GSS/Pithos update: iPhone, Android, Mongo & getting to 2.0

It has been a while since my last post on GSS/Pithos, and there have been quite a few important announcements that I have not blogged about, only tweeted:

• a shiny native iPhone client
• a brand new Android client
• a back-office application for statistics and admin tasks
• public folders for hosting static web sites
• a slew of performance, usability and bug fixes for the web & desktop clients
• new user registration and quota upgrade workflows

And since the code base has been adopted by other European NRNs, we have also improved our release engineering process and documentation. This however was just the tip of the iceberg. What we are now close to completing, is nothing short of an almost complete rewrite of the core GSS server, aiming at even greater performance and scalability. We are about to release GSS 2.0.

The initial plan was far more modest. We had identified the PostgreSQL installation that stores all of the the system’s data besides the actual files, as a future bottleneck when user adoption would start to accelerate. In such an event there were a few low-to-medium-cost solutions that we could pursue, like replication and manual sharding, but since the problem seemed best suited for a NoSQL solution, we decided to bite the bullet and attempt a transition.

Our field search included lots of products, like Mongo, Couch, Cassandra, Riak, Redis, HBase, MemcacheDB and others. The basic requirements were high write throughput, sharding, replication and easy migration. Most of the products boasted high performance for writes and lots supported replication. In the end we discarded the ones that didn’t support sharding, which limited the options a lot, and we investigated the various APIs, trying to figure out what it would mean to rewrite GSS for one of these. In the end we chose Mongo for its rich query support, since it was the best fit for our problem at hand, making the transition of the existing code base easier. Although we’d love to have had the opportunity to try different ports for each of those datastores, it is most likely that we would have had to rewrite the entire server from scratch, which was not an option.

The initial plan was to try to isolate the changes to the data access layer, essentially building a new adapter for the middle tier that would appear almost identical to the existing code. This effort was fruitless however, since we could not find satisfactory solutions for various issues, like transactions, data mapping, entity lifecycle, etc. We came to the conclusion that even if we managed to build a not-too-leaky abstraction for Mongo, it would have probably killed the performance, making the switch from PostgreSQL pointless.

So we ditched JPA and container-managed transactions and decided to make a simpler POJO mapping to Mongo using Morphia, a Java data mapper for Mongo that takes care of the nitty-gritty details and repetitive coding, for only a minor performance hit. Since Mongo only supports atomicity for a single query, we reverted to manual transaction handling and decided to ditch the EJBs that formed the backbone of GSS altogether, since they didn’t buy us anything at that point. The new GSS server would be a set of servlets, service and domain objects, wired together via Google Guice. Since we could host the new server in a plain servlet container, we went from JBoss to Jetty, which vastly reduced the server startup time from 60 seconds down to 2. As a consequence, development round-trip times went down, and coding was fast again. Lots of object copying and conversions from entities to DTOs and back again were now unnecessary, since there was no more a transaction boundary at the session bean layer. This kept memory usage low, increased cache locality and boosted performance. There are lots of interesting details about our approach, but they would be better discussed in separate blog posts. If you are interested, you can already check out the code in the gss2 branch of the GSS repository.

The next generation GSS server is not production ready yet, but we'll be getting there soon, and so far it has been a great ride!

GSS authentication

As I've described before, the GSS API is a REST-like interface for interacting with a GSS-based service, like Pithos. Using regular HTTP commands a client can upload, download, browse and modify files and folders stored in the GSS server. These interactions have two important properties: they store no client state to the server and they are not encrypted. I have already mentioned the most important benefits from the stateless architecture of the GSS protocol, namely scalability and loose coupling along the client-server boundary. In the same line of thought, SSL/TLS encryption of the transport was avoided for scalability reasons. Consequently, these two properties of the communication protocol, lead to another requirement: message authentication for each API call.

