I am trying to send some very large files (>200MB) through an Http output stream from a Java client to a servlet running in Tomcat.
My protocol currently packages the file contents in a byte[] and that is placed a a Map<String, Object> along with some metadata (filename, etc.), each part under a "standard" key ("FILENAME" -> "Foo", "CONTENTS" -> byte[], "USERID" -> 1234, etc.). The Map is written to the URL connection output stream (urlConnection.getOutputStream()). This works well when the file contents are small (<25MB), but I am running into Tomcat memory issues (OutOfMemoryError) when the file size is very large.
I thought of sending the metadata Map first, followed by the file contents, and finally by a checksum on the file data. The receiver servlet can then read the metadata from its input stream, then read bytes until the entire file is finished, finally followed by reading the checksum.
Would it be better to send the metadata in connection headers? If so, how? If I send the metadata down the socket first, followed by the file contents, is there some kind of standard protocol for doing this?
You will almost certainly want to use a multipart POST to send the data to the server. Then on the server you can use something like commons-fileupload to process the upload.
The good thing about commons-fileupload is that it understands that the server may not have enough memory to buffer large files and will automatically stream the uploaded data to disk once it exceeds a certain size, which is quite helpful in avoiding OutOfMemoryError type problems.
Otherwise you are going to have to implement something comparable yourself. It doesn't really make much difference how you package and send your data, so long as the server can 1) parse the upload and 2) redirect data to a file so that it doesn't ever have to buffer the entire request in memory at once. As mentioned both of these come free if you use commons-fileupload, so that's definitely what I'd recommend.
I don't have a direct answer for you but you might consider using FTP instead. Apache Mina provides FTPLets, essentially servlets that respond to FTP events (see http://mina.apache.org/ftpserver/ftplet.html for details).
This would allow you to push your data in any format without requiring the receiving end to accommodate the entire data in memory.
Regards.
Related
I am using Spring-MVC and I need to send a MP4 file back to the user. The MP4 files are, of course, very large in size (> 2 GB).
I found this SO thread Downloading a file from spring controllers, which shows how to stream back a binary file, which should theoretically work for my case. However, what I am concerned about is efficiency.
In one case, an answer may implicate to load all the bytes into memory.
byte[] data = SomeFileUtil.loadBytes(new File("somefile.mp4"));
In another case, an answer suggest using IOUtils.
InputStream is = new FileInputStream(new File("somefile.mp4"));
OutputStream os = response.getOutputStream();
IOUtils.copy(is, os);
I wonder if either of these are more memory efficient than simply defining a resource mapping?
<resources mapping="/videos/**" location="/path/to/videos/"/>
The resource mapping may work, except that I need to protect all requests to these videos, and I do not think resource mapping will lend itself to logic that protects the content.
Is there another way to stream back binary data (namely, MP4)? I'd like something that's memory efficient.
I would think that defining a resource mapping would be the cleanest way of handling it. With regards to protecting access, you can simply add /videos/** to your security configuration and define what access you allow for it via something like
<security:intercept-url pattern="/videos/**" access="ROLE_USER, ROLE_ADMIN"/>
or whatever access you desire.
Also, you might consider saving these large mp4's to a cloud storage and/or CDN such as Amazon S3 (with our without CloudFront).
Then you can generate unique urls which will last as long as you want them to. Then the download is handled by Amazon rather than having to use the computing power, data space, and memory of your web server to serve up the large resource files. Also, if you use something like CloudFront, you can configure it for streaming rather than download.
Loading the entire file into memory is worse, as well as using more memory and being non-scalable. You don't transmit any data until you've loaded it all, which adds all that latency.
I am using URL class in java and I want to read bytes through Input Stream from a specific byte location in the stream instead of using skip() function which takes a lot of time to get to that specific location.
I suppose it is not possible and here is why: when you send GET request, remote server does not know that you are interested in bytes from 100 till 200 - he sends you full document/file. So you need to read them, but don't need to handle them - that is why skip is slow.
But: I am sure that you can tell server (some of them support it, some - don't) that you want 100+ bytes of file.
Also: see this to get in-depth knowledge about skip mechanics: How does the skip() method in InputStream work?
The nature of streams mean you will need to read through all the data to get to the specific place you want to start from. You will not get faster than skip() unfortunately.
The simple answer is that you can't.
If you perform a GET that requests the entire file, you will have to use skip() to get to the part that you want. (And in fact, the slowness is most likely because the server has to send all of the data that is being skipped to the client. That is how TCP/IP works ...)
However, there is a possible alternative. The HTTP 1.1 specification supports partial fetching documents using the Range header. If your server supports this, then you can request the server to send you just the range of the document that you are interested in. However, you may need to deal with the case where the server ignores the Range header and sends the entire document anyway.
