Filesystem Indexing with libferris
Normalization of EA in a Relational Database
To resolve a query on a normalized EA, four tables are involved. The docmap table stores the file's URL and a synthetic integral key, the docid. The attrmap table stores the name of an EA and assigns a synthetic integral key, the attrid. One of many valuemap tables are used, depending on the type of the value, but they all follow a similar style. For example, strlookup assigns varchar values a synthetic integral key, the vid. And finally, a join table, docattrs, joins a docid with an attrid and a vid to record that a file has an attribute with a given value. Thus, to resolve a query (width<=800) if the width EA is normalized requires looking up the attrid and vid to join in the docattrs table, producing a list of docids that have a width satisfying the query.
Normalized EAs are stored directly into the docmap table as a column. To resolve the width query above, the relational database index on the docmap.width column is used to find the matching docids directly. This normalization is a space-time trade-off. Generally, for EAs for which you are planning to search often, you would consider inlining them. Many EAs that are not part of a stat(2) call or deemed very interesting should be left indexed in the attrmap, valuemap and docattrs tables.
For this example, I use my user name and a dbname of lj. The second command below creates the EA index using the non-interactive fcreate tool. The third command then adds all the JPEG files in my shared image directory to that index. You also can use feaindexadd with the -d option to list file paths explicitly on the command line. Without -d, feaindexadd tries to recurse into the paths you supply:
$ mkdir /tmp/ea-index $ fcreate --create-type=eaindexpostgresql \ --target-path=/tmp/ea-index dbname=lj user=ben # if you have setup new db, append db-exists=1 $ find /usr/share/backgrounds/images \ -name "*.jpg" \ | feaindexadd -P /tmp/ea-index --filelist-stdin
My image directory contains 42 JPEG images. Here, I query the index:
$ feaindexquery -P /tmp/ea-index '(width>=640)' Found 34 matches at the following locations: file:///usr/share/backgrounds/images/dewdop_leaf.jpg ... $ feaindexquery -P /tmp/ea-index '(size>=100k)' Found 42 matches at the following locations: file:///usr/share/backgrounds/images/dewdop_leaf.jpg ... $ feaindexquery -P /tmp/ea-index \ '(&(width<=800)(size>=100k))' Found 19 matches at the following locations: file:///usr/.../images/space/apollo08_earthrise.jpg ...
The EA index query syntax is based on “The String Representation of LDAP Search Filters” as described in RFC 2254. This is a simple syntax, providing a small set of comparative operators to make lvalue operator rvalue terms and a means to combine these terms with Boolean and (&), or (|) and not (!) operations. All terms are contained in parentheses, with operators preceding their arguments. The operators are kept simple: == for equality, <= and >= for value ranges and =~ for regex matches.
The ODBC (optionally) and PostgreSQL (always) EA indexing plugins allow you to store many versions of EAs for a file in the index. Having many versions of metadata for a file allows you to query for files based on the EA values those files once had.
To use this functionality, you have to select a time range to match the search against when querying by using a special EA. The time-restricting EAs are atime, ferris-current-time, multiversion-mtime and multiversion-atime. The last two EAs match against the mtime and atime for the file you are seeking. The ferris-current-time EA for a version of a file's index data is the time when that file was being indexed. If no time range is selected, only the latest version of metadata for each file is considered when executing a query.
Time restrictions can be given as a string, and libferris tries its best to work out the format of your time string. In the tests/timeparsing directory of the libferris distribution is a timeparse tool that accepts time values and tells you what libferris makes of your time string. More details on the permissible time strings are given in the libferris FAQ item (see Resources).
The following example of a time-based query looks for all image files that were indexed over a year ago with a given width range:
$ feaindexquery -P /tmp/ea-index \ '(&(width>=1600)(ferris-current-time<=1 year ago))'
If a large image file was indexed two years ago and subsequently replaced with a thumbnail image and re-indexed, the above query returns the file. This is because one of its versions of metadata matches the given query.
Handling the time restriction for EA queries by using the same interface as querying on EA values allows you to use all the standard query mechanisms to select your matching time range. For example, I could select documents that were indexed in 2003 with a given width or those with a specific owner that were modified in the last month:
## note, all one line $ feaindexquery -P /tmp/ea-index ' (| (& (width>=1600)(ferris-current-time>=begin 2003) (ferris-current-time<=end 2003) ) (& (owner-name==sarusama) (multiversion-mtime>=end last month) ) )
|Non-Linux FOSS: libnotify, OS X Style||Jun 18, 2013|
|Containers—Not Virtual Machines—Are the Future Cloud||Jun 17, 2013|
|Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer||Jun 12, 2013|
|Weechat, Irssi's Little Brother||Jun 11, 2013|
|One Tail Just Isn't Enough||Jun 07, 2013|
|Introduction to MapReduce with Hadoop on Linux||Jun 05, 2013|
- Containers—Not Virtual Machines—Are the Future Cloud
- Non-Linux FOSS: libnotify, OS X Style
- Linux Systems Administrator
- Validate an E-Mail Address with PHP, the Right Way
- Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer
- Senior Perl Developer
- Technical Support Rep
- UX Designer
- RSS Feeds
- Introduction to MapReduce with Hadoop on Linux
43 min 1 sec ago
45 min 33 sec ago
47 min 43 sec ago
- Bought photoshop CS5 for developing a website :(
4 hours 13 sec ago
- What the author describes
5 hours 26 min ago
- Reply to comment | Linux Journal
9 hours 36 min ago
- Reply to comment | Linux Journal
10 hours 21 min ago
- Didn't read
10 hours 32 min ago
- Reply to comment | Linux Journal
10 hours 37 min ago
- Poul-Henning Kamp: welcome to
12 hours 47 min ago
Free Webinar: Hadoop
How to Build an Optimal Hadoop Cluster to Store and Maintain Unlimited Amounts of Data Using Microservers
Realizing the promise of Apache® Hadoop® requires the effective deployment of compute, memory, storage and networking to achieve optimal results. With its flexibility and multitude of options, it is easy to over or under provision the server infrastructure, resulting in poor performance and high TCO. Join us for an in depth, technical discussion with industry experts from leading Hadoop and server companies who will provide insights into the key considerations for designing and deploying an optimal Hadoop cluster.
Some of key questions to be discussed are:
- What is the “typical” Hadoop cluster and what should be installed on the different machine types?
- Why should you consider the typical workload patterns when making your hardware decisions?
- Are all microservers created equal for Hadoop deployments?
- How do I plan for expansion if I require more compute, memory, storage or networking?