Since no session state is stored in the server, the client must authenticate each request as if it were the first one. Traditional web applications use an initial authentication interaction between client and server, that creates a unique session ID, which is transmitted with each subsequent request in that same session. While using this ID, the client does not need to authenticate again to the server. Stateless protocols, like WebDAV for instance, cannot rely on such an ID and have to transmit authentication data in each call. Ultimately the tradeoff is a minor increase in the message payload, in return for a big boost in scalability. The HTTP specification defines an Authorization header for use in such cases and WebDAV uses the HTTP Digest Access Authentication scheme. The GSS API uses a slight variation in that theme, blended with some ideas from OAuth request signing.

Essentially the standard HTTP Authorization header is populated with a concatenation of the username (for identifying the user making the request) and a HMAC-SHA1 signature of the request. The request signature is obtained by applying a secret authentication token to a concatenated string of the HTTP method name, the date and time of the request and the actual requested path. The GSS API page has all the details. The inclusion of the date and time helps thwart replay attacks using a previously sniffed signature. Furthermore, a separate header with timestamp information, X-GSS-Date, is used to thwart replay attacks with a full copy of the payload. The authentication token is issued securely by the server to the client and is the time-limited secret that is shared between the client and server. Since it is not the user password that is used as a shared secret, were the authentication token to be compromised, any ill effects would have been limited to the time period the token was valid. There are two ways for client applications to obtain the user's authentication token, and they are both described in detail in the API documentation.

In my experience, protocol descriptions are one thing and working code is a totally different one. In that spirit, I'm closing this post with a few tested implementations of the GSS API signature calculation, for client applications in different languages. Here is the method for signing a GSS API request in Java (full code here):

public static String sign(String httpMethod, String timestamp,
        String path, String token) {
  String input = httpMethod + timestamp + path;
  String signed = null;

  try {
    System.err.println("Token:" + token);
    // Get an HMAC-SHA1 key from the authentication token.
    System.err.println("Input: " + input);
    SecretKeySpec signingKey = new SecretKeySpec(
        Base64.decodeBase64(token.getBytes()), "HmacSHA1");

    // Get an HMAC-SHA1 Mac instance and initialize with the signing key.
    Mac hmac = Mac.getInstance("HmacSHA1");
    hmac.init(signingKey);

    // Compute the HMAC on the input data bytes.
    byte[] rawMac = hmac.doFinal(input.getBytes());

    // Do base 64 encoding.
    signed = new String(Base64.encodeBase64(rawMac), "US-ASCII");

  } catch (InvalidKeyException ikex) {
    System.err.println("Fatal key exception: " + ikex.getMessage());
    ikex.printStackTrace();
  } catch (UnsupportedEncodingException ueex) {
    System.err.println("Fatal encoding exception: "
                + ueex.getMessage());
  } catch (NoSuchAlgorithmException nsaex) {
    System.err.println("Fatal algorithm exception: "
                + nsaex.getMessage());
    nsaex.printStackTrace();
  }

  if (signed == null)
    System.exit(-1);
  System.err.println("Signed: " + signed);
  return signed;
}

Here is a method for signing GSS API requests in JavaScript, that uses the SHA-1 JavaScript implementation by Paul Johnston (full code here):

function sign(method, time, resource, token) {
  var q = resource.indexOf('?');
  var res = q == -1? resource: resource.substring(0, q);
  var data = method + time + res;
  // Use strict RFC compliance
  b64pad = "=";
  return b64_hmac_sha1(atob(token), data);
}

Here is a Tcl implementation by Alexios Zavras (full code here):

proc ::rest::_sign {id} {
  variable $id
  upvar 0 $id data
  set str2sign ""
  append str2sign $data(method)
  append str2sign $data(date)
  set idx [string first ? $data(path)]
  if {$idx < 0} {
    set str $data(path)
  } else {
    incr idx -1
    set str [string range $data(path) 0 $idx]
  }
  # append str2sign [::util::url::encode $str]
  append str2sign $str
  puts "SIGN $str2sign"
  set sig [::base64::encode [::sha1::hmac -bin -key $::GG(gsstoken) $str2sign]]
  set data(signature) $sig
}

I'd love to show implementations of the signature calculation for other languages as well, but since my time is scarce these days, I could use some help here. If you've written one yourself and you'd like to share, leave a comment and I'll update this post, with proper attribution of course.

