DB2 Forecast is Partly Cloudy Necessity is the mother of invention, or so the saying goes. If this is indeed true, the current economic slump experienced by all major economies around the world is bound to produce a lot of invention. The most damaging factor has been the fear and uncertainty that business leaders are feeling regardless of the actual performance of their companies. The doom and gloom propagated by the media is making even the most bullish of the business leaders reevaluate their companies’ spending. Many of the IT projects, even those that are committed and funded, are finding it more difficult to proceed. IT projects that require large capital investments are undergoing more rigorous scrutiny and ever higher levels of approval are required for even lower levels of capital budgets. Indeed, capital spent is under attack and more IT leaders are looking to reduce CAPEX (capital expenditures) and replace it with OPEX (Operating Expense) wherever possible. Strangely enough, this trend did not start with the onset of the current economic crisis. It is actually the years of rapid growth before the current slump that prompted many in our industry to reevaluate the way we procure and deliver IT infrastructure. Indeed, in an article “Servers: Why Thrifty Isn't Nifty” written for Forbes.com Kenneth Brill illustrated very clearly that the model we’ve used so far is based on costs that are not well understood. When these costs are added up the picture we get is this model of procuring IT resources is not scalable, maybe not even sustainable. According to the data presented in the article, a single low end $2,500 server ends up costing as much as $15,400 when utility costs are taken in to account.

Figure 1 Utility cost of a $2500 server

And when one looks at a standard rack of blade servers, the cost is a staggering $1 million in facility capital. Not only do facility costs represent a significant chunk of the total IT spend, they are also the fastest growing portion with 20% annual growth rates. This is by no means a radical or an isolated view. Indeed, IDC published research on global annual server spending that clearly validates these statements.

Figure 2 Global Annual Server Spending (Source: IBM Corporate Strategy Analysis of IDC data)

Despite these costs, we are experiencing ever increasing demands for more computing capacity. For example, storage shipments are increasing 54% annually and the number of servers in the US is projected to reach 15 million by 2010 up from 10 million in 2005. At the same time, industry statistics tell us that 85% of computing capacity in distributed environments sits idle. The drive towards reduced capital expenditures and a desire to increase utilization of the existing compute capacity forms the backdrop in the recent rise in popularity in the notion of Cloud Computing. Cloud Computing overall is a very broad subject and there are a lot of really good articles on this topic. In this article I wanted to provide a very quick introduction and focus on what Cloud Computing means to database professionals and DB2 in particular. To understand where DB2 fits in to Cloud Computing paradigm, it is important to understand that Cloud Computing does not represent any new technology. Much like Centralized Computing in the 1960s and 70s, and the Client-Server Computing in the 1980s, Cloud Computing changes the model for delivering and consuming IT infrastructure and services. NIST provides this description of the three delivery models for Cloud Computing: Cloud Software as a Service (SaaS). The capability provided to the consumer is to use the provider’s applications running on a cloud infrastructure and accessible from various client devices through a thin client interface such as a Web browser

(e.g., web-based email). The consumer does not manage or control the underlying cloud infrastructure, network, servers, operating systems, storage, or even individual application capabilities, with the possible exception of limited userspecific application configuration settings. Cloud Platform as a Service (PaaS). The capability provided to the consumer is to deploy onto the cloud infrastructure consumer-created applications using programming languages and tools supported by the provider (e.g., java, python, .Net). The consumer does not manage or control the underlying cloud infrastructure, network, servers, operating systems, or storage, but the consumer has control over the deployed applications and possibly application hosting environment configurations. Cloud Infrastructure as a Service (IaaS). The capability provided to the consumer is to provision processing, storage, networks, and other fundamental computing resources where the consumer is able to deploy and run arbitrary software, which can include operating systems and applications. The consumer does not manage or control the underlying cloud infrastructure but has control over operating systems, storage, deployed applications, and possibly select networking components (e.g., firewalls, load balancers). DB2 strategy is to participate in all 3 delivery models. We intend to provide DB2 as the DBMS for third party and IBM Software as a Service (SaaS) and Platform as a Service (PaaS) offerings. DB2 has a lot of technological underpinnings that make it the right platform for SaaS and PaaS providers. Corent Technology chose DB2 to deliver their SaaS-Engine and SaaS-Factory software as a service offerings as well as SaaS-Powerplant platform as a service. There is a short video on ChannelDB2.com that shows how one can create a true SaaS application using Corent SaaS-Factory on DB2 in about 10 minutes. When delivered as part of SaaS/PaaS, DB2 is virtually invisible to the consumer of the service or a developer utilizing the PaaS platform. Unless working for a SaaS/PaaS provider, DB2 professionals will have very limited direct exposure to DB2 that underpins applications delivered using these models. Infrastructure as a Service (IaaS) represents an entirely different opportunity for DB2 professionals. In essence, IaaS (often called Utility Computing) provides DB2 shops with alternative methods of provisioning hardware to run DB2 servers. The benefit of utility computing lies in its ability to dramatically accelerate provisioning and deployment of required computing capacity, use this capacity on a “pay only what you use” model and shift the spending from capital to operating expense. Compared to traditional way of deploying a database server which requires weeks if not months for securing capital budget, purchasing hardware, arranging for data center space and server hook up, cloud computing offers ability to get a DB2 server up and running in about 10 minutes. And not only can you get your project off the ground very quickly, you will only pay for the time the server is up and running. When you are finished with the project, simply shut the server down and you will stop incurring any charges. For many, employing Cloud Computing will allow to move on with projects that would have

