Big Data needs Big Storage and storage is at the end a physical device. Until now, most storage devices are hard disks that require mechanical movements. In this tutorial we will discuss the storage performance challenges.

What are the storage performance challenges?

A common hard drive available today (December 2012) has 15,000 (Seagate, 2013) revolutions per minute (rpm) and a desktop hard drive has some 7,200-rpm. In any case, this means that there is significant latency involved until the reading head is in place. The mechanical approach to storage has been around for decades and scientists as well as engineers complain about storage performance.

In-memory was always faster than hard disk storage and the network speed is higher than what can be done with hard disks. (Anthes, 2012) states that disk based storage is about 10-100 times slower than a network and about 1,000 times slower than main memory. This means that there is a significant “bottleneck” when it comes to delivering data from a disk-based storage to an application.

As big data is about storing and analyzing data, this is a major challenge to Big Data Applications. It doesn’t help us much if we have enough compute power to analyze data but our disks simply can’t deliver the data in a fast way.

Data is distributed

When we look at supercomputers nowadays, they are often measured in cores and Teraflops (Top 500 Supercomputers Site, 2012). This is basically good if you want to do whatever kind of calculation such as the human genome. However, this doesn’t tell us anything about disk performance if we want to store or analyze data. (Zverina, 2011) cites Allan Snavely when he proposes to include the disk performance in such metrics as well:

“I’d like to propose that we routinely compare machines using the metric of data motion capacity, or their ability to move data quickly” – Allan Snavely

Allan Snavely also stated that with increasing data size – hard disks are getting higher in capacity but access time stays the same – it is harder to find data.

This can be illustrated easily: you have an external hard disk with the capacity of 1 TB. The hard disk operates with 7,200 rpm and a cache of 16MB. There are 1,000 Videos stored on this hard drive, each with a size of 1 GB. This would fill the entire hard disk. If you now change to a larger system as your videos grow, you would change to a 2 TB system.

If this System is full, you won’t be able to transfer the videos to another system in the same time as you did with the 1 TB hard drive. It is very likely that your 2 TB System now needs about twice as much time to transfer the data. Whereas compute performance grows, the performance to access data stays about the same. Given the growth of data and storage capacity, it even gets slower. Allan Snavely (Zverina, 2011) describes this with the following statement:

“The number of cycles for computers to access data is getting longer – in fact disks are getting slower all the time as their capacity goes up but access times stay the same. It now takes twice as long to examine a disk every year, or put another way, this doubling of capacity halves the accessibility to any random data on a given media.”

How to overcome these challenges?

In the same article, Snavely suggests to include the following metrics in a computer’s performance: DRAM, flash memory, and disk capacity.

But what can enterprises do to achieve higher through output of their systems? There is already some research about that and most resources point towards Solid State Disks as Storage (SSD). Solid State Disks are getting commodity hardware in high end Personal Computers, but they are not that common for servers and distributed systems yet.

SSDs normally have better performance but lower disk space and the price per GB is more expensive. If we talk about large-scale databases that have the need for performance, SSDs might be a better choice. The San Diego Supercomputing Center (SDSC) built a supercomputer with SSDs. This computer is called “Gordon” and can handle Data up to 100 times faster as with normal drives (Zverina, 2011).

Another prototype, called “Moneta” (Anthes, 2012) used a phase change memory to boost I/O performance. The performance was about 9.5 times faster as a normal RAID-System and about 2.8 times faster as a flash-based raid system.

There is significant research around this topic as the performance of storage is a problem to large-scale data centric systems as we now have with Big Data Applications.

I hope you enjoyed the first part of this tutorial about big data technology. This tutorial is part of the Big Data Tutorial. Make sure to read the entire tutorials.

One of the most important things is to partition data in an environment. Especially with large-scale systems, this is very important, as not everything can be stored on a limited number of systems.

How to partition data?