Retrying transactions with exponential backoff

The method I described in my previous post about retrying transactions could use some improvement. A deficiency that will only become relevant in highly congested servers is the constant retry interval. When two or more transactions try to commit at the same time and fail, with the code from the last post they will retry probably simultaneously again. Random runtime events (like process/thread scheduler decisions, JIT compiler invocations, etc.) might help avoid collisions, but in general the collided transactions may well collide again. And again. And then again, until the specified maximum number of retries is reached.

The general method to alleviate such problems is to randomize the retry intervals. The most well-known algorithm in this category is called exponential backoff. This is one way to implement it for the utility method in my previous post (full code in gss):

public T tryExecute(final Callable<T> command) throws Exception {
  T returnValue = null;
  // Schedule a Future task to call the command after delay milliseconds.
  int delay = 0;
  ScheduledExecutorService executor = Executors.newScheduledThreadPool(1);
  for (int i = 0; i < TRANSACTION_RETRIES; i++) {
    final int retry = i;
    ScheduledFuture<T> future = executor.schedule(new Callable<T>() {

      @Override
      public T call() throws Exception {
        return command.call();
      }
    }, delay, TimeUnit.MILLISECONDS);

    try {
      returnValue = future.get();
      break;
    } catch (ExecutionException e) {
      Throwable cause = e.getCause();
      if (!(cause instanceof EJBTransactionRolledbackException) ||
          retry == TRANSACTION_RETRIES - 1) {
        executor.shutdownNow();
        throw new Exception(cause);
      }
      delay = MIN_TIMEOUT + (int) (MIN_TIMEOUT * Math.random() * (i + 1));
      String origCause = cause.getCause() == null ?
          cause.getClass().getName() :
          cause.getCause().getClass().getName();
      logger.info("Transaction retry #" + (i+1) + " scheduled in " + delay +
          " msec due to " + origCause);
    }

  }
  executor.shutdownNow();
  return returnValue;
}

The notable changes in this version are the delayed invocations of the Callable and the retry interval calculation. The former is accomplished using a ScheduledFuture from the excellent utilities in java.util.concurrent, which gets executed after a random time period. The interval calculation is something that can be implemented in a number of ways, either with monotonically increasing intervals or not. I opted for the formula above since it provided the fastest collision elimination in my tests, much faster than the monotonically increasing interval formulas I tried. The MIN_TIMEOUT constant is more of a black art. It should be tuned to the particular hardware and network setup in order to attain maximum efficiency: the minimum number of retries with the minimum interval between retries.

Another issue raised in a comment in my previous post was that the exception I am using as guard, EJBTransactionRolledbackException, may be too generic for this purpose. This is definitely true, EJBTransactionRolledbackException just wraps an OptimisticLockException as one would expect for this case, which in turn wraps the Hibernate-specific StaleObjectStateException that is thrown initially. However, contention in a database table, index or row might not necessarily result in optimistic locking violations, but in deadlocks as well, when one transaction holds exclusive locks on resources that are needed by the other and vice versa. Deadlock detection is performed by the DBMS, which forces a rollback of both transactions. This time however the initial exception (at least in my testing, might be DBMS or JDBC driver-specific) is GenericJDBCException which gets wrapped in a PersistenceException, which in turn is put inside a EJBTransactionRolledbackException before being received by our call site. Therefore, the generic nature of the guard is not a problem in this case. On the contrary it covers more relevant cases and one might argue even that there is no better service that can be offered to the caller, than retrying in all such occasions.

Retrying transactions in Java

When building a web-based system, the most often encountered persistence solution is the use of a relational DBMS. This provides many benefits, but probably the most important one is that its behavior and usage are widely understood after so many decades of use. The simplicity of the ACID transactional model, coupled with the nice fit of the HTTP request/response protocol, make for a killer combination.

