.net 4.0 It will be loaded with UL and LI tags and if you now see page source then it will look like following. Which will have must lesser HTML then it was earlier like following. So isn’t that great performance enhancement?.. It’s very cool. If you still like old way doing with tables then in asp.net 4.0 there is property called ‘RenderingMode’ is given. So you can set RenderingMode=Table then it will load menu control with table otherwise it will load menu control with UL and LI Tags. That’s it..Stay tuned for more..Happy programming.. Technorati Tags: Menu,Asp.NET 4.0

    Read the article

  • Small Business and Organic SEO - A Win Win Situation

    Small business owners have to run on tiny budgets and that becomes constraint for effective publicity. The online marketing performance also suffers due to this crucial factor. The remedy lies in organic SEO, which efficiently works for the website of small business and supports the placement in higher rankings in search results. It is a win win situation for small business owners.

    Read the article

  • Update of SAE Benchmark Presentation to M6/T5/ZFS

    - by uwes
    Strategic Applications Engineering (SAE) published in March an updated Benchmark presentation showing the performance of Oracle systems, software and Virtualization. SPARC M6/T5/ZFS Benchmarks March 2014 The presentation is available via our eSTEP portal.  You will need to provide your email address and the pin below to access the downloads. Link to the portal is shown below. URL: http://launch.oracle.com/ PIN: eSTEP_2011 The material can be found under tab eSTEP Download Located under: Recent Updates and Miscellaneous

    Read the article

  • Thread placement policies on NUMA systems - update

    - by Dave
    In a prior blog entry I noted that Solaris used a "maximum dispersal" placement policy to assign nascent threads to their initial processors. The general idea is that threads should be placed as far away from each other as possible in the resource topology in order to reduce resource contention between concurrently running threads. This policy assumes that resource contention -- pipelines, memory channel contention, destructive interference in the shared caches, etc -- will likely outweigh (a) any potential communication benefits we might achieve by packing our threads more densely onto a subset of the NUMA nodes, and (b) benefits of NUMA affinity between memory allocated by one thread and accessed by other threads. We want our threads spread widely over the system and not packed together. Conceptually, when placing a new thread, the kernel picks the least loaded node NUMA node (the node with lowest aggregate load average), and then the least loaded core on that node, etc. Furthermore, the kernel places threads onto resources -- sockets, cores, pipelines, etc -- without regard to the thread's process membership. That is, initial placement is process-agnostic. Keep reading, though. This description is incorrect. On Solaris 10 on a SPARC T5440 with 4 x T2+ NUMA nodes, if the system is otherwise unloaded and we launch a process that creates 20 compute-bound concurrent threads, then typically we'll see a perfect balance with 5 threads on each node. We see similar behavior on an 8-node x86 x4800 system, where each node has 8 cores and each core is 2-way hyperthreaded. So far so good; this behavior seems in agreement with the policy I described in the 1st paragraph. I recently tried the same experiment on a 4-node T4-4 running Solaris 11. Both the T5440 and T4-4 are 4-node systems that expose 256 logical thread contexts. To my surprise, all 20 threads were placed onto just one NUMA node while the other 3 nodes remained completely idle. I checked the usual suspects such as processor sets inadvertently left around by colleagues, processors left offline, and power management policies, but the system was configured normally. I then launched multiple concurrent instances of the process, and, interestingly, all the threads from the 1st process landed on one node, all the threads from the 2nd process landed on another node, and so on. This happened even if I interleaved thread creating between the processes, so I was relatively sure the effect didn't related to thread creation time, but rather that placement was a function of process membership. I this point I consulted the Solaris sources and talked with folks in the Solaris group. The new Solaris 11 behavior is intentional. The kernel is no longer using a simple maximum dispersal policy, and thread placement is process membership-aware. Now, even if other nodes are completely unloaded, the kernel will still try to pack new threads onto the home lgroup (socket) of the primordial thread until the load average of that node reaches 50%, after which it will pick the next least loaded node as the process's new favorite node for placement. On the T4-4 we have 64 logical thread contexts (strands) per socket (lgroup), so if we launch 48 concurrent threads we will find 32 placed on one node and 16 on some other node. If we launch 64 threads we'll find 32 and 32. That means we can end up with our threads clustered on a small subset of the nodes in a way that's quite different that what we've seen on Solaris 10. So we have a policy that allows process-aware packing but reverts to spreading threads onto other nodes if a node becomes too saturated. It turns out this policy was enabled in Solaris 10, but certain bugs suppressed the mixed packing/spreading behavior. There are configuration variables in /etc/system that allow us to dial the affinity between nascent threads and their primordial thread up and down: see lgrp_expand_proc_thresh, specifically. In the OpenSolaris source code the key routine is mpo_update_tunables(). This method reads the /etc/system variables and sets up some global variables that will subsequently be used by the dispatcher, which calls lgrp_choose() in lgrp.c to place nascent threads. Lgrp_expand_proc_thresh controls how loaded an lgroup must be before we'll consider homing a process's threads to another lgroup. Tune this value lower to have it spread your process's threads out more. To recap, the 'new' policy is as follows. Threads from the same process are packed onto a subset of the strands of a socket (50% for T-series). Once that socket reaches the 50% threshold the kernel then picks another preferred socket for that process. Threads from unrelated processes are spread across sockets. More precisely, different processes may have different preferred sockets (lgroups). Beware that I've simplified and elided details for the purposes of explication. The truth is in the code. Remarks: It's worth noting that initial thread placement is just that. If there's a gross imbalance between the load on different nodes then the kernel will migrate threads to achieve a better and more even distribution over the set of available nodes. Once a thread runs and gains some affinity for a node, however, it becomes "stickier" under the assumption that the thread has residual cache residency on that node, and that memory allocated by that thread resides on that node given the default "first-touch" page-level NUMA allocation policy. Exactly how the various policies interact and which have precedence under what circumstances could the topic of a future blog entry. The scheduler is work-conserving. The x4800 mentioned above is an interesting system. Each of the 8 sockets houses an Intel 7500-series processor. Each processor has 3 coherent QPI links and the system is arranged as a glueless 8-socket twisted ladder "mobius" topology. Nodes are either 1 or 2 hops distant over the QPI links. As an aside the mapping of logical CPUIDs to physical resources is rather interesting on Solaris/x4800. On SPARC/Solaris the CPUID layout is strictly geographic, with the highest order bits identifying the socket, the next lower bits identifying the core within that socket, following by the pipeline (if present) and finally the logical thread context ("strand") on the core. But on Solaris on the x4800 the CPUID layout is as follows. [6:6] identifies the hyperthread on a core; bits [5:3] identify the socket, or package in Intel terminology; bits [2:0] identify the core within a socket. Such low-level details should be of interest only if you're binding threads -- a bad idea, the kernel typically handles placement best -- or if you're writing NUMA-aware code that's aware of the ambient placement and makes decisions accordingly. Solaris introduced the so-called critical-threads mechanism, which is expressed by putting a thread into the FX scheduling class at priority 60. The critical-threads mechanism applies to placement on cores, not on sockets, however. That is, it's an intra-socket policy, not an inter-socket policy. Solaris 11 introduces the Power Aware Dispatcher (PAD) which packs threads instead of spreading them out in an attempt to be able to keep sockets or cores at lower power levels. Maximum dispersal may be good for performance but is anathema to power management. PAD is off by default, but power management polices constitute yet another confounding factor with respect to scheduling and dispatching. If your threads communicate heavily -- one thread reads cache lines last written by some other thread -- then the new dense packing policy may improve performance by reducing traffic on the coherent interconnect. On the other hand if your threads in your process communicate rarely, then it's possible the new packing policy might result on contention on shared computing resources. Unfortunately there's no simple litmus test that says whether packing or spreading is optimal in a given situation. The answer varies by system load, application, number of threads, and platform hardware characteristics. Currently we don't have the necessary tools and sensoria to decide at runtime, so we're reduced to an empirical approach where we run trials and try to decide on a placement policy. The situation is quite frustrating. Relatedly, it's often hard to determine just the right level of concurrency to optimize throughput. (Understanding constructive vs destructive interference in the shared caches would be a good start. We could augment the lines with a small tag field indicating which strand last installed or accessed a line. Given that, we could augment the CPU with performance counters for misses where a thread evicts a line it installed vs misses where a thread displaces a line installed by some other thread.)

    Read the article

< Previous Page | 248 249 250 251 252 253 254 255 256 257 258 259  | Next Page >