Search Results

Search found 32994 results on 1320 pages for 'second level cache'.

Page 295/1320 | < Previous Page | 291 292 293 294 295 296 297 298 299 300 301 302  | Next Page >

  • Using DNS in iproute2

    - by Oliver
    In my setup I can redirect the default gateway based on the source address. Let's say a user is connected through tun0 (10.2.0.0/16) is redirect to another vpn. That works fine! ip rule add from 10.2.0.10 lookup vpn1 In a second rule I redirect the default gateway to another gateway if the user access a certain ip adress: ip rule add from 10.2.0.10 to 94.142.154.71 lookup vpn2 If I access the page on 94.142.154.71 (myip.is) the user is correctly routed and I can see the ip of the second vpn. On any other pages the ip address of vpn1 is shown. But how do I tell iproute2 that all request at e. g. google.com should be redirected through vpn2?

    Read the article

  • Ignoring Robots - Or Better Yet, Counting Them Separately

    - by [email protected]
    It is quite common to have web sessions that are undesirable from the point of view of analytics. For example, when there are either internal or external robots that check the site's health, index it or just extract information from it. These robotic session do not behave like humans and if their volume is high enough they can sway the statistics and models.One easy way to deal with these sessions is to define a partitioning variable for all the models that is a flag indicating whether the session is "Normal" or "Robot". Then all the reports and the predictions can use the "Normal" partition, while the counts and statistics for Robots are still available.In order for this to work, though, it is necessary to have two conditions:1. It is possible to identify the Robotic sessions.2. No learning happens before the identification of the session as a robot.The first point is obvious, but the second may require some explanation. While the default in RTD is to learn at the end of the session, it is possible to learn in any entry point. This is a setting for each model. There are various reasons to learn in a specific entry point, for example if there is a desire to capture exactly and precisely the data in the session at the time the event happened as opposed to including changes to the end of the session.In any case, if RTD has already learned on the session before the identification of a robot was done there is no way to retract this learning.Identifying the robotic sessions can be done through the use of rules and heuristics. For example we may use some of the following:Maintain a list of known robotic IPs or domainsDetect very long sessions, lasting more than a few hours or visiting more than 500 pagesDetect "robotic" behaviors like a methodic click on all the link of every pageDetect a session with 10 pages clicked at exactly 20 second intervalsDetect extensive non-linear navigationNow, an interesting experiment would be to use the flag above as an output of a model to see if there are more subtle characteristics of robots such that a model can be used to detect robots, even if they fall through the cracks of rules and heuristics.In any case, the basic and simple technique of partitioning the models by the type of session is simple to implement and provides a lot of advantages.

    Read the article

  • MySQL is killing the server IO.

    - by OneOfOne
    I manage a fairly large/busy vBulletin forums (~2-3k requests per second, running on gigenet cloud), the database is ~ 10 GB (~9 milion posts, ~60 queries per second), lately MySQL have been grinding the disk like there's no tomorrow according to iotop and slowing the site. The last idea I can think of is using replication, but I'm not sure how much that would help and worried about database sync. I'm out of ideas, any tips on how to improve the situation would be highly appreciated. Specs : Debian Lenny 64bit ~12Ghz (6 cores) CPU, 7520gb RAM, 160gb disk. Kernel : 2.6.32-4-amd64 mysqld Ver 5.1.54-0.dotdeb.0 for debian-linux-gnu on x86_64 ((Debian)) Other software: vBulletin 3.8.4 memcached 1.2.2 PHP 5.3.5-0.dotdeb.0 (fpm-fcgi) (built: Jan 7 2011 00:07:27) lighttpd/1.4.28 (ssl) - a light and fast webserver PHP and vBulletin are configured to use memcached. MySQL Settings : [mysqld] key_buffer = 128M max_allowed_packet = 16M thread_cache_size = 8 myisam-recover = BACKUP max_connections = 1024 query_cache_limit = 2M query_cache_size = 128M expire_logs_days = 10 max_binlog_size = 100M key_buffer_size = 128M join_buffer_size = 8M tmp_table_size = 16M max_heap_table_size = 16M table_cache = 96