The simplicity starts to break down however, when the load on the DBMS server increases, causing transactions to start failing. The DBMS can guarantee that no data corruption will occur, but handling such cases is not that simple. The scalable Optimistic Locking model requires transactions to be resubmitted when failure occurs. Since this is supposed to be rather rare (otherwise optimistic locking should probably not be used), developers often cheat by not dealing with such errors, letting them bubble up to the user, who is then responsible for resubmitting the transaction. This leads to simple, elegant codebases and unhappy users.

Adding ORM to the mix only makes things worse. In JavaEE, the JPA API hides much of the complexity of dealing with relational databases, but corner cases become harder. Add the container-managed transaction boundaries provided by session EJBs and one might be tempted to raise his hands up in despair.

However a solution is not as hard as it sounds. Provided that session EJB methods are sufficiently free from side effects, one can replay a transaction for a number of times from the call site (external to the session bean) with a small amount of code and a small added complexity to the call sites. Since I seem to come up against this requirement quite often, I thought I should describe one way to do it. The following tryExecute() helper method, retries any action supplied in a standard Callable interface for a number of times, in the face of optimistic locking violations:

public T tryExecute(Callable<T> command) throws Exception {
  T returnValue = null;
  for (int i = 0; i < TRANSACTION_RETRIES; i++) {
    try {
      returnValue = command.call();
      break;
    } catch(EJBTransactionRolledbackException trbe) {
      if (i == TRANSACTION_RETRIES - 1)
        throw trbe;
      } catch (Exception e) {
        throw e;
      }
      logger.debug("Transaction retry: " + (i+1));
    }
  return returnValue;
}

By abstracting away the session bean method code in a parameterized class we can reuse this method in every place that needs to retry transactions. Using a Callable as a parameter lets us return any values received from the session bean. For supplying bean methods that do not return values, we could create Runnable instances and wrap them to Callable using Executors.callable(). You can see the original code here.

If you use the above code in a JavaSE environment without EJB, you should probably need to catch OptimisticLockException for JPA, or even StaleObjectStateException, if you don't mind depending on a Hibernate API. I use the above helper method like this in my code:

file = new TransactionHelper<FileHeaderDTO>().tryExecute(new Callable<FileHeaderDTO>() {
  @Override
  public FileHeaderDTO call() throws Exception {
    return sessionBean.createFile(user.getId(), folder.getId(), name, mimeType, uploadedFile);
  }
});

Java's verbosity makes this not that easy to understand at first sight, but what happens is very simple indeed: I wrap a one-liner call to the session bean method createFile(), in a Callable and supply it to tryExecute(). The result from the bean method is received in a local variable as usual. One thing to keep in mind though is that since Java doesn't have closures, the above anonymous function can only receive final variables as parameters. In my example user, folder, name, mimeType and uploadedFile are all final. You might have to resort to using temporary local variables for storing final values on some occasions.

An introduction to the GSS API

Application Programming Interface or API design is one of my favorite topics in programming, probably because it is both a science and an art. It is a science because there are widely accepted principles about how to design an API, but at the same time applying these principles within the constraints of a given programming language, requires the finesse of an experienced practicioner. Therefore it is with great pleasure that I'll try to explain the ins and outs of the GSS REST-like API, as promised before. As I've already mentioned, GSS is both the name of the source code project in Google Code, as well as the GRNET-sponsored service for the Greek research and academic network (although, it's official name after leaving the beta stage will be Pithos). Since anyone can use the open-source code to setup a GSS service, in this post I'll use generic examples, so anyone writing a client for the GRNET service should modify them accordingly.

When developing an application in a particular programming language, we are used to thinking about the APIs presented to us by the various libraries, which are invariably specified in that same language. For instance, for communicating with an HTTP server from a Java program we might use the HttpClient library API. This library presents a set of Java classes, interfaces and methods for interacting with web servers. These classes hide the underlying complexity of making the low-level HTTP protocol operations, allowing our mental process to remain constantly in a Java world. We could however interact with a web server without such a library, opting to implement the HTTP protocol interactions ourselves instead. Unfortunately, there is no such higher-level library for GSS yet, wrapping the low-level HTTP communications. Therefore this post will present a low-level API, in the sense that one has to make direct HTTP calls in his chosen programming language. The good news is that the following discussion is useful for programmers with any background, since there is support for the ubiquitus HTTP protocol in every modern programming language.