I have an JAX-RS web service that calls a db2 z/os database and returns about 240mb of data in a resultset. I am then creating an OutputStream to send this data to the client by looping through the resultset and adding a few XML tags for my output.
I am confused about what to use PrintWriter, BufferedWriter or OutputStreamWriter. I am looking for the fastest way to deliver the data. I also don't want the JVM to hold onto this data any longer than it needs to, so I don't use up it's memory.
Any help is appreciated.
You should use
BufferedWriter
Call .flush() frequently
Enable gzip for best compression
Start thinking about a different way of doing this. Can your data be paginated? Do you need all the data in one request.
If you are sending a large binary data, you probably don't want to use xml. When xml is used, binary data is usually represented using base64 which becomes larger than the original binary and uses quite a lot of CPU for the conversion into base64.
If I were you, I'd send the binary separate from the xml. If you are using WebService, MTOM attachment could help. Otherwise you could send the reference to the binary data in the xml, and let the app. download the binary data separately.
As for the fastest way to send binary, if you are using weblogic, just writing on the response's outputstram would be ok. That output stream is most probably buffered and whatever you do probably won't change the performance anyways.
Turning on gzip could also help depending on what you are sending (e.g. if you are sending jpeg (stuff that is already compressed) or something, it won't help a lot but if you are sending raw text then it can help a lot, etc.).
One solution (which might not work for you) is to spawn a job / thread that creates a file and then notifies the user when the file is ready to download, in this way you're not tied to the bandwidth of the client connection (and you can even compress the file properly, before the client downloads it)
Some Business Intelligence and data crunching applications do this, specially if the process takes some time to generate the data.
The output max speed will me limited by network bandwith and i am shure any Java OutputStream will be much more faster than you will notice the difference.
The choice depends on the data to send: is that text (lines) PrintWriter is easy, is that a byte array take OutputStream.
To hold not too much data in the buffers you should call flush() any x kb maybe.
You should never use PrintWriter to output data over a network. First of all, it creates platform-dependent line breaks. Second, it silently catches all I/O exceptions, which makes it hard for you to deal with those exceptions.
And if you're sending 240 MB as XML, then you're definitely doing something wrong. Before you start worrying about which stream class to use, try to reduce the amount of data.
EDIT:
The advice about PrintWriter (and PrintStream) came from a book by Elliotte Rusty Harold. I can't remember which one, but it was a few years ago. I think that ServletResponse.getWriter() was added to the API after that book was written - so it looks like Sun didn't follow Rusty's advice. I still think it was good advice - for the reasons stated above, and because it can tempt implementation authors to violate the API contract
in order to get predictable behavior.
An extra requirement is that the attachments can be stored as a stream, as there might be potentially very large binaries that has to be saved. Videos etc.
I have looked at Voldemort and other key value stores, but they all seem to expect byte arrays, which is completely out of the question.
This should, preferrably, be written in Java, and be embeddable.
The use case is:
I have written a HTTP Cache library which has multiple backends.
I have a Memory based one (using hashmap and Byte arrays), Derby database, persistent hashmap with file attachment, EHCache with file attachment.
I was hoping there was something out there which didn't use the file system, or if it does, it's transparent from the API.
I am storing the Headers with some more meta information in a datastore. But I also need to store the payload of the HTTP response.
The HTTP response payload might be VERY big, thats why I need to use streaming.
Why is a byte[] value out of the question? Any object graph can be serialized into a byte array!
Have you looked at sleepycat's Berkeley DB (it's free)?
EDIT - having seen jhedding's comment, it seems like you need to store data which is too big to fit into a single JVM in one go. Have you:
Checked that it won't fot into a 64-bit JVM?
Tried using a network file system? (NAS or whatever)
I am developing a webapplication which reads very large files > 50MB and display it. The Java Spring backend will read these files and its content with CXF. My problem is, after it reads a 50 megabyte file, the size of used heap is increasing by 500 megabytes. I read this fils as a String and this String will be sent to frontend. Is there any idea, tricks how can I reduce the Java heap usage? I tried nio, spring's Resource class, but nothing helped.
A dirty way to do this is to have the Spring #Controller method accept an OutputStream or Writer argument—the framework will supply the raw output stream of the HTTP response and you can write directly into it. This sidesteps all the nice logic of content type management etc.
A better way is to define a custom type which will be returned from the controller method and a matching HttpMessageConverter which will (for example) use the information in that object to open the appropriate file and write its contents into the output stream.
In both cases you will not read the file into RAM; you'll use a single, fixed-size buffer to transfer the data directly from the disk to the HTTP response.