Monthly Archive: December 2015

The Object rEvolution


It’s our pleasure to host Ryan Peterson, Chief Solution Strategist at EMC, as a guest blogger to expand upon another great step in our partnership to deliver compelling customer solutions through joint engineering efforts. Follow Ryan @BigDataRyan.

Object storage isn’t a new concept and EMC’s been innovating around it since the beginning. Take our Centera and Atmos products as key examples. The first Centera was created around the idea that objects could store much higher quantities of data than a file system in a single store while the other aspect of Centera was a rich set of security and compliancy features file systems had not been able to achieve. Data shredding for example was a feature required by governments and law firms. We all know some politicians who need a Centera system Atmos on the other hand was designed with a completely different base requirement. The goal was to support a geo-parity environment mostly seen in large enterprise customers and with service providers. In the Atmos design, data written to one location would be protected by other locations and yet share a common namespace. The design inspired many large internet-scale companies you likely use today and some of them are even backed by an Atmos system.

But when you are innovating from scratch, you make design decisions that leave things out and you learn from the 25,000 current EMC object storage customers. So we started with a new baseline of code and added in many of the components of Centera and Atmos to create something new, exciting, and dare I say revolutionary. Enter Elastic Cloud Storage (ECS) which can scale from one rack in one data center to many racks in many data centers thus encompassing the design requirements of both Atmos and Centera and with new data protection features that increase performance from the original design such as local replica process, erasure coding for high performance, and geo-protection using XOR to reduce overhead. ECS changes the game!

But with the advent of new technologies in the world such as is near and dear to my heart, Hadoop, the design needed to include the capability to analyze the data in the entire global namespace and do it efficiently.

ECS includes a mapping to Hadoop using the Hadoop Compatible File System (HCFS) guidelines the same way you might see a Lustre or Gluster connect. The metadata controllers in ECS provide the namespace context and allow Hadoop to be able to see the data on that system the same way it would if it were looking at HDFS. In fact, it’s as simple as using a different URI string to connect and you don’t have to remove your HDFS DAS if you don’t want to. Simply take your existing Hadoop cluster and point to viprfs:// like shown below. Hadoop will automatically open a series of connections to access the data at the fastest possible rate.

Now before we wanted to go out and tell the world about this solution, we really wanted to enlist the support of the Hadoop distributions and we wanted to test it thoroughly. The picture below is a setup of 10 racks of ECS running the Hortonworks and Pivotal distributions of Hadoop. This is one of others like it that seek to simplify the implementation process, validate all things are functional, and provides us a place to test scenarios our customers bring to us.

Our friends at Hortonworks really did an amazing job going through all of the features of Hadoop and validating each and every line of Apache code works on ECS. Click here to see all of the certifications that have already been completed with our geo-scale object platform and Hortonworks.

So what? What does this mean to you? Let’s get serious and clear. Never before has there been an opportunity to purchase your own Analytics-Ready-Cloud-in-a-Box. So who are the customers that might care?

If you have a need for data to be spread across geographies such as Americas, Europe, and Asia; or even New York, Chicago, and Los Angeles, then relying on a single name space to support that environment while keeping the data in a state that can be quickly accessed and analyzed should be top of mind. Thus far, we’ve seen customers in the following segments (to name a few and not exhaustive):

Internet of Things (IoT) such as Connected Cars, Home Automation, Turbines, and Smartphone Backups
Geo-scale Archive – data that you might have sent to tape or offsited stays inexpensive and analytics accessible
Service Providers, Telcos, and Web 2.0 companies that need to service the application generation

Let’s compare this with the existing technologies used in Public Cloud providers not using ECS. Data is collected in multi-tenant object systems, is copied to another platform for analysis (a Cloud Data Lake so to speak) and the results pushed back into your primary system. Amazon’s S3 and EMR are a good example of that type of legacy cloud architecture. With ECS, we remove the need to move data by allowing analysis to happen against the data set where it sits. Now that’s Revolutionary!

If you have requirements that you believe are met with ECS, whether you want to host the equipment yourself or are looking for an ECS-enabled Public Cloud Service Provider, reach out to your EMC representative or discuss with our friends at Hortonworks. We can meet your needs with this rEvolutionary architecture.

For more information, you can watch this video of my colleagues Nikhil & Priya discussing the internals of the platform and how it works with Hadoop.

You can also download our Hadoop on ECS White Paper.

The post The Object rEvolution appeared first on Hortonworks.

