What would be the best directory structure for a massive amount of file.
Considering i have more than 20 million of files using number_id as file names (ex. 13842985.xml).
if would go with something like
filename : 13842985.xml
directory : 1/3/8/13842985.xml
How can i do this properly wherein all files are scattered evenly on each directories and subdirectories.
You could create the directory structure like a trie.
Change your method slightly to this instead:
filename : 13842985.xml
directory : 842/985/13842985.xml # use the 6 last to create the directory name
I am assuming the filenames are somewhat random. This scheme will create 1000 top folders, each containing 1000 subfolders. By starting from the last digits instead of the first, you will be protected against long filenames:
filename : 138429851234.xml
directory : 851/234/138429851234.xml
Hope this helps!
Edit: By hashing the filename first and using this number instead, you'll avoid degenerate cases (for example, varying only in the beginning).
Do some benchmarking to figure out where the tradeoff between having to scan through multiple directorie becomes cheaper than having to scan through "many" files in a single directory.
At some point the file system overhead of opening/scanning/security-checking/etc... on each directory layer you add will be higher than the savings from having to parse the directory to find the single file you want. That's the level you'd do your split/layering cutoff.
Related
I know how to use glob() to fetch all image files in a directory, but I want to save retrieval time and only fetch the ones I need in the first place.
I am building a car dealership website, and there is a directory where all the vehicle photos get stored. Photos that are associated with a vehicle for sale start with the letter "v" and then the database ID, and then a dot before the model of vehicle.
Here is a sample list of files in a directory:
v313.2014.toyota.camry.0.jpg
v313.2014.toyota.camry.1.jpg
fordfusion.jpg
fordfusion2.jpg
v87.2015.honda.civic.0.jpg
v87.2015.honda.civic.1.jpg
2014.ford.escape.0.jpg
2014.ford.escape.1.jpg
Out of those files, only fordfusion.jpg, fordfusion2.jpg, 2014.ford.escape.0.jpg, 2014.ford.escape.1.jpg should be returned by glob().
I hope this is possible without retrieving all the image files and then going through the array with a regex because 90% of the images being fetched wouldn't be necessary.
I hope this is possible without retrieving all the image files and then going through the array with a regex because 90% of the images being fetched wouldn't be necessary.
Unless there is an extremely large number of files in the directory, this isn't worth worrying about. glob() internally has to iterate through all files in the folder to check their names against the pattern anyway; doing it in PHP code with a regular expression will perform equally well.
If there really is a very large number of files in the directory… don't do that. Large directories perform very poorly in general, and many filesystems have limits on the number of files in a folder. (For instance, the ext3 file system, common on older Linux systems, has a limit of around 32,768 files in a single directory.) Split them up into multiple directories.
To answer the question directly, though, there is no way to do this with a glob() pattern. It's possible to match all the files that do have names starting that way, but there's no way to invert the match. (You could check for [^v]* and v[^0-9]* as two separate patterns, but there's no way to combine them into a single pattern.)
if you're confident that files you don't want to retrieve starts with the letter v or starts with v followed by any digit you can try use the following regex
^[^v]+$|^v[^\d].+
check the Demo
What's the most efficient way to grab a list of new files after a given date in php, or perhaps using a system call?
I have full control over how the files are stored as I receive them, so I thought maybe storing them in a folder structure like year/month/day/filename would be best, then all I have to do is scan for the directories greater than or equal to the date I want to retrieve using scandir and casting the directory name to int values. But I am not sure if I'm missing something that would make this easier/faster. I'm interested in the most efficient way of doing this as there will be a lot of files building up over time and I don't want to have to rescan old directories. Basically the directory structure should lend itself well to efficient manual filtering but I wanted to check to see if I'm missing something.
Simple example usage:
'2012/12/1' contains files test1.txt, test2.txt
'2012/12/2' => test3.txt, test4.txt
'2011/11/1' => test5.txt
'2011/11/2' => test6.txt
If I search for files on or after 2011/11/2, then I want everything except test5.txt to be returned.