A RESTful API models its entities as resources. Resources are identified by Uniform Resource Identifiers, or URIs. There are four kinds of resources in GSS: files, folders, users & groups. These resources have a number of properties that contain various attributes. The API models these entities and their properties in the JSON format. There is also a fifth entity that is not modeled as a resource, but is important enough to warrant special mention: permissions.

Users · Users are the entities that represent the actual users of the system. They are used to login to the service and separate namespaces of files and folders. User entities have attributes like full name, e-mail, username, authentication token, creation/modification times, groups, etc. The URI of a user with username paul would be:

http://host.domain/gss/rest/paul/

The JSON representation of this user would be something like this:

{
    "name": "Paul Smith",
    "username": "paul",
    "email": "paul@gmail.com",
    "files": "http://hostname.domain/gss/rest/paul/files",
    "trash": "http://hostname.domain/gss/rest/paul/trash",
    "shared": "http://hostname.domain/gss/rest/paul/shared",
    "others": "http://hostname.domain/gss/rest/paul/others",
    "tags": "http://hostname.domain/gss/rest/paul/tags",
    "groups": "http://hostname.domain/gss/rest/paul/groups",
    "creationDate": 1223372769275,
    "modificationDate": 1223372769275,
    "quota": {
        "totalFiles": 7,
        "totalBytes": 429330,
        "bytesRemaining": 10736988910
    }
}

Groups · Groups are entities used to organize users for easier sharing of files and folders among peers. They can be used to facilitate sharing files to multiple users at once. Groups belong to the user who created them and cannot be shared. The URI of a group named work created by the user with username paul would be:

http://host.domain/gss/rest/paul/groups/work

The JSON representation of this group would be something like this:

[
    "http://hostname.domain/gss/rest/paul/groups/work/tom",
    "http://hostname.domain/gss/rest/paul/groups/work/jim",
    "http://hostname.domain/gss/rest/paul/groups/work/mary"
]

Files · Files are the most basic resources in GSS. They represent actual operating system files from the client's computer that have been augmented with extra metadata for storage, retrieval and sharing purposes. Familiar metadata from modern file systems are also maintained in GSS, like file name, creation/modification times, creator, modifier, tags, permissions, etc. Furthermore, files can be versioned in GSS. Updating versioned files retains the previous versions, while updating an unversioned file replaces irrevocably the old file contents. The URI of a file named doc.txt located in the root folder of the user with username paul would be:

http://host.domain/gss/rest/paul/files/doc.txt

The JSON representation of the metadata in this file would be something like this:

{
   "name": "doc.txt",
   "creationDate": 1232449958563,
   "createdBy": "paul",
   "readForAll": true,
   "modifiedBy": "paul",
   "owner": "paul",
   "modificationDate": 1232449944444,
   "deleted": false,
   "versioned": true,
   "version": 1,
   "size": 802,
   "content": "text/plain",
   "uri": "http://hostname.domain/gss/rest/paul/files/doc.txt",
   "folder": {
       "uri": "http://hostname/gss/rest/aaitest@uth.gr/files/",
       "name": "Paul Smith"
   },
   "path": "/",
   "tags": [
       "work",
       "personal"
   ],
   "permissions": [
       {
           "modifyACL": true,
           "write": true,
           "read": true,
           "user": "paul"
       },
       {
           "modifyACL": false,
           "write": true,
           "read": true,
           "group": "work"
       }
   ]
} 

Folders · Folders are resources that are used for grouping files. They represent the file system concept of folders or directories and can be used to mirror a client's computer file system on GSS. Familiar metadata from modern file systems are also maintained in GSS, like folder name, creation/modification times, creator, modifier, permissions, etc. The URI of a folder named documents located in the root folder of the user with username paul would be:

http://host.domain/gss/rest/paul/files/documents

The JSON representation of this folder would be something like this:

