Recherche avancée

Sur d’autres sites (5039)

• Processing Big Data Problems

  8 janvier 2011, par Multimedia Mike — Big Data

  I’m becoming more interested in big data problems, i.e., extracting useful information out of absurdly sized sets of input data. I know it’s a growing field and there is a lot to read on the subject. But you know how I roll— just think of a problem to solve and dive right in.

  Here’s how my adventure unfolded.

  The Corpus
  I need to run a command line program on a set of files I have collected. This corpus is on the order of 350,000 files. The files range from 7 bytes to 175 MB. Combined, they occupy around 164 GB of storage space.

  Oh, and said storage space resides on an external, USB 2.0-connected hard drive. Stop laughing.

  A file is named according to the SHA-1 hash of its data. The files are organized in a directory hierarchy according to the first 6 hex digits of the SHA-1 hash (e.g., a file named a4d5832f... is stored in a4/d5/83/a4d5832f...). All of this file hash, path, and size information is stored in an SQLite database.

  First Pass
  I wrote a Python script that read all the filenames from the database, fed them into a pool of worker processes using Python’s multiprocessing module, and wrote some resulting data for each file back to the SQLite database. My Eee PC has a single-core, hyperthreaded Atom which presents 2 CPUs to the system. Thus, 2 worker threads crunched the corpus. It took awhile. It took somewhere on the order of 9 or 10 or maybe even 12 hours. It took long enough that I’m in no hurry to re-run the test and get more precise numbers.

  At least I extracted my initial set of data from the corpus. Or did I ?

  Think About The Future
  
  A few days later, I went back to revisit the data only to notice that the SQLite database was corrupted. To add insult to that bit of injury, the script I had written to process the data was also completely corrupted (overwritten with something unrelated to Python code). BTW, this is was on a RAID brick configured for redundancy. So that’s strike 3 in my personal dealings with RAID technology.

  I moved the corpus to a different external drive and also verified the files after writing (easy to do since I already had the SHA-1 hashes on record).

  The corrupted script was pretty simple to rewrite, even a little better than before. Then I got to re-run it. However, this run was on a faster machine, a hyperthreaded, quad-core beast that exposes 8 CPUs to the system. The reason I wasn’t too concerned about the poor performance with my Eee PC is that I knew I was going to be able to run in on this monster later.

  So I let the rewritten script rip. The script gave me little updates regarding its progress. As it did so, I ran some rough calculations and realized that it wasn’t predicted to finish much sooner than it would have if I were running it on the Eee PC.

  Limiting Factors
  It had been suggested to me that I/O bandwidth of the external USB drive might be a limiting factor. This is when I started to take that idea very seriously.

  The first idea I had was to move the SQLite database to a different drive. The script records data to the database for every file processed, though it only commits once every 100 UPDATEs, so at least it’s not constantly syncing the disc. I ran before and after tests with a small subset of the corpus and noticed a substantial speedup thanks to this policy chance.

  Then I remembered hearing something about "atime" which is access time. Linux filesystems, per default, record the time that a file was last accessed. You can watch this in action by running 'stat <file> ; cat <file> > /dev/null ; stat <file>' and observe that the "Access" field has been updated to NOW(). This also means that every single file that gets read from the external drive still causes an additional write. To avoid this, I started mounting the external drive with '-o noatime' which instructs Linux not to record "last accessed" time for files.

  On the limited subset test, this more than doubled script performance. I then wondered about mounting the external drive as read-only. This had the same performance as noatime. I thought about using both options together but verified that access times are not updated for a read-only filesystem.

  A Note On Profiling
  Once you start accessing files in Linux, those files start getting cached in RAM. Thus, if you profile, say, reading a gigabyte file from a disk and get 31 MB/sec, and then repeat the same test, you’re likely to see the test complete instantaneously. That’s because the file is already sitting in memory, cached. This is useful in general application use, but not if you’re trying to profile disk performance.

  Thus, in between runs, do (as root) 'sync; echo 3 > /proc/sys/vm/drop_caches' in order to wipe caches (explained here).

  Even Better ?
  I re-ran the test using these little improvements. Now it takes somewhere around 5 or 6 hours to run.

  I contrived an artificially large file on the external drive and did some 'dd' tests to measure what the drive could really do. The drive consistently measured a bit over 31 MB/sec. If I could read and process the data at 30 MB/sec, the script would be done in about 95 minutes.

  But it’s probably rather unreasonable to expect that kind of transfer rate for lots of smaller files scattered around a filesystem. However, it can’t be that helpful to have 8 different processes constantly asking the HD for 8 different files at any one time.

  So I wrote a script called stream-corpus.py which simply fetched all the filenames from the database and loaded the contents of each in turn, leaving the data to be garbage-collected at Python’s leisure. This test completed in 174 minutes, just shy of 3 hours. I computed an average read speed of around 17 MB/sec.

  Single-Reader Script
  I began to theorize that if I only have one thread reading, performance should improve greatly. To test this hypothesis without having to do a lot of extra work, I cleared the caches and ran stream-corpus.py until 'top' reported that about half of the real memory had been filled with data. Then I let the main processing script loose on the data. As both scripts were using sorted lists of files, they iterated over the filenames in the same order.

  Result : The processing script tore through the files that had obviously been cached thanks to stream-corpus.py, degrading drastically once it had caught up to the streaming script.

  Thus, I was incented to reorganize the processing script just slightly. Now, there is a reader thread which reads each file and stuffs the name of the file into an IPC queue that one of the worker threads can pick up and process. Note that no file data is exchanged between threads. No need— the operating system is already implicitly holding onto the file data, waiting in case someone asks for it again before something needs that bit of RAM. Technically, this approach accesses each file multiple times. But it makes little practical difference thanks to caching.

  Result : About 183 minutes to process the complete corpus (which works out to a little over 16 MB/sec).

  Why Multiprocess
  Is it even worthwhile to bother multithreading this operation ? Monitoring the whole operation via 'top', most instances of the processing script are barely using any CPU time. Indeed, it’s likely that only one of the worker threads is doing any work most of the time, pulling a file out of the IPC queue as soon the reader thread triggers its load into cache. Right now, the processing is usually pretty quick. There are cases where the processing (external program) might hang (one of the reasons I’m running this project is to find those cases) ; the multiprocessing architecture at least allows other processes to take over until a hanging process is timed out and killed by its monitoring process.

  Further, the processing is pretty simple now but is likely to get more intensive in future iterations. Plus, there’s the possibility that I might move everything onto a more appropriately-connected storage medium which should help alleviate the bottleneck bravely battled in this post.

  There’s also the theoretical possibility that the reader thread could read too far ahead of the processing threads. Obviously, that’s not too much of an issue in the current setup. But to guard against it, the processes could share a variable that tracks the total number of bytes that have been processed. The reader thread adds filesizes to the count while the processing threads subtract file sizes. The reader thread would delay reading more if the number got above a certain threshold.

  Leftovers
  I wondered if the order of accessing the files mattered. I didn’t write them to the drive in any special order. The drive is formatted with Linux ext3. I ran stream-corpus.py on all the filenames sorted by filename (remember the SHA-1 naming convention described above) and also by sorting them randomly.

  Result : It helps immensely for the filenames to be sorted. The sorted variant was a little more than twice as fast as the random variant. Maybe it has to do with accessing all the files in a single directory before moving onto another directory.

  Further, I have long been under the impression that the best read speed you can expect from USB 2.0 was 27 Mbytes/sec (even though 480 Mbit/sec is bandied about in relation to the spec). This comes from profiling I performed with an external enclosure that supports both USB 2.0 and FireWire-400 (and eSata). FW-400 was able to read the same file at nearly 40 Mbytes/sec that USB 2.0 could only read at 27 Mbytes/sec. Other sources I have read corroborate this number. But this test (using different hardware), achieved over 31 Mbytes/sec.

• Fixed a link in the API docs

  4 juin 2014, par JamesMGreene

  Fixed a link in the API docs

• Simply beyond ridiculous

  7 mai 2010, par Dark Shikari — H.265, speed

  For the past few years, various improvements on H.264 have been periodically proposed, ranging from larger transforms to better intra prediction. These finally came together in the JCT-VC meeting this past April, where over two dozen proposals were made for a next-generation video coding standard. Of course, all of these were in very rough-draft form ; it will likely take years to filter it down into a usable standard. In the process, they’ll pick the most useful features (hopefully) from each proposal and combine them into something a bit more sane. But, of course, it all has to start somewhere.

  A number of features were common : larger block sizes, larger transform sizes, fancier interpolation filters, improved intra prediction schemes, improved motion vector prediction, increased internal bit depth, new entropy coding schemes, and so forth. A lot of these are potentially quite promising and resolve a lot of complaints I’ve had about H.264, so I decided to try out the proposal that appeared the most interesting : the Samsung+BBC proposal (A124), which claims compression improvements of around 40%.

  The proposal combines a bouillabaisse of new features, ranging from a 12-tap interpolation filter to 12thpel motion compensation and transforms as large as 64×64. Overall, I would say it’s a good proposal and I don’t doubt their results given the sheer volume of useful features they’ve dumped into it. I was a bit worried about complexity, however, as 12-tap interpolation filters don’t exactly scream “fast”.

  I prepared myself for the slowness of an unoptimized encoder implementation, compiled their tool, and started a test encode with their recommended settings.

  I waited. The first frame, an I-frame, completed.

  I took a nap.

  I waited. The second frame, a P-frame, was done.

  I played a game of Settlers.

  I waited. The third frame, a B-frame, was done.

  I worked on a term paper.

  I waited. The fourth frame, a B-frame, was done.

  After a full 6 hours, 8 frames had encoded. Yes, at this rate, it would take a full two weeks to encode 10 seconds of HD video. On a Core i7. This is not merely slow ; this is over 1000 times slower than x264 on “placebo” mode. This is so slow that it is not merely impractical ; it is impossible to even test. This encoder is apparently designed for some sort of hypothetical future computer from space. And word from other developers is that the Intel proposal is even slower.

  This has led me to suspect that there is a great deal of cheating going on in the H.265 proposals. The goal of the proposals, of course, is to pick the best feature set for the next generation video compression standard. But there is an extra motivation : organizations whose features get accepted get patents on the resulting standard, and thus income. With such large sums of money in the picture, dishonesty becomes all the more profitable.

  There is a set of rules, of course, to limit how the proposals can optimize their encoders. If different encoders use different optimization techniques, the results will no longer be comparable — remember, they are trying to compare compression features, not methods of optimizing encoder-side decisions. Thus all encoders are required to use a constant quantizer, specified frame types, and so forth. But there are no limits on how slow an encoder can be or what algorithms it can use.

  It would be one thing if the proposed encoder was a mere 10 times slower than the current reference ; that would be reasonable, given the low level of optimization and higher complexity of the new standard. But this is beyond ridiculous. With the prize given to whoever can eke out the most PSNR at a given quantizer at the lowest bitrate (with no limits on speed), we’re just going to get an arms race of slow encoders, with every company trying to use the most ridiculous optimizations possible, even if they involve encoding the frame 100,000 times over to choose the optimal parameters. And the end result will be as I encountered here : encoders so slow that they are simply impossible to even test.

  Such an arms race certainly does little good in optimizing for reality where we don’t have 30 years to encode an HD movie : a feature that gives great compression improvements is useless if it’s impossible to optimize for in a reasonable amount of time. Certainly once the standard is finalized practical encoders will be written — but it makes no sense to optimize the standard for a use-case that doesn’t exist. And even attempting to “optimize” anything is difficult when encoding a few seconds of video takes weeks.

  Update : The people involved have contacted me and insist that there was in fact no cheating going on. This is probably correct ; the problem appears to be that the rules that were set out were simply not strict enough, making many changes that I would intuitively consider “cheating” to be perfectly allowed, and thus everyone can do it.

  I would like to apologize if I implied that the results weren’t valid ; they are — the Samsung-BBC proposal is definitely one of the best, which is why I picked it to test with. It’s just that I think any situation in which it’s impossible to test your own software is unreasonable, and thus the entire situation is an inherently broken one, given the lax rules, slow baseline encoder, and no restrictions on compute time.