[OT] Re: Effects of turning off sendFile in the NIO connector
On 23/03/2014 12:41, Martin Gainty wrote: Date: Sat, 22 Mar 2014 14:24:01 -0400 Subject: Effects of turning off sendFile in the NIO connector From: tomcat.ran...@gmail.com To: users@tomcat.apache.org What effect would setting useSendfile=false have on a web application using the NIO connector? I'm asking because I may want to use gzip compression in the connector. The docs state: *There is a tradeoff between using compression (saving your bandwidth) and using the sendfile feature (saving your CPU cycles). If the connector supports the sendfile feature, e.g. the NIO connector, using sendfile will take precedence over compression. The symptoms will be that static files greater that 48 Kb will be sent uncompressed.* It's trivial that adding compression uses CPU cycles, but does that imply that turning sendFile off even without enabling compression would increase CPU cycles? It's worth mentioning that the site serves a large (8mg) SWF file. I believe that was one of the pluses of NIO/sendFile, that it was good with sending large files under heavy traffic? MGwhen you enable sendfile support with request attr org.apache.tomcat.sendfile.support = true MGYou will need to set these 3 header attributes org.apache.tomcat.sendfile.filename: Canonical filename of the file which will be sent as a String org.apache.tomcat.sendfile.start: Start offset as a Long org.apache.tomcat.sendfile.end: End offset as a Long While this is relevant to how sendfile works under the covers and would be relevant if the OP was trying to use sendfile from their own servlet (rather than just using the sendfile support Tomcat provides for static content via the default servlet), is is not at all relevant to the question the OP asked. MGhtitps://tomcat.apache.org/tomcat-6.0-doc/aio.html The OP stated they are using Tomcat 7, not Tomcat 6. You failed to copy and paste that URL correctly. MGCompression: MGset compression=on @ Connector MGhttps://tomcat.apache.org/tomcat-7.0-doc/config/http.html MGI did not read that TC cannot use sendfile with any compressed Stream? MGcan you show us the URL? Clearly you did not bother to read the description for the compression attribute on the URL you just quoted. It clearly states that sendfile and compression may not be used together. Martin, Replies from you that miss the point of the original question are a regular occurrence. I'll go further and state that it is very unusual for a reply of yours to be relevant and useful to the person asking for help. Not only is it rare for a reply of yours to be helpful, in many cases your replies mislead the person seeking help and cause them to waste considerable time. Such responses are detrimental to the health of this community. Therefore, with the health of this community and mind and wearing my list owner's hat, I am asking you to take more care when you reply. If the current pattern continues then you will be unsubscribed from this mailing list and prevented from resubscribing. Mark P.S. For folks thinking that this is rather a harsh message to post on a public mailing list, a number of mailing list regulars have tried to steer Martin in the right direction off-list only to be consistently meet with a refusal by Martin to accept that there is anything wrong with any of his posts. We also only really need compression on XML data, the site has minimal HTML, SWF's don't really benefit from gzip and some binary data we send back and forth is already compressed. I could manually implement compression on XML at the application level and within the SWF, if turning off sendFile will have negative consequences. Tomcat 7.0.42 RHEL6 ~4T outbound traffic/day Best, John - To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
RE: Effects of turning off sendFile in the NIO connector
Date: Sat, 22 Mar 2014 14:24:01 -0400 Subject: Effects of turning off sendFile in the NIO connector From: tomcat.ran...@gmail.com To: users@tomcat.apache.org What effect would setting useSendfile=false have on a web application using the NIO connector? I'm asking because I may want to use gzip compression in the connector. The docs state: *There is a tradeoff between using compression (saving your bandwidth) and using the sendfile feature (saving your CPU cycles). If the connector supports the sendfile feature, e.g. the NIO connector, using sendfile will take precedence over compression. The symptoms will be that static files greater that 48 Kb will be sent uncompressed.