Thanks in advance for any insight!
edit: the storing and actual processing of files are two separate processes, so I can't just process them as they come in which would obviously be the best solution.
Generally speaking I create directories like YYYY/MM/DD to store my files, often with another level for different sources. Sometimes I'll use YYYY-MM/DD or something similar. Note that there are only 3652 days in a decade, so you could even have a single level like YYYY-MM-DD and not get directories that are so large that they're hard to work with. If you have a filesystem that indexes directories, you can easily have 10s of thousands of files in a directory, otherwise one thousand should probably be your upper limit.
To process the files, I don't bother doing any actual searching of directory names. Since I know what date I'm interested in, I can simply generate the paths and scan only the directories containing files in the proper date range.
For example, let's say I want to process all files for the past week:
for $date = today() - 7 to today():
$path = strftime("%Y/%m/%d", $date)
for $filename in getFiles($path):
processFile($path, $filename)
It looks like you are on either linux or mac based on how you wrote your path.
The find command can return a list of files modified (or accessed) within a certain date.
// find files that were modified less than 30m ago
$filelist = system("find /path/to/files -type f -mmin -30");
I think system calls should be used sparingly since they reduce portability.
Storing in directories as you mentioned makes sense as it will reduce the search space.
I'm currently building an application that will generate a large number of images (a few tens of thousand of images, possibly more but not in the near future at least). And I want to be able to determine whether a file exists or not and also send it to clients over http (I'm using apache is my web server).
What is the best way to do this? I thought about splitting the images to a few folders and reduce the number of files in each directory. For example lets say that I decide that each file name will begin with a lower letter from the abc. Than I create 26 directories and when I want to look for a file I will add the name of the directory first. For example If I want a file called "funnyimage2.jpg" I will save it inside a directory called "f". I can add layers to that structure if that is required.
To be honest I'm not even sure if just saving all the files in one directory isn't just as good, so if you could add an explanation as to why your solution is better it would be very helpful.
p.s
My application is written in PHP and I intend to use file_exists to check if a file exists or not.
Do it with a hash, such as md5 or sha1 and then use 2 characters for each segment of the path. If you go 4 levels deep you'll always be good:
f4/a7/b4/66/funnyimage.jpg
Oh an the reason its slow to dump it all in 1 directory, is because most filesystems don't store filenames in a B-TREE or similar structure. It will have to scan the entire directory to find a file often times.
The reason a hash is great, is because it has really good distribution. 26 directories may not cut it, especially if lots of images have a filename like "image0001.jpg"
Since ext3 aims to be backwards compatible with the earlier ext2, many of the on-disk structures are similar to those of ext2. Consequently, ext3 lacks recent features, such as extents, dynamic allocation of inodes, and block suballocation.[15] A directory can have at most 31998 subdirectories, because an inode can have at most 32000 links.[16]
A directory on a unix file system is just a file that lists filenames and what inode contains the actual file data. As such, scanning a directory for a particular filename boils down to the equivalent operation of opening a text file and scanning for a line with a particular piece of text.
At some point, the overhead of opening that directory "file" and scanning for your filename will outweigh the overhead of using multiple sub-directories. Generally, this won't happen until there's many thousands of files. You should benchmark your system/server to find where the crossover point is.
After that, it's a simple matter of deciding how to split your filenames into subdirectories. If you're allowing only alpha-numeric characters, then maybe a split based on the first 2 characters (1,296 possible subdirs) might make more sense than a single dir with 10,000 files.
Of course, for every additional level of splitting you add, you're forcing the system to open yet another directory "file" and scan for your filename, so don't go too deep on the splits.
your setup is okay. Keep going this way
It seems that you are on the right path. Another post at ServerFault seems to confirm that you are doing the right thing.
I think linux has a limit to the amount of files a directory can contain; it might be best to split them up.
