Hi Rafal,
On Apr 22, 2011, at 4:27 AM, Rafal Szczepanski wrote: > Hi Chuck, > > Thank you for your email. As I already told Kieran I couldn't answer before - > sorry for that. Please see my comments inline. > > On Thu, Mar 24, 2011 at 8:22 PM, Chuck Hill <[email protected]> wrote: > Hi Rafal, > > On Mar 24, 2011, at 7:50 AM, Rafal Szczepanski wrote: >> >> I would like to ask you about the optimal number of WorkerThreads >> (application settings: WOWorkerThreadCountMin & WOWorkerThreadCount). > > That depends. :-) > If someone says "that depends" I always start to fear :-) Deployment tuning and optimization depends on so many variables that it is more an art than a science. > I would say that a min of 4 is reasonable and that max should not be high > enough to allow the instance to get backlogged with requests. How many > instances are you running? > Well. We are running multiple applications on one server. The application > that is causing the problems consists of 4 instances. That is reasonable. > > Are they spread over multiple servers? > Not at this moment. If you have multiple app servers, distributing the instances over multiple servers is usually the best setup. > ` > Is concurrent request handling enabled? > Yes. We needed to enable it because we are calling third party web services > and slow response was causing the whole application to freeze. Having it enabled is good. Calling slow third party webservices in the R-R loop is probably not the best plan. If you are doing that, you may need to increases the listen queue size and max worker threads. I would view that as a work-around, not a solution. > > How much heap space are you allocating to each instance? > Currently the instances are running with 1.5GB each. That should be enough. :-) > > Are the app servers swapping? > Nope. The sum of the memory assigned is less than the server's memory. > > > >> One of the my applications is serving sometimes more than 2000 pages per >> hour. >> The current setting is 16 WorkerThreads but I noticed that from time to time >> this number is not enough and the server is creating more WorkerThreads >> which is freezing for a while the application. > > I think the more likely interpretation of this is that your app froze (was > slow to process requests), so it created more worker threads to record the > unprocessed requests waiting in line. WorkerThreads are created in response > to slowness, creating them does not cause slowness. > Well you're right - I have taken a look on one of the instances using > VisualVM. Usually there was one WorkerThread that was causing other > WorkerThreads to wait, hence when new requests were coming WebObjects was > creating additional ones. Slow database access (lots of SQL statements, or slow queries) will cause this problem. > > > >> Do you know if the number of WorkerThreads affects application’s >> performance? If not then perhaps I should start with 256 right from the >> beginning…. > > It does not affect performance directly, controls how many requests can be > queued up in an application before wotaskd considers it too full to send more > requests to. Making more workerthreads just means that the app will queue up > more pending requests. This is usually a Bad Thing (tm) as it will take so > long to process them and return a response that the user will click stop or > click the link again, thus using another worker thread and making things > worse. Find out what is making the app slow and fix that. > It seems that the problem was created by one page that was running a slow > query. As I have written before other WorkerThreads (which I assume wanted to > query the db) were waiting until the first query is finished. I have modified > the query so it runs faster and the situation improved significantly. You could also create a different EOF stack which would give you a different database connection to run this slower query in. That will not make it faster, but will eliminate the impact on the other requests. >> I’ve got also another question do you know some real life examples of >> applications built in WebObjects that are serving huge number of >> requests/heavy load? > > The iTunes music store is the classic example, as well as the Apple Store. > There are lots of others around that process a heavy load. You just need to > optimize for your specific case. > That's obviously a very promising example but I wonder how much of original > WebObjects is used there... :-) > That may be a good point. But they also have rather unique optimization requirements. What do you consider a huge number of requests/heavy load? GVC.SiteMaker at the University of Michigan gets over 1 million unique visits a month and it is only partially optimized. There is a guitar tabs site that was recently mentioned here that gets more than that. Chuck -- Chuck Hill Senior Consultant / VP Development Come to WOWODC this July for unparalleled WO learning opportunities and real peer to peer problem solving! Network, socialize, and enjoy a great cosmopolitan city. See you there! http://www.wocommunity.org/wowodc11/
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Do not post admin requests to the list. They will be ignored. Webobjects-dev mailing list ([email protected]) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com This email sent to [email protected]
