I would like to send a very large (~8GB) datastructure through the network, so I use the Marshal module to transform it into Bytes.
My problem is that the memory doubles, because we need to store both representations (initial data and Marshaled data).
Is there a simple way to Marshal into a Stream instead ? This would avoid to have the full Marshalled representation of the initial datastructure.
I thought of Marshaling to an out_channel in which I opened a pipe with a second thread and reading from the pipe in the main thread into s Stream, but I guess there might be a simpler solution.
Thanks !
Edit: Answer to a comment:
In the toplevel :
let a = Array.make (1024*1024*1024) 0. ;; (* Takes 8GB of RAM *)
let data = Marshal.to_bytes a [Marshal.Closures] ;; (* Takes an extra 8GB *)
It's not possible. You would have to modify the Marshal module to stream the data as it marshals something and to reconstruct the data in place without buffering it all first.
In the short run it might be simpler to implement your own specialized marshal function specific to your data. For an 8GiB array you might want to switch to using BigArray so you can send/recv the data without having to copy it.
Note: A 8GiB array will use 16GiB if the GC ever copies it, at least temporary.
From what I understand, MPI only allows to send data packets with a known size, not a stream of data. You could implement a custom stream type that split an incoming flow of data to packets of constant, small size (on close, you flush whatever remains in the buffer).
Also, you only can marshall arbitrary long data to a channel, because otherwise you take up too many space.
And then, you need to have a way to connect the channel to the stream, which AFAIK is not easily possible. Maybe you could start antoer ocaml process: the process would convert the flow of bytes (you can wrap a custom stream over Stream.of_channel) and send it through MPI. The main process would marshall data to the process's input channel.
Related
Standard way to open a file in Dart as a stream is to use file.openRead() which returns a Stream<List<int>>.
The next standard step is to transform this stream with utf8.decoder SteamTranformer which returns Stream<String>.
I noticed that with the files I've tried this resulting stream only emits a single event with the whole file content represented as one string. But I feel like this should not be a general case since otherwise the API wouldn't need to return a stream of strings, a Future<String> would suffice.
Could you explain how can I observe the behavior when this stream emits more than one event? Is this dependent on the file size / disk IO rate / some buffers size?
It depends on file size and buffer size, and however the file operations are implemented.
If you read a large file, you will very likely get multiple events of a limited size. The UTF-8 decoder decodes chunks eagerly, so you should get roughly the same number of chunks after decoding. It might carry a few bytes across chunk boundaries, but the rest of the bytes are decoded as soon as possible.
Checking on my local machine, the buffer size seems to be 65536 bytes. Reading a file larger than that gives me multiple chunks.
I am writing an application which sends and receives packages using UDP. However, the documentation of recv_from states:
If a message is too long to fit in the supplied buffer, excess bytes may be discarded.
Is there any way to receive all bytes and write them into a vector? Do I really have to allocate an array with the maximum packet length (which, as far as I know, is 65,507 bytes for IPv4) in order to be sure to receive all data? That seems a bit much for me.
Check out the next method in the docs, UdpSocket::peek_from (emphasis mine):
Receives a single datagram message on the socket, without removing it from the queue.
You can use this method to read a known fixed amount of data, such as a header which contains the length of the entire packet. You can use crates like byteorder to decode the appropriate part of the header, use that to allocate exactly the right amount of space, then call recv_from.
This does require that the protocol you are implementing always provides that total size information at a known location.
Now, is this a good idea?
As ArtemGr states:
Because extra system calls are much more expensive than getting some space from the stack.
And from the linked question:
Obviously at some point you will start wondering if doubling the number of system calls to save memory is worth it. I think it isn't.
With the recent Spectre / Meltdown events, now's a pretty good time to be be reminded to avoid extra syscalls.
You could, as suggested, just allocate a "big enough" array ahead of time. You'll need to track how many bytes you've actually read vs allocated though. I recommend something like arrayvec to make it easier.
You could instead implement a pool of pre-allocated buffers on the heap. When you read from the socket, you use a buffer or create a new one. When you are done with the buffer, you put it back in the pool for reuse. That way, you incur the memory allocation once and are only passing around small Vecs on the stack.
See also:
How can I create a stack-allocated vector-like container?
How large should my recv buffer be when calling recv in the socket library
How to read UDP packet with variable length in C
I have a binary file that I want to parse. The file is broken up into records that are 1024 bytes each. The high level steps needed are:
Read 1024 bytes at a time from the file.
Parse each 1024-byte "record" (chunk) and place the parsed data into a map or struct.
Return the parsed data to the user and any error(s).
I'm not looking for code, just design/approach help.
Due to I/O constraints, I don't think it makes sense to attempt concurrent reads from the file. However, I see no reason why the 1024-byte records can't be parsed using goroutines so that multiple 1024-byte records are being parsed concurrently. I'm new to Go, so I wanted to see if this makes sense or if there is a better (faster) way:
A main function opens the file and reads 1024 bytes at a time into byte arrays (records).
The records are passed to a function that parses the data into a map or struct. The parser function would be called as a goroutine on each record.
The parsed maps/structs are appended to a slice via a channel. I would preallocate the underlying array managed by the slice as the file size (in bytes) divided by 1024 as this should be the exact number of elements (assuming no errors).
I'd have to make sure I don't run out of memory as well, as the file can be anywhere from a few hundred MB up to 256 TB (rare, but possible). Does this make sense or am I thinking about this problem incorrectly? Will this be slower than simply parsing the file in a linear fashion as I read it 1024 bytes at a time, or will parsing these records concurrently as byte arrays perform better? Or am I thinking about the problem all wrong?
I'm not looking for code, just design/approach help.
Cross-posted on Software Engineering
This is an instance of the producer-consumer problem, where the producer is your main function that generates 1024-byte records and the consumers should process these records and send them to a channel so they are added to the final slice. There are a few questions tagged producer-consumer and Go, they should get you started. As for what is fastest in your case, it depends on so many things that it is really not possible to answer. The best solution may be anywhere from a completely sequential implementation to a cluster of servers in which the records are moved around by RabbitMQ or something similar.
Well, even Embarcadero states that it is not guaranteed to return accurate result of the bytes ready to read in the socket buffer, but if you look at it, when you place -1 at Socket.ReceiveBuf (this is what ReceiveLength wraps) it calls ioctlsocket with FIONREAD to determine the amount of data pending in the network's input buffer that can be read from socket s.
so, how is it not safe or bad ?
e.g: ioctlsocket(Socket.SocketHandle, FIONREAD, Longint(i));
The documentation you mention specifically says (emphasis mine)
Note: ReceiveLength is not guaranteed to be accurate for streaming socket connections.
This means that the length is not known ahead of time because it's being supplied by a stream of data. Obviously, if you don't know how big the data is that's being sent ahead of time, you can't properly set the length the client should expect.
Consider it like generic code to copy a file. If you don't know ahead of time how big the file is you'll be copying, you can't predict how many bytes you'll be copying. In the case of the socket, the stream size that's supplying the socket isn't known in advance (for instance, for data being generated real-time and sent), so there's no way to inform the client socket how much to expect.
There is a question I have been wondering about for ages and I was hoping someone could give me an answer to rest my mind.
Let's assume that I have an input stream (like a file/socket/pipe) and want to parse the incoming data. Let's assume that each block of incoming data is split by a newline, like most common internet protocols. This application could just as well be parsing html, xml or any other smart data structure. The point is that the data is split into logical blocks by a delimiter rather than a fixed length. How can I buffer the data to wait for the delimiter to appear?
The answer seems simple enough: just have a large enough byte/char array to fit the entire thing.
But what if the delimiter comes after the buffer is full? This is actually a question about how to fit a dynamic block of data in a fixed size block. I can only really think of a few alternatives:
Increase the buffer size when needed. This may require heavy memory reallocation, and may lead to resource exhaustion from specially crafted streams (or perhaps even denial of service in the case of sockets where we want to protect ourselves against exhaustion attacks and drop connections that try to exhaust resources...and an attacker starts sending fake, oversized, packets to trigger the protection).
Start overwriting old data by using a circular buffer. Perhaps not an ideal method since the logical block would become incomplete.
Dump new data when the buffer is full. However, this way the delimiter will never be found, so this choice is obviously not a good option.
Just make the fixed size buffer damn large and assume all incoming logical data blocks is within its bounds...and if it ever fills, just interpret the full buffer as a logical block...
In either case I feel we must assume that the logical blocks will never exceed a certain size...
Any thoughts on this topic? Obviously there must be a way since the higher level languages offer some sort of buffering mechanisms with their readLine() stream methods.
Is there any "best way" to solve this or is there always a tradeoff? I really appreciate all thoughts and ideas on this topic since this question has been haunting me everytime I have needed to write a parser of some sort.
There are normally two techniques for this
1) What I think readline uses - if the buffer fills return the data without the delimiter on the end
2) When the buffer fills, remeber it filled, keep reading until you get the delimiter and report an error (or truncate the record at the buffer size)
Options (2) and (3) are out as you are losing data in both cases. Option (4) of a huge fixed size buffer would not solve the problem as it is just not possible to know what size is large enough? Is it all the physical memory + swap space + the free space available in all disks everywhere in the known universe?
Resizing the buffer looks like the best solution. Say realloc to twice the size and continue writing. There is always a chance of a specially constructed stream like a DoS that tries to bring down the system. My first thought was so set an arbitrarily large size as the max_size for the buffer. However, if we could do that, we could just set that as the size of the large buffer. So, resizing the buffer looks like the best option to me.
If the protocol or you do not define a upper bound for the length of each block then I don't see how you can prevent memory exhaustion edge cases.
Assuming that there is an upper bound using a fixed size block seems like a good approach for reasonably sized limits.
If the limits are high enough that a single fixed buffer will be inefficient then I would suggest using a data structure that is implemented internally as a linked list of fixed size buffers.
Why do you have to wait to start processing?
Generally alternative 4 is sound. It doesn't however, require an "assumption", rather a definition. You simply declare that blocks are smaller than 8K and be done with it. It's not difficult to do.
Further, there's alternative 5: Start processing partial buffers. This works unless you have designed a truly pathological protocol that sends critical data at the very end of the block.
HTML, XML, JSON/YAML, etc., can all be parsed incrementally. You don't require a delimeter to do useful processing.