{
   "name": "documents",
   "owner": "paul",
   "deleted": false,
   "createdBy": "paul",
   "creationDate": 1223372795825,
   "modifiedBy": "paul",
   "modificationDate": 1223372795825,
   "parent": {
       "uri": "http://hostname.domain/gss/rest/paul/files/",
       "name": "Paul Smith"
   },
   "files": [
       {
           "name": "notes.txt",
           "owner": "paul",
           "creationDate": 1233758218866,
           "deleted":false,
           "size":4567,
           "content": "text/plain",
           "version": 1,
           "uri": "http://hostname.domain/gss/rest/paul/files/documents/notes.txt",
           "folder": {
               "uri": "http://hostname.domain/gss/rest/paul/files/documents/",
               "name": "documents"
           },
           "path": "/documents/"
       }
   ],
   "folders": [],
   "permissions": [
       {
           "modifyACL": true,
           "write": true,
           "read": true,
           "user": "paul"
       },
       {
           "modifyACL": false,
           "write": true,
           "read": true,
           "group": "work"
       }
   ]
}

Working with these resources is accomplished by sending HTTP protocol requests to the resource URI with GET, HEAD, DELETE, POST, PUT methods. GET requests retrieve the resource representation, either the file contents, or the JSON representations for the resources specified above. HEAD requests for files return just the metadata of the file and DELETE requests remove the resource from the system. PUT requests upload files to the system from the client, while POST requests perform various modifications to the resources, like renaming, moving, copying, moving files to the trash, restoring them from the trash, creating folders and more. The operations are numerous and I hope to cover them in more detail in a future post.

One important aspect of every RESTful API is the use of URIs to allow the client to maintain a stateful conversation. For example, fetching the user URI would provide the files URI for fetching the available files and folders. Fetching the files URI would in turn return the URIs for the particular files and folders contained in the root folder (along with other folder properties). Returning to the parent of the current folder would entail following the URI contained in the parent property. This mechanism removes the state handling from the server and puts the burden on the client, providing excellent scalability for the service. Furthermore, since the URIs are treated opaquely by the client, the API allows client reuse across server deployments. A client can target multiple GSS services, as long as they speak the same RESTful API. Moreover, links from service A can refer to resources in service B without a problem (in the same authentication domain, e.g. the same Shibboleth federation). This is the same as using a single web browser to communicate with multiple web servers, by following links among them.

GSS architecture

Handling a large number of concurrent connections requires many servers. Not only because scaling vertically (throwing bigger hardware at the problem) is very costly, but also because even if an application can be designed to scale vertically, the underlying stack probably can not. Java applications for instance, like GSS, run on the JVM and although the latter is an excellent piece of engineering, using huge amounts of heap is not something it's tuned for. Big Iron servers with many cores and 20+ GB of RAM are usually running more than one JVM, since garbage collection is not all that efficient with huge heaps. And since running application instances with a 4-8 GB heap size can be done with cheap off-the-shelf hardware, why spend big bucks on Big Iron?

So having a large number of servers is a sane choice, but brings it's own set of problems. Unless one partitions users to servers (having all requests a particular user makes be delivered to the same server), all servers must have a consistent view of the system data, in order to deliver meaningful results. Assigning user requests to particular servers, usually requires expensive application layer load-balancers or customized application code on each server, so it would rarely be your first option. Having all servers work on the same data is a more tractable problem, since it can be solved by having the application state being replicated among server nodes. Usually, only a small part of the application state needs to be replicated, for each user, that is the part which concerns his current session. But even though session clustering solutions have been a well studied field and implementations abound, having no session to replicate is an even better option.