With your method, you can have the same exact image with many different file names. Also, you'll have more images that start with "t" than you would "q" so the directory would still get large. You might want to store them as MD5-HASH.jpg instead. This will eliminate duplicates and have a more even distribution over 36 directories.
Edit: Like Evert mentions, you can do a multi-level directory structure to keep the directory size even smaller.
Basically i have simple form which user uses for files uploading. Files should be stored under /files/ directory with some subdirectories for almost equally splitting files. e.g. /files/sub1/sub2/file1.txt
Also i need to not to store equal files (by filename).
I have own solution. Calculate sha1 from filename. Take first 5 symbols - abcde for example and put file in /files/a/b/c/d/e/ this works well, but gives situation when one folder contains 4k files, 2nd 6k files. Is there any way to make files count be more closer to each other? Max files count can be 10k or 10kk.
Thanks for any help.
P.S. May be i explained something wrong, so once again :) Task is simple - you have only html and php (without any db) and files directory where you should store only uploaded files without any own data. You should develop script that can handle storing uploads to files directory without storing duplicates (by filename) and split uploaded files by subdirectories by files count in each directory (optimal and count files in each directory should be close to each other).
I have no idea why you want it taht way. But if you REALLY have to do it this way, iI would suggest you set a limit how many bytes are stored in each folder. Everytime you have to save the data you open a log with
the current sub
the total number of bytes written to that directory
If necesary you create a new sub diretory(you coulduse th current timestempbecause it wont repeat) and reset the bytecount
Then you save the file and increment the byte count by the number of bytes written.
I highly doubt it is worth the work, but I do not really know why you want to distribute the files that way.
I have an PHP website which creates and stores HTML template files on server based on user input.one user can create many templates.So to store the template files and associate them with the DB record ,what I do is-
"templates" is the table which hold other information about the template such as who created it etc. with unique auto-increment id as template_id
for example -
if template id is 1001
I convert it to hex which is 03e9
Now I split the hex number into 03 & e9 (after two numbers) becomes folder and e9 becomes
file with some extension as "e9.tpl"
This is how I can find out template from the file system if I know the template ID.I dont need to separately store the path to the file.
is it a good approach ? any shortfalls of this approach ? is there any other approach better than this ?
What are the advantages / disadvantages of storing the path to file in the database itself ? for example to enable using different discs serving templates etc.?
If the ID in the DB table is already UNIQUE, why transform the id for the filesystem at all?
Just add a file 1001.tpl and you are all set. If you want to have template files sorted into folders, use the User ID (which I assume to be UNIQUE too), so you get folder 124/1001.tpl.
Depending on your deployment process, you will want to keep the created files outside the application folder, so not accidently delete them when updating the application.
Are you doing this because you are worried that you might run out of file entries/inodes in the directory? In ext3 the practical limit is somewhere around 100.000 files (and 32.000 dirs).
Creating a directory structure on the fly is better done using modulu as in $dir = $id % 1000 and then put the new template in that dir ($dir/$id.tpl). That strategy will create max 1000 dirs and you have thus made it possible to handle around 100.000.000 files.
I don't see any reason for messing with hexadecimal values or substrings.
If you have to hit the database to get the id, you may be just as well off storing the template in it as well. But there's nothing categorically wrong with storing them on the file system. I generally would.
When you hit 65,536, you'll get 0x10000. Make sure your code can handle that. I'd be more apt to store 0x1234 like: 1/1234.tpl, just for the sake of clarity. Note that by virtue of sequential IDs, your folders will fill up sequentially.
I'd probably not even convert them to hex. You could use a modulus operator to determine which folder to put them in. Figure out how many files you are likely to have and use that to determine how many folders you want.
For example:
$path = ($id % NUMBER_OF_FOLDERS) . "/$id.tpl"
where $id is the template id in decimal.
I don't understand the point in separating the hex into two parts to create different folders... That could create hundreds and hundreds of different folders which would become a complete mess on your server. Why not just store on the templates in one single folder with the hex value as the file name, such as 03e9.tpl?