Now Scale Up is Out and Scale Out is In

Saturday, August 14, 2010

Rahul Neel Mani


Being at the helm of product development for nearly two decades in the company, Roland Slee, Vice President, Oracle Database Product Management Team, in conversation with Rahul Neel Mani, thinks it is about time the industry reconsiders the way it computes....

Q:Do you think the computing will be done as the industry has predicted or will it change as time passes?

A: I see three primary drivers in the way the world computes. In the current scenario, IT has not yet delivered economies of scale. When there are economies of scale there is consolidation and centralisation.

But even today corporates including telcos, banks, manufacturing organisations etc. use hundreds of application silos, which are fragmented and distributed simply because everything is driven by money.

If you build a larger, critical system on which your business depends, the price of that growing system outpaces its capacity. From both price and risk points of view, it makes sense to have ten small systems rather than one large system.

This has to change. To help IT truly mature, we have to change the architecture in a way that enterprises can deploy one big system that is cheaper, less risky and requires little maintenance. It hasn’t happened yet. We are at the tipping point.

Second important thing is a single, unified technical architecture. If you look at the data centre of any large corporation, you will find systems fragmented not only for economic reasons but also for technical reasons.

In every business you need to work with at least three styles of computing – transact, analyse and collaborate. Traditionally, both at software and hardware level, these different styles of computing have different architectures. OLTP systems, data mart/data warehousing systems and communication and collaboration systems in an organisation are all architected differently.

However, the way business process flows, one would ideally want to capture transactions, understand their implications and collaborate with others to take action on the insight – in a continuous flow, within one process. In today’s world, the technical environment wouldn’t allow you to put these things together.

Today, corporations spend most of their time working in three different silos – moving the data from one process to another. This has to change to bring the processes closer to the data and not the other way round.

Thirdly, we need to bridge the gap between the real and virtual worlds. Computer systems are considered to be virtual reality for businesses, but it has a lot of gaps.

Typically today, when you look at the warehouse management system, it shows a specific product lying at a specific location but which actually isn’t there. Ideally, the system should be the virtual reality of the real world. Normally, the system is wrong. The strategic agenda for IT is to make this virtual reality correct.

What does IT? IT is transactions processing. Transactions record business events. In case of a warehouse, the product moves out of the shelf and if the systems record that, it should be reflected. Today, most companies spend their time updating systems to keep pace with reality.

We need to move to a world where automatic sensing, location-based services and ubiquitous network access are put together which result into self maintaining of the systems. This will result in continuous monitoring of reality without human beings required to manually key-in the transactions. This will free a lot of time from transaction processing.

Q:What is Oracle doing to help CIOs achieve economies of scale without spending huge money on IT?

A: First, let’s talk about how to create economies of scale. The fundamental reason why traditional IT architectures are not able to deliver economies of scale is because of the inherent incapability of the systems in doing so.

If you ask any traditional system vendor how to build a scalable system, the answer would be to buy a bigger computer. Bigger computers become exponentially expensive. You get reverse economies of scale. The bigger a system you create, the more internal friction there is, which leads to lesser efficiency.

This is precisely why the price rises exponentially. Hypothetically even if you can buy the biggest computer which has about 128 CPUs, do you have an estimate how much will it cost to add a 129th CPU? The answer is - infinitely expensive. The crux is that there is a ceiling on scalability, when you are trying to build a scalable environment through bigger, powerful computers.

Secondly, there is an inherent risk in doing this. When you are trying to consolidate workloads into large systems, instead of having thousands of users to one system, you are perhaps exposing hundreds of thousands of users to one system. If that system goes down, your entire global enterprise comes to a grinding halt.

You cannot afford that sort of downtime. And the cost of avoiding downtime also rises exponentially.

Thus the solution here is to move from ‘scale up’ architecture to ‘scale-out’ architecture. Instead of putting clever hardware and software to create scalable and continuously available architecture, CIOs should look at solving it through databases and middleware.

This is what Oracle has been doing for 30 years. We make it possible to bring together a very large number of standardised, lowcost computers allowing them to work in concert in a grid or in today’s fashionable language – in a cloud. We may be slow in adopting the language of cloud computing, but from an engineering perspective, we are ten years ahead of the competition.

Oracle is perhaps the only company which can manage information in the cloud in a way which is invisible to the applications. Oracles journey from scale-up to scale-out and from ‘big boxes’ to the ‘cloud’ has been invisible to applications. This is the magic. Oracle Exadata, the new database machine is the clearest indication of the strategic importance that we are talking about. Exadata is actually a private cloud.