For GSS we have implemented a stateless architecture for the core server, that should provide us with good scalability in a very cost-effective manner. The most important part in this architecture is the REST-like API that moves part of the responsibility for session state maintenance to the client applications, effectively distributing the system load to more systems than the available server pool. Furthermore, client requests can be authenticated without requiring an SSL/TLS transport layer (even though it can be used if extra privacy is required), which would entail higher load on the application servers or require expensive load balancers. In the server side, API requests are being handled by servlets that enlist the services of stateless session beans, for easier transaction management. Our persistence story so far is JPA with a DBMS backing, plus high-speed SAN storage for the file contents. If or when this becomes a bottleneck, we have various contingency plans, depending on the actual characteristics of the load that will be observed.


The above image depicts the path user requests will travel along, while various nodes interact in order to serve them. A key element in this diagram is the number of different servers that can be found, effectively specializing in their own particular domain. Although the system can be deployed on a single physical server (and regularly is, for development and testing), consisting of a number of standalone sub-services instead of a big monolithic service is a boon to scalability.

This high-level overview of the GSS architecture should help those interested to find their way around the open-source codebase and explain some of the design decisions. But the most interesting part from a user's point of view would be the REST-like API, that allows one to take advantage of the service for scratching his own itch.

So that will be the subject of my next post.

Introducing GSS

During my recent work-induced blog hiatus, I've been working on a new software system, called GSS. I've been more than enjoying the ride so far and since we have released the code as open-source, I thought discussing some of the experience I've gained might be interesting to others as well.

GSS is a network, er grid, er I mean cloud service, for providing access to a file system on a remote storage space. It is the name of both a service (currently in beta) for the Greek research and academic community and the open source software used for it, that can also be used by others for deploying such services. It is similar in some ways to services like iDrive, DropBox and drop.io, but it can also be regarded as a more high-level Amazon S3. Its purpose is to let the desktop computer's file system meet the cloud. The familiar file system metaphors of files and folders are used to store information in a remote storage space, that can be accessed from a variety of user and system interfaces, from any place in the world that has an Internet connection. All usual file manager operations are supported and users can share their files with selected other users or groups, or even make them public. Currently there are four user interfaces available, a web-based application, a desktop client, a WebDAV interface and an iPhone web application, in various stages of development. Underlying these user interfaces is a common REST-like API that can be used to extend the service in new, unanticipated ways.


The main focus of this service was to provide the users of the Greek research and academic community with a free, large storage space that can be used to store, access, backup and share their work, from as many computer systems as they want. Since the available user base is close to half a million (although the expected users of the service are projected to the low ten thousands), we needed a scalable system, that would be able to accommodate high network traffic and a high storage capacity at the same time. A Java Enterprise Edition server coupled with a GWT-based web client and a stateless architecture were our solution. In future posts I will describe the system architecture with all the gory details. The exposed virtual file system features file versioning, trash bin support, access control lists, tagging, full text search and more.

All of these features are presented through an API for third party developers to create scripts, applications or even full blown services that will fulfill their own particular needs or serve other niches. This API has a REST-like design and though it will probably fail a formal RESTful definition, it sports many of the advantages of such architectures:

• system resources such as users, groups, files and folders are represented by URIs

• GET, HEAD, POST, PUT and DELETE methods on resources have the expected semantics

• HTTP caching is explicitly supported via Last-Modified, ETag & If-* headers

• resource formats for everything besides files are simple JSON representations

• only authenticated requests are allowed, except for public resources

Users are authenticated through the GRNET Shibboleth infrastructure. User passwords are never transmitted to the GSS service. Instead GSS-issued authentication tokens are used by both client and server to sign the API requests after the initial user login. SSL transport can provide even stronger privacy guarantees, but it is not required, nor enabled by default.

The GSS code base is GPL-licensed and therefore anyone can use it as a starting point to implement his own file storage service. We have yet to provide binary downloads, due to the various dependencies, but the build instructions should be enough to get someone started. We are always interested in source or documentation patches, of course (did I mention it's open source?). Most importantly, the REST API will ensure that clients developed for one such service can be reused for every other one.

I will have much more to say about the API in a future post. In the meantime you can peruse the code and documentation, or even try it out yourself. I'd be very interested in any comments you might have.