The challenge of scaling up and scaling out

Additional computing capacity has been achieved by increasing the speed processors operate.

However, the speed of light became a limiting factor for a single processing unit, a single CPU, to handle its internal electrical connections. Marginal improvements in speed certainly will be available but order of magnitude improvements from speed increases are unlikely to materialize.

To increase computing capacity, hardware vendors began using additional processing units, CPUs, per chip by adding the equivalent of additional traffic lanes in a highway; same speed limit but offering more lanes and hence able to handle more traffic, more cars moving per unit of time.

We have now multi-CPU processing units of 2, 4, 8, 16, 32, 64 and more CPUs and soon likely will reach hundreds and probably in few years will count with thousands of CPUs in one computer unit.

The software to tap such capacity has not evolved as fast as the hardware. It is a serious challenge to develop applications that can operate correctly while running concurrently on more than one processing unit and a challenge also to use same or similar methods to distribute work among a network of computing nodes.

While there are proprietary and open source products to address software operation on clusters of computers, such as the widely used Message Passing Interface, MPI, in supercomputing applications, conventional programming languages lack tools and features to facilitate development of applications designed to operate correctly and efficiently by scaling within a computer unit, scaling up, and scaling out by distributing work among participating computing nodes.

The programming languages listed below receive attention these days partly because of corresponding claims of design, features and functions related to operating on multiprocessor systems and related constructs for concurrency and scalability.

  1. Scala. Excerpt from Scala's home page: "Scala is a general purpose programming language designed to express common programming patterns in a concise, elegant, and type-safe way. It smoothly integrates features of object-oriented and functional languages, enabling Java and other programmers to be more productive."
  2. Clojure. Excerpt from Clojure's home page: "Clojure is a dynamic programming language that targets the Java Virtual Machine (and the CLR ). It is designed to be a general-purpose language, combining the approachability and interactive development of a scripting language with an efficient and robust infrastructure for multi-threaded programming."
  3. Erlang. Excerpt from Erlang's home page: "Erlang is a programming language used to build massively scalable soft real-time systems with requirements on high availability. Some of its uses are in telecoms, banking, e-commerce, computer telephony and instant messaging. Erlang's run-time system has built-in support for concurrency, distribution and fault tolerance."
A recent development in the Scala community is the introduction of the Akka library that offers concurrency models and APIs for applications to operate correctly in a multi-CPU environment, but offers also APIs to go beyond one computing node and distribute work among participating nodes in a network. Akka's concurrency and distributed models and software constructs look and feel as it were a port of Erlang to Scala.
We believe that writing correct concurrent, fault-tolerant and scalable applications is too hard. Most of the time it's because we are using the wrong tools and the wrong level of abstraction.

Akka is there to change that. Using the Actor Model together with Software Transactional Memory we raise the abstraction level and provide a better platform to build correct concurrent and scalable applications.

Actors also provides the abstraction for transparent distribution and the basis for truly scalable and fault-tolerant applications.
A further development just announced this week is the partnership of the creator of Scala, Martin Odersky, and the creator of Akka, Jonas Bonér, to form the company named Typesafe offering an open source integration of Scala and Akka.

The software integration represented by Typesafe is good news to Scala and Java developers by offering an integrated set of tools, documentation, training and consulting in the area of applying concurrency constructs to build correct and efficient applications in muti-CPU and distributed computing environments.

The Akka library offers the Scala and Java APIs. Using the library one is certainly pleased to see the strong influence of Erlang.

Another Scala and Java library addressing the ability to scale up and out is GridGain. Here are excerpts re library from GridGain's home page.

GridGain 3.0 is the industry JVM-based post-functional distributed middleware that combined traditional object-oriented programming approach with comprehensive functional programming support.

GridGain is a High Performance Cloud Computing software that enables development of high performance Java and Scala applications that scale up on any grid or cloud infrastructure.

GridGain provides state of the art implementations for compute grids, data grids, and zero deployment. You can develop with GridGain using Java or Scala programming languages - both of which are supported natively.

The result - is a powerful, flexible and highly expressive APIs that are simple and productive to use.
Also this week we saw the introduction of Google's Go programming language at Google's I/O conference in San Francisco.

Google presents Go as designed specifically for concurrency and distribution and included as a third language alternative to Google's App Engine service: Python, Java and now, Go.

It appears that the software world is gradually catching up to benefit from the availability of tens, hundreds and even thousands of parallel computing lanes in one node and scale out by distributing work among participating computing nodes in a network.

No comments: