Hadoop HDFS Reviews & Product Details

it is effective for long time storage, and easily scalable. Review collected by and hosted on G2.com.

Writing to HDFS system is a little bit slow. Review collected by and hosted on G2.com.

HDFS supports features such as partitioning and replication that are actually mandatory to be present in a distributed environment. Of course, many optimizations should be done over the next years but the main concept will be always the same. Move code into the data and keep your data safe with no risk depending on the failures. What I like best in HDFS is the user interface which is pretty similar to a common local Linux filesystem. Moreover, HDFS is very compatible and that can be integrated with the majority of the frameworks that are used today like Hadoop and Spark. Last but not least, HDFS is open source and a huge community supports it. Review collected by and hosted on G2.com.

HDFS has some disadvantages as well. First of all, I would really like it to be more customizable and to provide more features to the user interface. By doing this, users will be free to play and experiment with new ideas which will be integrated with HDFS. I also have observed that someone has to be an expert in order to use it securely in his application and there is no much documentation about how to achieve this specifically in HDFS. Review collected by and hosted on G2.com.

Distributed and fault tolerant, transparent. Mimics Unix FS features which are familiar to many users. Good fit for tech savvy users (config files, command line interface). Fast enough. Review collected by and hosted on G2.com.

Not as fast as your local FS. More limited in tools and features than your familiar Unix environment. Bad fit for non-tech savvy users (config files, command line interface). Not entirely POSIX compliant, but gains in performance because of that. Review collected by and hosted on G2.com.

Distributed file storage made easy with using HDFS. I don't need to know where the files are stored physically in the server because HDFS exposed all the files as if it was a single storage with multiple backup (depending on your replication factor). In terms of using HDFS API, it is straightforward to use. Review collected by and hosted on G2.com.

Configuration. To get HDFS running might be easy or complicated depending on your experience. We are using Hadoop together with Cloudera, so that was really easy for us to get things started. However, as any other Hadoop components, fine-tuning HDFS can be tricky. Debugging HDFS can also be tricky, like suddenly HDFS doesn't allow write due to it being in safe mode. At the point I was using Hadoop, getting Hadoop to work with HA is also challenging, namenode was a single point of failure. HDFS also doesn't work well with lots of small files. For the average user, it can be daunting for them to access HDFS (I think HDFS has a web app running with limited functionality), for developers it would be no issue. Review collected by and hosted on G2.com.

It is a good way to make big storages and for distributed system. Review collected by and hosted on G2.com.

I don't know much about Hadoop HDFS. But if I have to answer the question then, it is hard to use whenever I wanted to control some files or directories. Commands are not comfortable for me. I wanted to use like Linux command. This is a little bit different with Linux's, I think. Review collected by and hosted on G2.com.

Hadoop is a collection of software that handles distributed file system (HDFS), and distributed processing mechanism on top of it (MapReduce). It is highly scalable and reliable. With Hadoop, users could specify their processing requirements on large datasets without worrying the details of underlying communication and data distribution. Hadoop can scale up easily to adapt to workflow increase. Automatic data replication mechanism in HDFS guarantees its reliability. Review collected by and hosted on G2.com.

Hadoop is written in Java and it is not fast. It cannot handle the data processing requests in real-time. Its processing layer, MapReduce, simplifies the processing logic by supporting only a Map and Reduce function, but it also introduces inconvenience to express complicated processing logic.

Hadoop adopts master-slave architecture, but the master is designed in single-node mode: when the master node is down, it is difficult to get recovered. Users have to purchase high-end hardware to prevent master-node failures. Review collected by and hosted on G2.com.

hdfs is high available and scalable, I can expand the storage only by adding several datanodes. And with hdfs, genome data can be easily analyzed by mapreduce. Review collected by and hosted on G2.com.

hdfs is not so good for small files, and the nfs-gate-way is also not very well. Review collected by and hosted on G2.com.

Well I am using Hadoop HDFS for HBase filesystems. I found it's really easy to deploy. I use Cloudera Manager as hadoop package, it could be more easy. If you have a lot of nodes, then truly you will have power from Hadoop HDFS Review collected by and hosted on G2.com.

It's quite troublesome for tuning HBase and HDFS. At first when we have few nodes it doesn't look better, but when we hit more nodes, performance gained. But still, a lot of tinkering to do. Review collected by and hosted on G2.com.

Hadoop is a no brainer for big data. The main killer feature is HDFS. Having a redundant WORM file system is amazingly useful. There's a reason Google invented it and Yahoo made it open source. 3 copies of your file just works. Cheap commodity servers, but still fast and stable. Never lose data, store everything. Access and use in multiple use cases. So many tools and other projects around Hadoop make it a must-have for all enterprises and startups. You add Spark which most distributions include and you can pretty much do everything you need. Ambari and Hue make it easy to set up now. Review collected by and hosted on G2.com.

There's a lot of stuff in Hadoop, also there's always 10 ways to do something and hard to know what's the best. Do you do Storm or one of 20 other frameworks. Should I store in Parquet, ORCFile, Avro or CSV or something else? Do you compress with SNAPPY or nothing. What level of encryption? Is Kerberos good enough for my security. Security is a bit lax and there's definitely a lot of things to configure. Review collected by and hosted on G2.com.

HDFS is reliable and solid, and in my experience with it there are very few problems using it. If you have your own data centre and you use Hadoop, it's the obvious choice for reliably storing your data. Review collected by and hosted on G2.com.

If your NameNodes all go down, then HDFS is pretty much useless as you won't know which file blocks are where and which files they belong to -- and I've read it's difficult to recover (or impossible) if you completely lose your NameNode file mappings. Fortunately I've never personally seen this occur. Review collected by and hosted on G2.com.