Inside a full rack Exadata system there are 22 servers – 14 storage servers and 8 database servers. You can scale it out up to eight full racks just plugging in cables to network. The architecture never changes; it is identical for any scenario. This is our answer to bring economies of scale to IT for the first time.

Q:What about the unified computing?

A: For the last 30 years, Oracle has been working to create better middleware and database technologies which can support transaction processing, decision support and data warehousing, and content management and collaboration on a completely unified, standardised technology stack. We recently released Oracle Enterprise Content Management (ECM) version 11G.

The tech stack for this is: Oracle fusion middleware running on Java; running against an Oracle Exadata database machine. Similarly, we have launched Oracle Business Intelligence application as well.

In Oracle it doesn’t matter what kind of applications you want to build, the architecture remains identical so that you can put the data in the middle of the architecture and bring the processes to the data. Later this year, we are going to announce the fusion applications. One of the things that they will integrate is the three styles of computing that we talked about –transactions, analysis, and collaboration.

Q: This may be true for a homogenous environment. What if the customer has an extremely heterogeneous environment?

A: Heterogeneity is a norm today. In fact customers should choose the best for their businesses. If I need to describe Oracle’s strategy in four simple words, they would be: functionally complete, open standardsbased, integrated and best of breed. Openness involves embracing the heterogeneity of the customer’s IT environment so that they can make the best choices for themselves.

As we build more complete, more best-of-breed solutions, in the long term most customers will find it suitable for their business to deploy what Oracle can offer. But along the way, they may make other choices which are unique for their own industry and it makes sense.

Oracle will just make it easier for the customers to adopt these solutions because whatever we do, we use Java, SOA and open standards for easy integration. However, as a best practice advice to customers, we suggest simplifying and standardising the infrastructure to derive better business value.

Q: The final piece remains unanswered. How can CIOs bridge the gap between the real and virtual environments?

A: The key concept here is the notion of extreme transaction processing. If transactions keep your system in sync with the real world and the real world is changing very quickly, you will, practically, need lots of transactions.

Today's customers are looking at extreme levels of performance, which is technically very challenging and expensive to satisfy - Oracle is now building a technology where that level of performance is woven into the architecture. Again Exadata is a good example of how one machine can support a million random I/Os per second.

Getting such a performance out of a box is absolutely unheard of. Customers have personally reported that when they adopt this technology, they see a 10-fold increase in performance. In the past, there was normal IT for normal purposes. When there was a need for extreme compute power, companies had to resort to proprietary systems. But today those extreme demands can be met by using normal, standardised off-the-shelf technology like Exadata.

And when you combined that with ubiquitous network like wireless network and smart portable mobile devices, it helps in keeping the systems perfectly in sync with reality. This is the trend. And the business solutions created by Oracle are well in tune with the trend.

When you put all these pieces together, you get a massively scalable, continuously available, standards-based grid to handle every kind of data for every purpose keeping in sync with reality. This is the future of enterprise computing.

Q:I am compelled to ask if all the acquisitions, done by Oracle till date, were aimed at creating this massively scalable and continuously available grid.

A: I have been studying Oracle since 1987, much before I started working for it. I am convinced that the strategy of the company hasn’t changed ever. It is just the tactics that change. The decision to acquire Sun Micro- systems is part of the latest tactics and not a change in strategy. I will try to paint a simple picture for you.

The business value in IT is like an upside down pyramid. The biggest value resides in core industry-specific applications. If you are selling a core banking system to a bank, that’s the conversation you will have in the CEO’s office and not the CIO’s office. But if you are selling some storage to a bank, you may probably talk to the CIO or the storage management team.

In Oracle, we started with database, went up in the middleware and database tools and then into horizontal applications, which we sold to everybody. After that we got into industry applications mostly through acquisitions. We very smartly integrated the front office to the middle office and the back office and offered a whole suite as one.

Progressively, the company is modernising, integrating and standardising the applications to deliver an open, complete, agile, industry-specific footprint for which we have literally no competitor.

Cross=Posted from CTO Forum

Possibly Related Articles:
Enterprise Security
Enterprise Security Oracle
Post Rating I Like this!
The views expressed in this post are the opinions of the Infosec Island member that posted this content. Infosec Island is not responsible for the content or messaging of this post.

Unauthorized reproduction of this article (in part or in whole) is prohibited without the express written permission of Infosec Island and the Infosec Island member that posted this content--this includes using our RSS feed for any purpose other than personal use.