otherwise been stuck in budget approval processes. This is especially applicable to proof of concept and other projects that are much harder to get off the ground in today’s economic reality. In any conversations on Cloud Computing, Amazon Web Services (AWS) is typically cited as an example. As most large IT consumers, Amazon built out its data centers with a lot of spare capacity. Much of that capacity was unused most of the time resulting in increased operating expenses. In 2006, Amazon turned this expense in to a source of income by offering this compute capacity to anyone with a credit card. For as little as 10 cents for an hour you can have a server and run whatever software you see fit. You get root access and in general, the experience is no different than working with your own dedicated server in your company’s own data center. The only real difference is that you don’t really have physical access to the server and can’t use removable media like CD-ROM or tape. Since the introduction the popularity of the AWS as the cloud computing provider has grown rapidly especially within startup and small and medium enterprises. For startups this is an especially attractive model as it allows them to direct scarce funding to developing their core business rather than spending it on building out their own IT infrastructure that may never actually be used. This model of consuming IT resources as a utility that is provided by a third party for public at large is called Public Cloud. While very popular with startups and small and medium enterprises there is hesitation on the part of many larger businesses to embrace public clouds. This hesitation is based on concerns about security, privacy and resiliency. I don’t mean to say that public clouds don’t deliver the level of security that large enterprises require but there is a perception that the level of security exposure is higher than what is provided by non-shared on premises IT systems. However, the attraction of the Cloud Computing benefits is very strong and many larger enterprises are looking to build their own cloud computing infrastructure for exclusive use of their own organizations. This deployment model is called Private Cloud. The reason Amazon is able to offer compute time for 10 cents an hour and still manage to make a profit is use of 3 key practices: 1. Standardization 2. Virtualization 3. Automation When building on premises systems, a lot of planning and thought goes in to picking the right processor architecture, storage subsystems and other aspects of the infrastructure. These technologies and components are picked for suitability for a particular application and workload. While producing a well suited infrastructure, this specialization increases costs. Cloud Computing foregoes specialization and builds out generic compute and storage capacity from standardized components. Result is a much lower cost infrastructure albeit not specifically tuned for any specific application or workload. Dedicating a physical server to a specific application or user even in a highly standardized environment is still wasteful of resources. This is where virtualization comes in to play. By virtualizing processing power, storage and networking, resources can be very quickly and efficiently allocated across the standardized set of servers, storage and network devices. Not needing to allocate the entire server or

storage capacity of a physical device to a specific application results in much higher utilization of resources and dramatically reduced costs. The other benefit of virtualization is added resiliency as virtualized resources can quite easily be moved around cloud infrastructure. One of the largest components of the IT costs is the human costs. Operating hundreds or even thousands of servers is an expensive proposition. From the user perspective, one of the most valued aspects of Cloud Computing is the idea of self service. Automation, the third pillar of cloud computing, is key to both reducing the operating costs and delivering the self service experience that is so valued by the users. Another variation on a deployment model is known as Hybrid Cloud. Hybrid Clouds are nothing more than a combination of use of a Private Cloud infrastructure in conjunction with Public Cloud use for situations when there is insufficient capacity available on the private cloud. Use of Hybrid Clouds allows CIOs to avoid overspending on Private Cloud infrastructure to accommodate spikes in demand through the use of Public Cloud resources. Anyone looking to deploy DB2 on a private, public or hybrid cloud will be looking at highly standardized, completely virtualized and extensively automated environments. DB2 has done a lot of work to enable a good fit in to just such environments. Some of this was accomplished by changes to the product, and some by close cooperation with our partners that create cloud computing ecosystem for DB2. On the standardization side, DB2 is fully enabled and supported on industry standard platforms that are popular in the Cloud. This includes both 32 and 64-bit Linux platforms that are the backbone of most cloud deployments. DB2 is optimized for Linux and is fully supported on a number of Linux distributions including Novel SUSE Linux Enterprise, Red Hat Enterprise Linux, and Ubuntu. DB2 also runs on other popular distributions of Linux such as CentOS but IBM does not provide formal support for these distributions. In addition to Linux, DB2 is also easily deployable on Windows servers in the Cloud, both 32 and 64-bit. Solaris x64 on Intel/AMD processors is another popular cloud computing platform and there is DB2 available for Solaris x64. It is also important to point out that not only is DB2 available and supported for all of these platforms, it is also available free of charges. IBM makes DB2 Express-C available for free and it is deployable for production use in private and public clouds. Naturally, there is DB2 for other platforms such as AIX, HP-UX and Solaris for SPARC as well as Linux for POWER and Linux for System z but these platforms are not very common in the world of Cloud Computing yet. Another key point related to standardization is the choice of system sizes. Most cloud providers make available smaller machines. For example, the largest Amazon EC2 instance (server) one can get is called Extra Large and it has 15GB of memory and an equivalent of 4 virtual cores with 8 compute units. (Note: This imposes severe constraints on database systems in the Cloud. DB2 has a very unique solution to this issue but that is a subject of another article). This type of instance costs $0.80/hour. The smallest EC2 instance costs $0.10/hour and has 1.7GB of memory and one virtual core. DB2 autonomic capabilities such as STMM are very helpful for making DB2 run well in resource constrained environments. In addition, the threaded engine architecture of the DB2 engine greatly reduces memory requirements and makes good use of multi-core processors.

DB2 is a good fit for virtualized environments demanded by the cloud computing. DB2 has been tailored and optimized to work with a number of industry leading hypervisors such as VMWare, XEN and KVM and others. On the business side, IBM has adjusted its license terms to make it easy and economical to run DB2 in virtualized environments. The new terms and conditions for virtualization are fairly unique in the industry and is a welcome change for those looking to dabble in Cloud Computing. DB2 also exploits virtualization to make provisioning of DB2 servers in the cloud a very simple affair. For example, DB2 is now available as a set of Amazon Machine Images (AMI) that make creating your own DB2 server in a public cloud an exercise that takes only a few minutes. There are two types of AMIs that are available: 1. Development AMIs and 2. Production AMIs You can also create your own AMI. There is a 6 minute video on ChannelDB2.com that illustrates how you could do that. Development AMI as you would expect is targeted at development activities. These are prepackaged instances of DB2 installed and configured on SUSE Linux Enterprise (SLES). All you need is to get an account on Amazon EC2 and you can use this AMI to boot up one or more of your very own development DB2 servers. There are no charges to pay IBM and you will only be paying the standard Amazon EC2 charges for the servers that you start. Naturally you do not pay when these servers are stopped. Production AMIs allow you to use both SLES and DB2 in a production environment. Just like the other AMIs they let you create your very own DB2 servers in minutes. However, the charges for use of these AMIs also include charges for the SLES as well as DB2. In other words, Amazon will bill you not just for the hardware use but for the entire stack that also includes the operating system and DB2. All of this is charged on a pay-only-what-you-use basis. For example, if you created a DB2 server on a Amazon EC2 Small Image you would be paying to Amazon $0.38 for every hour that your server is up and running. For more information I recommend taking a look at the IBM zone on the AWS website. So, now you know that you can get DB2 running on an Amazon cloud using either free DB2 or pay per hour of use. However, many of our customers already have sufficient DB2 licenses they may want to utilize these licenses for their cloud deployments. IBM has adjusted license terms and conditions to do exactly that. If you have DB2 licenses, these licenses can now be applied for on-premises deployments as always and these licenses can be applied to create DB2 servers on the cloud. This provides for maximum flexibility and allows customers to accelerate or restart stalled projects by eliminating requirements for funding hardware acquisition and using software licenses that have already been acquired. As I mentioned before, high degree of automation is the hallmark of Cloud Computing and is the attribute that delivers the self service aspect of the cloud that users appreciate so much. In addition to automating installation of DB2 by making it available as a set of packaged Amazon Machine Images (AMIs) DB2 is also available for deployment as a set of templates on RightScale.com. While RightScale.com uses a different process for creating DB2 server instances the net effect is the same, you can have a running DB2 server on an Amazon EC2 cloud in about 6 minutes. As a matter of fact, we

made the latest version of DB2 (version 9.7) available on RightScale.com the same day we made it available for deployment in an on-premises environment.

RightScale is the leader in provisioning, monitoring, management and automated administration of cloud computing resources. Most of enterprise DB2 customers will be looking for the high level of automation of the administrative procedures that RightScale delivers. It is impossible to cover a very broad subject like Cloud Computing in a short article. I am hoping that this article was enough of an introduction and perked your interest in Cloud Computing and the role that DB2 plays. If you want to learn more on this subject I recommend watching a replay of a webcast we did on DB2 and Cloud Computing. We definitely see Cloud Computing in the future of DB2. Our ambitions are nothing less than to make DB2 the best DBMS for both private and public clouds. As we make this happen, we will be spreading the message through an extensive series of articles, webinars, presentations and white papers on this topic. If you are interested, I recommend you subscribe to the FreeDB2.com RSS feed and stay tuned.

Leon Katsnelson works at the IBM Toronto Laboratory, which is home of DB2 Development. Leon has been working on DB2 since 1993 when DB2 mission was transferred to the IBM Toronto Lab. Leon and his team lead the work on adoption of Cloud Computing and other emerging technologies that help more people experience the benefits of DB2.