…read more

Hadoop on Remote Storage


The question regarding running Hadoop on a remote storage rises again and again by many independent developers, enterprise users and vendors. And there are still many discussions in community, with completely opposite opinions. I’d like to state here my personal view on this complex problem.

In this article I would call remote storage “NAS” for simplicity. I would also take as a given that remote storage is not the same HDFS, but something completely different – from standard storage arrays with LUNs mounted to the servers to different distributed storage systems. For all these systems I assume that they are remote, because unlike HDFS they don’t allow you to run your custom code on the storage nodes. And they are mostly “storages”, so they are using some kind of erasure encoding to save the space and make this solution more competitive.

If you are reading my blog for a long time, you might mention that it is the second version of this article. During the last year I was constantly thinking on this problem, and my position has shifted a bit, mostly based on the real world practice and experience.

Read IO Performance. For most of the Hadoop clusters the limiting factor in performance is IO. The more IO bandwidth you have, the faster your cluster would work. You won’t be surprised if I tell you that the IO bandwidth mostly depends on the amount of disks you have and their type. For example, a single SATA HDD can deliver you somewhat 50MB/sec in sequential scans, SAS HDD can give you 90MB/sec and SSD might achieve 300MB/sec. This is a simple math to calculate the total platform bandwidth given these numbers. Comparing DAS with NAS does not make much sense in this context, because both NAS and cluster with DAS might have the same amount of disks and thus would deliver comparable bandwidth. So again, considering infinite network bandwidth with zero latency, same RAID controllers and same number and type of drives used, DAS and NAS solutions would deliver the same read IO performance.
Write IO Performance. Here the things are getting a bit more complicated, and you should understand how exactly your NAS solution work to be able to compare it with Hadoop on DAS. HDFS stores a number of exact copies of the data, 3 by default. So if you write X GB of data, in fact they would occupy 3*X GB of disk space. And of course, the process of writing 3 copies of the data is 3 times slower than the process of writing a single copy. How does the most NAS storages work? NAS is an old industry and they clearly understood that storing many exact copies of the data is very wasteful, so most of them use some kind of erasure coding (like Reed-Solomon one). This allows you to achieve similar redundancy with storing 3 exact copies of the data with only 40% overhead with RS(10,4). But everything comes at cost, and the cost here is performance. For writing a single block in HDFS you have to just write it 3 times. With RD(10,4) to write a single block you have to calculate erasure codes for it either by reading other 9 blocks and writing out 4 of them, or having some kind of a caching layer with replication and background compaction process. In short, writing to it would always be slower than writing to the cluster with replication, this is like comparing RAID10 with RAID5, same logic of replication vs erasure coding.
Read IO performance (degraded). In case you have lost a single machine or single drive in Hadoop cluster with DAS, your read performance is not affected – you read the same data from a different node that is still alive. But what happens in NAS with RS(10,4)? Right, to restore a single block with RS(10,4) you have to read up to 13 blocks, which would make your system up to 13 times slower! Of course, in most cases you encode sequential blocks and then read sequential blocks, so you can restore the missing one easier. But still, your performance would degrade 2x in best scenario and up to 13x in worst:

And if you think that the degraded case is not very relevant for you, here is the statistics of Facebook Hadoop cluster:

Data Recovery. When you are losing the node and repliacing it, how long does it take to recover the redundancy of your system? For HDFS with DAS you are just copying the data for under-replicated blocks to the new node. For RS(10,4) you have to restore the missing blocks by reading all the other blocks in its group and performing computations on top of them. Usually it is 5x-10x slower:

Network. When you run a Hadoop cluster with DAS, Hadoop framework itself tries to schedule executers as close to the data as possible, usually making a preference to local IO. In the cluster with NAS, your IO is always remote, with no exceptions. So the network becomes a big pain point – you should plan it very carefully with no oversubscription both between the compute nodes, and between compute and storage. Network rarely becomes a bottleneck if you have enough 10GbE interfaces, but the switches should be good, and you need much more of them than in solution with DAS. Here’s the slide from Cisco’s presentation regarding this subject:

Local Storage. Having remote HDFS might look like a good option, but what about the local storage on the “compute” nodes? Usually people forget that the same MapReduce stores intermediate data on the local storage, and the same Spark puts all the shuffle intermediate data to the local storage. Plus the same Hive and Pig are translated into MR or Tez or Spark, storing their intermediate results on local storage as well. Thus even “compute” nodes should have enough local storage, and the safest option is to have the same amount of raw …read more