    Read the article

  • Postgres 9.0 locking up, 100% CPU

    - by Jake
    We are having a problem where our Postgres 9.0 server occasionally locks up and kills our webapp. Restarting Postgres fixes the problem. Here's what I've been able to observe: First, usage of one CPU jumps to 100% for a few minutes Disk operations drop to ~0 during this time Database operations drop to 0 (blocks and tuples per sec) Logs show during this time: WARNING: worker took too long to start; cancelled WARNING: worker took too long to start; cancelled No Queries in logs (only those over 200ms are logged) No unusually long-running queries logged before or during Then the second CPU jumps to 100% The number of postgres processes jumps from the usual 8-10 to ~20 Matched by a spike in Postgres Blocks per second (about twice normal) Logs show LOG: could not accept SSL connection: EOF detected Queries are running but slow Restarting postgres returns everything to normal Setup: Server: Amazon EC2 Large Ubuntu 10.04.2 LTS Postgres 9.0.3 Dedicated DB server Does anyone have any idea what's causing this? Or any suggestions about what else I should be checking out?

    Read the article

  • The internal storage of a DATETIME2 value

    - by Peter Larsson
    Today I went for investigating the internal storage of DATETIME2 datatype. What I found out was that for a datetime2 value with precision 0 (seconds only), SQL Server need 6 bytes to represent the value, but stores 7 bytes. This is because SQL Server add one byte that holds the precision for the datetime2 value. Start with this very simple repro declare @now datetime2(7) = '2010-12-15 21:04:03.6934231'   select  cast(cast(@now as datetime2(0)) as binary(7)),         cast(cast(@now as datetime2(1)) as binary(7)),         cast(cast(@now as datetime2(2)) as binary(7)),         cast(cast(@now as datetime2(3)) as binary(8)),         cast(cast(@now as datetime2(4)) as binary(8)),         cast(cast(@now as datetime2(5)) as binary(9)),         cast(cast(@now as datetime2(6)) as binary(9)),         cast(cast(@now as datetime2(7)) as binary(9)) Now we are going to copy and paste these binary values and investigate which value is representing what time part. Prefix  Ticks       Ticks         Days    Days    Original value ------  ----------  ------------  ------  ------  -------------------- 0x  00  442801             75844  A8330B  734120  0x00442801A8330B 0x  01  A5920B            758437  A8330B  734120  0x01A5920BA8330B  0x  02  71BA73           7584369  A8330B  734120  0x0271BA73A8330B 0x  03  6D488504        75843693  A8330B  734120  0x036D488504A8330B 0x  04  46D4342D       758436934  A8330B  734120  0x0446D4342DA8330B 0x  05  BE4A10C401    7584369342  A8330B  734120  0x05BE4A10C401A8330B 0x  06  6FEBA2A811   75843693423  A8330B  734120  0x066FEBA2A811A8330B 0x  07  57325D96B0  758436934231  A8330B  734120  0x0757325D96B0A8330B Let us use the following color schema Red - Prefix Green - Time part Blue - Day part What you can see is that the date part is equal in all cases, which makes sense since the precision doesm't affect the datepart. What would have been fun, is datetime2(negative) just like round accepts a negative value. -1 would mean rounding to 10 second, -2 rounding to minute, -3 rounding to 10 minutes, -4 rounding to hour and finally -5 rounding to 10 hour. -5 is pretty useless, but if you extend this thinking to -6, -7 and so on, you could actually get a datetime2 value which is accurate to the month only. Well, enough ranting about this. Let's get back to the table above. If you add 75844 second to midnight, you get 21:04:04, which is exactly what you got in the select statement above. And if you look at it, it makes perfect sense that each following value is 10 times greater when the precision is increased one step too. //Peter

    Read the article

  • WHM Cpanel Enable/Install MBstring module

    - by Phil Jackson
    Hi, this is my second post here and only my second day dealing with my own dedicated server. I have no previous experience in this field so please bear with me. I have successfully created a package and linked up my domain yesterday. I am getting an error regarding mb_convert_encoding() wich after doing some reading discover I have to enable or install a module?? Any way, my cpanel is "cPanel 11.25.0-R46156 - WHM 11.25.0 - X 3.9" and after some searching found mainsoftwareModule Installers but cannot find any to do with mbstrings. Could anyone point me in the right direction in what or where I should be looking?? Any help much appreciated, Regards, Phil

    Read the article

  • IP tables gateway

    - by WojonsTech
    I am trying to make an iptables gateway. I ordered 3 dedicated server from my hosting company all with dual nics. One server has been given all the ip addresses and is connected directly to the internet and has its other nic connected to a switch where the other servers are all connected also. I want to setup iptables so for example the ip address 50.0.2.4 comes into my gateway server it fowards all the traffic to a private ip address using the second nic. This way the second nic can do what ever it needs and can respond back also. I also want it setup that if any of the other servers needs to download anything over the inernet it is able to do so and by using the same ip address that is used for its incomming traffic. Lastly I would like to be able to setup dns and other needed networking stuff that i maybe not thinking about.