Partitioning is another factor for Big Data Applications. It is one of the factors of the CAP-Theorem (see 1.6.1) and is also important for scaling applications. Partitioning basically describes the ability to distribute a database over different servers. In Big Data Applications, it is often not possible to store everything on one (Josuttis, 2011)

Data Partitioning

Data Partitioning

The factors for partitioning illustrated in the Figure: Partitioning are described by (Rys, 2011). Functional partitioning is basically describing the service oriented architecture (SOA) approach (Josuttis, 2011). With SOA, different functions are provided by their own services. If we talk about a Web shop such as Amazon, there are a lot of different services involved. Some Services handle the Order Workflow; other Services handle the search and so on.
If there is high load on a specific service such as the shopping cart, new instances can be added on demand. This reduces the risk of an outage that would lead to loosing money. Building a service-oriented architecture simply doesn’t solve all problems for partitioning. Therefore, data also has to be partitioned. By data partitioning, all data is distributed over different servers. They can also be distributed geographically.
A partition key basically identifies partitioned Data. Since there is a lot of data available and single nodes may fail, it is necessary to partition data in the network. This means that data should be replicated and stored redundant in order to deal with node failures.

I hope you enjoyed the first part of this tutorial about big data technology. This tutorial is part of the Big Data Tutorial. Make sure to read the entire tutorials.

Scalable data processing is necessary for all platforms handling data. In today’s tutorial we will have a look at this.
Scalability is another factor of Big Data Applications described by (Rys, 2011). Whenever we talk about Big Data, it mainly involves high-scaling systems. Each Big Data Application should be built in a way that eases scaling. (Rys, 2011) describes several needs for scaling: user load scalability, data load scalability, computational scalability and scale agility.

What is scalable data?

The figure illustrates the different needs for scalability in Big Data environments as described by (Rys, 2011). Many applications such as Facebook (Fowler, 2012) have a lot of users. Applications should support the large user base and should stay prone to errors in case the application sees unexpected high user numbers. Various techniques can be applied to support different needs such as fast data access. A factor that often – but not only – comes with a high number of users is the data load.
(Rys, 2011) describes that some or many users can produce this data. However, things such as sensors and other devices that do not directly relate to users can also produce large datasets. Computational scalability is the ability to scale to large datasets. Data is often analyzed and this needs compute power on the analysis side. Distributed algorithms such as Map/Reduce require a lot of nodes in order to perform queries and analyze in a performing manner.
Scale agility describes the possibility to change the environment of a system. This basically means that new instances such as compute can be added or removed on-demand. This requires a high level of automation and virtualization and is very similar to what can be done in cloud computing environments. Several Platforms such as Amazon EC2, Windows Azure, OpenStack, Eucalyptus and others enable this level of self-service that is a great support to scaling agility for Big Data environments.

I hope you enjoyed the first part of this tutorial about big data technology. This tutorial is part of the Big Data Tutorial. Make sure to read the entire tutorials.

Agility is an important factor to Big Data Applications. Agile data needs to fulfill 3 different agility factors which are: model agility, operational agility and programming ability. (Rys, 2011)

Data agility

Data agility

Agile data: model agility

Model agility means how easy it is to change the Data Model. Traditionally, in SQL Systems it is rather hard to change a schema. Other Systems such as non-relational Databases allow easy change to the Database. If we look at Key/Value Storages such as DynamoDB (Amazon Web Services, 2013), the change to a Model is very easy. Databases in fast changing systems such as Social Media Applications, Online Shops and other require model agility. Updates to such systems occur frequently, often weekly to daily (Paul, 2012).

Operational agility

In distributed environments, it is often necessary to change operational aspects of a System. New Servers get added often, also with different aspects such as Operating System and Hardware. Database systems should stay tolerant to operational changes, as this is a crucial factor to growth.

Programming agility

Database Systems should support the software developers. This is when programming agility comes into play. Programming agility describes the approach that the Database and all associated SDK’s should easy the live of a developer that is working with the Database itself. Furthermore, it should also support fast development.

I hope you enjoyed the first part of this tutorial about big data technology. This tutorial is part of the Big Data Tutorial. Make sure to read the entire tutorials.