• With XAP's on-demand scalability we can just add more and more commodity servers gradually to meet [required] growth.”
• Cees de Groot, Lead Architect, eBay subsidiary Marktplaats

In this page: The Challenge | GigaSpaces XAP Delivers | How GigaSpaces XAP Does It | GigaSpaces XAP Track Record | Definitions and Assumptions - Hardware Savings

The Challenge

Many IT organizations are finding it challenging to support unexpected business growth and sudden peaks in demand. Software applications often need to scale up quickly to meet dramatically higher traffic or operation volumes. As soon as the capacity of a single machine is exhausted and the application needs to “scale out” across additional machines, scalability becomes difficult and expensive. In the best case, there are diminishing returns from each additional machine, until at some point adding more machines actually decreases performance. In the worst case, applications need to undergo a re-architecture and re-write of key components, which can take months and cost millions.

Read more about the hardware cost of scaling.

GigaSpaces XAP Delivers

With GigaSpaces eXtreme Application Platform (XAP), you’ll need to purchase much less hardware in order to scale up. The following table shows how much hardware you can expect to save. It depends on how far you want to scale, because the more you scale, the more hardware you’ll have to buy for each incremental increase in throughput. It also depends on the contention – how congested your application architecture is, the severity of the bottlenecks, and as a result how difficult it is to scale.

(*) For these combinations of scaling factor and contention, scalability is theoretically impossible. Therefore, if this is the contention level you are starting off with, it is not meaningful to provide a savings figure. Nevertheless, GigaSpaces XAP has proven experience scaling applications to these levels.

These figures were computed using the most broadly-accepted model of software scalability – see our definitions and assumptions below.

Additional Savings

Want more details about the financial benefits to your organization?

Let us know and a GigaSpaces expert will be in touch to help you quantify the benefits of inexpensive scalability.

How GigaSpaces XAP Does It

• Single Platform – simplifies your architecture for inherent scalability
• Most enterprise applications rely on numerous pieces of middleware. GigaSpaces XAP unites all these different parts into one platform, providing all the middleware functions in-memory, in-process, including data storage and management, messaging, and processing for software services. GigaSpaces XAP makes your architecture simple and inherently scalable, allowing you to scale up on any hardware, and ensuring that your application behaves exactly the same at any scale. This saves most of the difficulty and cost associated with scalability today.
• Scalable On-Demand – unlimited, linear scalability providing a low, predictable cost of scaling
• Traditional software architecture has centralized components, primarily the database, which quickly become overloaded when the application grows. GigaSpaces XAP solves this problem by packaging the entire application, from the load balancer on the front end to the database on the back end, into a lightweight scalable unit, running entirely in-memory. Simply duplicating this unit on additional machines gives you unlimited, linear scalability. XAP distributes the data between these in-memory units, releasing the database bottleneck. XAP also balances the available resources on-the-fly, by automatically deploying the scalable units based on custom policies and service levels you define

 Learn more about GigaSpaces XAP

GigaSpaces XAP Track Record

GigaSpaces XAP has a long history of achieving scalability for the most demanding applications and the most extreme scenarios. Here are just a few of our war stories and benchmarks.

Customer Stories

• iPhone activation platform at a major UK mobile operator – the operator was planning an extremely hyped launch of iPhone in Britain, and expected an unknown peak number of activations in the first hours of the launch. XAP was initially tested for several thousand activations per second, but on the day of the launch, the actual peak load was five times the expected level! Nevertheless, XAP met the unexpected load and did not drop a single activation. Even more impressively, XAP acted as a buffer, protecting the back-end systems from the high loads, and ensuring their high availability as well.
• Transaction Lifecycle Management solution used by 75 of the world's top 100 banks – before GigaSpaces, the solution was limited to one physical machine. Even with an expensive high-end machine, customers were limited to a maximum volume of trades they could process and validate at the end of the trading day. Today the same maximum volume is processed 500% faster, on four commodity Intel boxes. To go beyond this volume, customers can simply add more machines – they can use eight commodity boxes to process 2X of the previous maximum limit; or even 80 boxes to process 20X.

All customers

Benchmarks

• Web application scalability – conducted on the classic eCommerce application example, Pet Clinic, running on commodity hardware and MySQL. GigaSpaces XAP scaled linearly up to 1.4 billion page views per day, with latency of only 6 ms.
• JEE scalability – a reference application running on a traditional JEE architecture was compared to an identical application running on XAP. XAP provided a 500% increase in throughput with 1/10 the latency on the same hardware.
• Scaling linearly on multi-core – the test was conducted on a Sun T5240 Sparc and Sun Fire x4450-Intel machine. XAP exhibited linear scalability up to 1,400,000 operations per second on 60 threads.

All benchmarks

Definitions and Assumptions – Hardware Savings

Here are the basic definitions and assumptions underlying our figures at the top of this page:

•  “Contention” refers to the fraction of time the application spends waiting for a centralized resource. 10% is extremely low, and is usually associated with standalone event-driven applications with no high availability. On the upper end of the scale, 40-50% contention is usually experienced by applications with a central database, integration with back-end systems such as ERP, ESBs, etc.
• Results are based on a broadly-accepted model of software scalability known as Amdahl's Law, which provides a theoretical maximum for scalability in the presence of contention, and a theoretical minimum for the number of machines required to scale.
• These are absolute minimum results – according to the predictions of Amdahl's Law, you will definitely save more, given these levels of contention. Keep in mind that you can only make a rough estimate of your actual software contention.
• These results apply to any size of application – even if you have only 2 machines, GigaSpaces XAP will deliver the same hardware savings when you scale up. Even if you need only one more machine to scale up, we can save some of this machine's capacity, allowing it to be used for other projects.
• The reason for the very high savings figures in extreme scaling scenarios is that due to non-linear scalability, you’d need to add as much as 10-20 times the number of existing machines just to achieve a scaling ratio of 2X, 3X or 4X. In these extreme cases, XAP makes a very big difference.
• Assumptions about usage of XAP – these figures are only valid if the application uses XAP as its application server, messaging server, and In-Memory Data Grid, with no use of aggregated queries on all data.

If you’re interested, see a detailed description of our computation.