    Read the article

  • How do I reset the time on my computer, without turning it on?

    - by chipperyman573
    Alright, so today, I did something very stupid: Do an experement on my computer without backing it up. So I saw that the calander in windows could only go up to 12/31/2999 (or something like that). I was wondering if I set the time to 11:59:59 PM, if it would crash my computer, thinking that if I did I could just restart it from the recovery disc or something. Well, I was right: It did crash it. However, I can't turn my computer on AT ALL. When I try to, it plays a 1-2 second beep, 1 second silence, repeated a total of 3 times. My manufacturer is Dell. I'm sending this from my phone, I appolgise for all typos. My last backup was from a few months ago, that won't work.

    Read the article

  • Engagement: Don’t Forget Your Employees!

    - by Kellsey Ruppel
    By Mark Brown, Sr. Director, Oracle WebCenter  This week we want to focus on Employee Engagement, and how it is critical to your business. Today we hear and read a great deal about “Customer Engagement” – and rightly so, it is those customers, whether they be traditional paying customers, citizens, students, club members, or whomever it is that are “paying the bills”.  A more engaged customer is more likely to make it easier to pay those bills by buying more, giving good reviews, or spreading the word of how wonderful their experience was. But what about those who are providing those services, those who design and make those goods; why is it that all too often they are left out of conversations concerning engagement? In fact, it is critical that we consider our employees as customers since they are using internal systems that run your organization the same way customers use external systems. Studies have shown that an organization in which the employees feel “engaged” or better able to make decisions, do their jobs, and are connected to their peers have better return to their stakeholders. (shareholders).  On the surface this seems obvious, happy employees are more productive employees. But it leads to the question – how many of our existing policies, systems and processes are actually reducing that level of engagement? Let’s look at a couple examples. If posting new information that may be of great value to everyone in the larger organization is hard to do because we use an antiquated system, then we’re making it hard to share and increasing the potential for duplicate work. If it is not trivially obvious how to create and publish this post, then chances are very high that I’ll put it on the bottom of my queue. And finally, when critical information is spread across various systems, intranet sites, workgroups and peoples inboxes, then it is very hard to learn and grow from that information.  These may sound trivial, but how often do we push things off not because it is intellectually challenging, we may have the answer at our fingertips, but because it is hard to make that information readily available.  If an engaged employee is a productive employee, then what can we do to increase their level of engagement? We can start by looking for opportunities to provide self-documenting self-service solutions. Our newer employees grew up using simplified web interfaces everyday and they loathe calling a help-desk unless it is the last resort. Sadly, many of our enterprise applications have not kept pace and we all still have processes that are based on sending an email -- like discount approvals, vacation requests, or even offer-letter approvals.   My suggestion is to pick one highly visible, high-impact process where employees are either reticent to execute on the process or openly complain about how cumbersome it is and look at the mechanism for that process. If there are better ways, streamlined steps, better UIs that could be done, then you have a candidate to reconfigure that process and make it more engaging. Looking to better engage your employees? Start here!

    Read the article

  • Path of md device wrong after reboot

    - by flammi88
    I have to set up a software raid (level1) on a Ubuntu server 12.04. It should serve files in the network via Samba. The server has the following disks: 250gb Sata hdd (Ubuntu is installed on that drive) 2 TB Sata hdd (first disk in raid array, data disk) 2 TB Sata hdd (second data disk) I created one partition on every data disk with the type Linux raid autodetect. In the second step I created the raid1 with the following command: mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sdb1 /dev/sdc1 After that, I added the array to the mdconf: mdadm --examine --scan >> /etc/mdadm/mdadm.conf The problem is: After a reboot the array is not available on the path /dev/md0. Instead of that it gets reassembled as /dev/md/0 but it is not very reliable. Has anybody a solution for this issue?

    Read the article

  • Share laptop Wi-Fi with router set up as access point via ethernet cables

    - by obie
    I have an ADSL modem which serves as a wireless router. There's a laptop connected to it through cable. On the other room I have a laptop which is connected through the router wirelessly. Since I have other appliances that need to get connected but are not Wi-Fi capable I have another Wi-Fi router. How can I share the second laptop's Wi-Fi with the second router so that the router can then serve as an access point to give internet to my other appliances through ethernet cables? I want something like this: INTERNETMODEM (WIFI)LAPTOP 1 (WIFI)LAPTOP 2 (WIFI)Router 2 (ETHERNET)DREAMBOX & WDTV What I've tried so far is opened the admin interface of router 2 then set static IP 192.168.100.13 and disabled the DHCP. The DHCP of the main router starts with 192.168.100.1 then I hooked up router 2 with the laptop 2 through Cat5 cable but for some reason it's not working.

    Read the article

  • Friday Fun: Shape Fold

    - by Asian Angel
    This week’s game comes with lots of puzzle solving, brain-teasing goodness to keep you busy. On each level you will need to rotate, twist, and/or move the hinged puzzle pieces into their proper shape. Do you have the patience and skill to solve all the puzzles or will you be forced to admit defeat? 7 Ways To Free Up Hard Disk Space On Windows HTG Explains: How System Restore Works in Windows HTG Explains: How Antivirus Software Works

    Read the article

  • Can I use Ubuntu One to sync data fiies between two remote computers

    - by Sleepy John
    I've got two computers, both running Ubuntu with files in their home folders sync'd in to Ubuntu One. I'd like to know if it's possible to make Ubuntu One automatically download data changes that have been uploaded automatically to Ubuntu One from one computer to the equivalent data file in the other. Clarifying a bit further, I've installed Red Notebook in both computers and so they each have their own /.rednotebook/data folder containing a series of .txt files corresponding to the monthly entries in each of them. These are sync'd to upload any changes to those .txt files to Ubuntu One. My question is can I, and if so how, do I make Ubuntu One automatically download and replace those .txt files in the other computer after they've been updated and uploaded from the first computer? I did labouriously manage to download all those text files which had been uploaded from the first computer, from Ubuntu One one-by-one to the second computer, but what I want to do is automate this process and that's where I'm stuck. I'm aware that things could get a bit complicated if both my computers were on-line at the same time and both were simultaneously making different Red Notebook entries, so that's not the scenario I'm trying to cover. All I want to achieve is that whatever updates to the files have been uploaded by one computer, will automatically be downloaded to the same-named files in the other computer as soon as that second computer appears on line and detects that Ubuntu One has matching but more recent sync'd files than the ones it's holding.

    Read the article

  • Worldwide Web Camps

    - by ScottGu
    Over the next few weeks Microsoft is sponsoring a number of free Web Camp events around the world.  These provide a great way to learn about ASP.NET 4, ASP.NET MVC 2, and Visual Studio 2010. The Web Camps are two day events.  The camps aren’t conferences where you sit quietly for hours and people talk at you – they are intended to be interactive.  The first day is focused on learning through presentations that are heavy on coding demos.  The second day is focused on you building real applications using what you’ve learned.  The second day includes hands-on labs, and you’ll join small development teams with other attendees and work on a project together. We’ve got some great speakers lined up for the events – including Scott Hanselman, James Senior, Jon Galloway, Rachel Appel, Dan Wahlin, Christian Wenz and more.  I’ll also be presenting at one of the camps. Below is the schedule of the remaining events (the sold-out Toronto camp was a few days ago): Moscow May 19-19 Beijing May 21-22 Shanghai May 24-25 Mountain View May 27-28 Sydney May 28-29 Singapore June 04-05 London June 04-05 Munich June 07-08 Chicago June 11-12 Redmond, WA June 18-19 New York June 25-26 Many locations are sold out already but we still have some seats left in a few of them.  Registration and attendance to all of the events is completely free.  You can register to attend at www.webcamps.ms. Hope this helps, Scott

    Read the article

  • When to favor ASP.NET WebForms over MVC

    - by P.Brian.Mackey
    I know Microsoft has said "ASP.NET MVC is not a replacement for WebForms". Some developers say WebForms is faster to develop than MVC, but I believe this all comes down to comfort level with the technology; so I don't want any answers in this direction. Given that ASP.NET MVC gives a developer more control over our application, why is WebForms not considered obsolete? When should I favor WebForms over MVC for new development?

    Read the article

  • Two network cards in one Windows 7 PC

    - by Martin
    Hi, I've upgraded from Vista to Windows 7. On Vista I used to run one network card (Intel 82566DC -2 Gigabit) for my internet which was connected to a Dovado USB mobile broadband router connected to a 3G modem. The IP setup on the card is set on DHCP. The second card (3com Ethernet link) was used to connect to our town's local WUG (wireless user group) and all worked fine. Now the problem I have with Windows 7 is that when I connect to the internet with the (Intel) first card, the internet work fine but as soon as I connect the second card to the wug my internet drops. If I then disconnect the first card (internet one) and try to connect to the wug it works. It seems that the two cards are fighting each other. Is there maybe something I can do to fix the problem? As you probably can gather from this post is that I am not an expert on PC's so please if you do answer do I in simple terms. :) Thanks, Martin

    Read the article

  • Zabbix - Some of the monitored items don't refreh

    - by Niro
    I'm experiencing a strange issue with Zabbix monitoring a MySQL server. Most of the data from the server such as MySQL queries per second and MySQL uptime , Buffers memory etc. update nicely while some data like CPU iowait time (avg1) , Host local time ,MySQL number of threads and other items which were monitored in the past has last check time of about a week ago. I can't find any logic in this, for example Mysql number of threads and Mysql queries per second are obtained in a similar way so it does not make sense one of them is monitored and one is not. Please help- how can I fix this? Update - I used zabbix_get from the zabbix server to check one of the items on the zabbix client and it works so the problem must be on the zabbix server side

    Read the article

  • SQL University: Parallelism Week - Part 3, Settings and Options

    - by Adam Machanic
    Congratulations! You've made it back for the the third and final installment of Parallelism Week here at SQL University . So far we've covered the fundamentals of multitasking vs. parallel processing and delved into how parallel query plans actually work . Today we'll take a look at the settings and options that influence intra-query parallelism and discuss how best to set things up in various situations. Instance-Level Configuration Your database server probably has more than one logical processor....(read more)

    Read the article

  • How To Use Regular Expressions for Data Validation and Cleanup

    You need to provide data validation at the server level for complex strings like phone numbers, email addresses, etc. You may also need to do data cleanup / standardization before moving it from source to target. Although SQL Server provides a fair number of string functions, the code developed with these built-in functions can become complex and hard to maintain or reuse. The Future of SQL Server Monitoring "Being web-based, SQL Monitor 2.0 enables you to check on your servers from almost any location" Jonathan Allen.Try SQL Monitor now.

    Read the article

  • Creating the concept of Time

    - by Jamie Dixon
    So I've reached the point in my exploration of gaming where I'd like to impliment the concept of time into my little demo I've been building. What are some common methodologies for creating the concept of time passing within a game? My thoughts so far: My game loop tendes to spend a fair bit of time sitting around waiting or user input so any time system will likely need to be run in a seperate thread. What I've currently done is create a BackgroundWorker passing in a method that contains a loop triggering every second. This is working fine and I can output information to the console from here etc. Inside this loop I have a DateTime object that is incrimented by 1 minute for every realtime second. (the game begins in the year 01/01/01) Is this a standard way of acheiving this result or are there more tried and tested methods? I'm also curious about how to go about performing time based actions (reducing player energy, moving entities around the game board, life/death etc). Thanks for any pointers or advice. I've searched around however I'm not familiar enough with the terms and so my searches are yeilding little result on this one.

    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

  • Installing Multiple OWB Patches

    - by [email protected]
    When an OUBI bug requires a fix to the Oracle Warehouse Builder (OWB) code, the fix is delivered as an MDL export file that will need to be imported and deployed in OWB. If more than one bug is being patched, then a recent question came in that asked if it would be possible to import one after the next, and then do the deploy steps once? The answer is Yes, all of the imports can be done before any of objects are deployed. Once all of the objects have been deployed, then the TCL scripts that need to be rerun can be run, and then the objects that were changed can be deployed. The order that the MDL files are loaded does not matter unless the same object is in two or more MDL files. In that case, the latest MDL file should be loaded last. For example, if two MDL files both contain changes to the SPLMAP_F_RECENT_CREW mapping, and one was created on January 2, 2009 and the second one was created on March 14, 2009, then the January 2 file should be loaded first and the March 14 file should be loaded second. Note that if the MDL files are always loaded in the order that they were created by Release Services, then this will work correctly.

    Read the article

< Previous Page | 291 292 293 294 295 296 297 298 299 300 301 302  | Next Page >