|
View:
New views
18 Messages
—
Rating Filter:
Alert me
|
|
|
DWR Scalabilty And Early Closing ModeWe have a project that will run on a Websphere 6.1 server that needs to send data to the web client in real or near real time. The data will be read from a JMS topic. We are considering using DWR "Reverse AJAX" to help push this data from the web server to the web client. There are some concerns about how this would scale as well as failover. There is also a concern about the "push" vs. "pull" concept.
I read some about Early Closing Mode in DWR. If I understood this correctly, DWR will close the http connection after 60 seconds if there is no activity. A question I have is does it release the socket or thread once that connection is closed? Would this approach be able to scale to thousands of clients or would a dedicated server component like Tibco "Ajax Messaging Service" which uses LightStreamer be needed in order to handle large numbers of clients? The disadvantage I read about Early Closing Mode is that many clients may possibly try to re-connect at the same time. How big of an issue is this? Any ideas on this would be greatly appreciated. Thanks r123 |
|
|
Re: DWR Scalabilty And Early Closing ModeDWR does a lot to protect the server against overloading. In the case of Websphere, there is no custom plugin for their async servlet, so we fall back to the servlet-api implementation which backs off connection numbers to prevent overloading things. Early closing mode is there to prevent buggy proxies messing with the responses - it's not about reducing load. What kind of latencies are important to you? If you're ok with latencies of > 60 seconds, then you can configure DWR to poll, which should reduce the server load. Joe. On Sun, Mar 16, 2008 at 4:30 PM, r123 <ronberna@...> wrote:
|
|
|
Re: DWR Scalabilty And Early Closing ModeSo would DWR in Full Streaming or Early Closing mode running on Websphere be able to handle the 10,000 concurrent users that LightStreamer advertises that it can handle? Also, is it possible to integrate DWR with LightStreamer or Flex Blaze DS servers?
|
|
|
Re: DWR Scalabilty And Early Closing ModeLightstreamer does a lot of stuff that DWR does not currently do, so I fully expect that it will scale better. There is an issue to get DWR to use lightstreamer as a conduit: http://getahead.org/bugs/browse/DWR-49, however this is very low priority. Joe On Wed, Mar 19, 2008 at 4:00 AM, r123 <ronberna@...> wrote:
|
|
|
Re: DWR Scalabilty And Early Closing ModeSo does this issue prevent DWR from currently working at all with Lightstreamer or Blaze DS?
Thanks Ron
|
|
|
Re: DWR Scalabilty And Early Closing ModeI think it depends on what you mean by "working with". Are you asking for DWR to use them as a network transport, or to be able to sit in the same page and still function. Joe.
On Thu, Apr 3, 2008 at 9:59 PM, r123 <ronberna@...> wrote:
|
|
|
Re: DWR Scalabilty And Early Closing ModeI guess have the ability to sit in the same page and function. What I am trying to determine is if at some point we decide to use a LightStreamer or Blaze DS server, would we have to remove DWR completely. Right now we are using DWR to access Java objects on the client. We are trying to determine whether or not to use DWR in polling or comet mode. A concern is the scalabilty. If we need to have a true push application with thousands of clients, then a dedicated server like LightStreamer or Blaze DS may be needed (i'm guessing).
|
|
|
Re: DWR Scalabilty And Early Closing ModeDWR will work just fine in the same page as either. There might be issues with the 2 connection limit, but you could configure DWR to use polling if that is an issue. I expect that BlazeDS's connections won't contribute to the 2 connection limit anyway. Also all browsers are upping their limits - it looks like we'll all be taking about the 6 connection limit soon, which is a much smaller problem. There are a number of different aspects to scalability. Lightstreamer is excellent at scaling to high data-rates, I'm not sure how it copes with scaling to a large number of connected clients though, which is a different matter. Note: I'm not saying that it's bad, just that I don't know how it copes. My advice would be that DWR is easy to implement, so you might be best off taking the easy route, and considering others if it can't grow to meet your needs. Joe. On Sun, Apr 6, 2008 at 3:07 AM, r123 <ronberna@...> wrote:
|
|
|
Re: DWR Scalabilty And Early Closing ModeThat is good to hear. This may be a dumb question, but in your opinion, what would be the main benefit of using DWR? I've been to the website, but would like to hear your thoughts. We are currently having to justify using DWR. What would it give us as opposed to just writing your own ajax functions. We are in the process of doing a POC using DWR with TIBCO GI for a project. I find it helps alot when we can actually access the java object from the client and then add that object to the GI matrix. It saves us from having to create our own CDF mapping. As it turns out ease of use cannot be the deciding factor (crazy I know). Just wanted to get your thoughts on why would I use DWR?
Thanks r123
|
|
|
Re: DWR Scalabilty And Early Closing ModeIf you are doing any kind of comet (which from the thread, you are), then it's a total no-brainer. DWR (or some other library that manages the complexity) will save you many months effort. Browse cometdaily.com or do a google search, and read up on the options and the issues: If, after that, you still want to write your own, you're a masochist. http://cometdaily.com/2008/02/22/comet-and-http-pipelining/ http://cometdaily.com/2008/02/05/buzzword-overload/ http://cometdaily.com/2007/10/25/http-streaming-and-internet-explorer/ http://cometdaily.com/2007/11/18/ie-activexhtmlfile-transport-part-ii/ http://www.slideshare.net/joewalker/comet-making-the-web-a-2way-medium http://en.wikipedia.org/wiki/Comet_(programming) Joe. On Tue, Apr 8, 2008 at 5:06 PM, r123 <ronberna@...> wrote:
|