* It's trivial that adding compression uses CPU cycles, but does that imply that turning sendFile off even without enabling compression would increase CPU cycles? It's worth mentioning that the site serves a large (8mg) SWF file. I believe that was one of the pluses of NIO/sendFile, that it was good with sending large files under heavy traffic? MGwhen you enable sendfile support with request attr org.apache.tomcat.sendfile.support = true MGYou will need to set these 3 header attributes org.apache.tomcat.sendfile.filename: Canonical filename of the file which will be sent as a String org.apache.tomcat.sendfile.start: Start offset as a Long org.apache.tomcat.sendfile.end: End offset as a Long MGhtitps://tomcat.apache.org/tomcat-6.0-doc/aio.html MGCompression: MGset compression=on @ Connector MGhttps://tomcat.apache.org/tomcat-7.0-doc/config/http.html MGI did not read that TC cannot use sendfile with any compressed Stream? MGcan you show us the URL? MGThanks We also only really need compression on XML data, the site has minimal HTML, SWF's don't really benefit from gzip and some binary data we send back and forth is already compressed. I could manually implement compression on XML at the application level and within the SWF, if turning off sendFile will have negative consequences. Tomcat 7.0.42 RHEL6 ~4T outbound traffic/day Best, John
Re: Effects of turning off sendFile in the NIO connector
MGwhen you enable sendfile support with request attr org.apache.tomcat.sendfile.support = true MGYou will need to set these 3 header attributes org.apache.tomcat.sendfile.filename: Canonical filename of the file which will be sent as a String org.apache.tomcat.sendfile.start: Start offset as a Long org.apache.tomcat.sendfile.end: End offset as a Long MGhtitps://tomcat.apache.org/tomcat-6.0-doc/aio.html MGCompression: MGset compression=on @ Connector MGhttps://tomcat.apache.org/tomcat-7.0-doc/config/http.html MGI did not read that TC cannot use sendfile with any compressed Stream? MGcan you show us the URL? MGThanks We also only really need compression on XML data, the site has minimal HTML, SWF's don't really benefit from gzip and some binary data we send back and forth is already compressed. I could manually implement compression on XML at the application level and within the SWF, if turning off sendFile will have negative consequences. Tomcat 7.0.42 RHEL6 ~4T outbound traffic/day Best, John Your first link refers to using sendFile for asynchronous writes from a servlet. Any servlet can instruct Tomcat to perform a sendfile call by setting the appropriate request attributes. Your answer is not accurate. I don't need to do anything explicitly with the headers. You should look at the documentation regarding the HTTP NIO Connector - which you already have as the second link in your reply. It discusses sendFile and compression, and how you cannot use both the NIO connector and compression if sendFile is on (on is the default).
Re: Effects of turning off sendFile in the NIO connector
On 23/03/2014 19:37, John Smith wrote: We also only really need compression on XML data, the site has minimal HTML, SWF's don't really benefit from gzip and some binary data we send back and forth is already compressed. I could manually implement compression on XML at the application level and within the SWF, if turning off sendFile will have negative consequences. Tomcat 7.0.42 RHEL6 ~4T outbound traffic/day Best, John John The consequences for disabling sendFile are extremely hard to quantify as there are so many variables. I would normally expect there to be more CPU load but how much more? No idea. It might be impossible to detect, it might leaver your CPUs pegged at 100%. The only way you will know for sure for your application is to test it with your application. Mark P.S. As you've probably figured out having been Martined, the best location for any response from Martin Gainty is /dev/null. I keep debating dropping him from the list as he causes far more harm than good. A topic for a separate thread I think. - To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
Re: Effects of turning off sendFile in the NIO connector
John The consequences for disabling sendFile are extremely hard to quantify as there are so many variables. I would normally expect there to be more CPU load but how much more? No idea. It might be impossible to detect, it might leaver your CPUs pegged at 100%. The only way you will know for sure for your application is to test it with your application. Mark P.S. As you've probably figured out having been Martined, the best location for any response from Martin Gainty is /dev/null. I keep debating dropping him from the list as he causes far more harm than good. A topic for a separate thread I think. Mark, Thanks for the answer. As suggested, I'll test it and see what happens...kind of figured I'd have to do that :) Appreciate the heads-up on MG. I was wondering why the answer seemed off. Best, Alec
