Raising the bar for enterprise IT
There is no doubt that cloud computing has put a lot of pressure on IT organizations to find a way to reduce costs and, more important, bring agility to business processes. Line of business users are increasingly bringing expectations from their consumer experiences to the front door of IT. The thinking is that if it takes a few minutes to get a Gmail account and it is free, why does it take weeks to get an Exchange account on a corporate email system? In an organization of any size, it takes weeks if not months to get budget approved to buy a server. It takes even longer to order, get it delivered, racked and cabled, fitted out with an operating system at the right level and with the right patches. Yet, I routinely start a virtual server on the cloud with DB2 and a content management system installed on it in a matter of about 6 minutes and I never have to get any budget approved or apply patches. I pay as little as 2 cents per hour for this server and only for the hours that I actually have it powered up. Today’s enterprise IT has some high expectations to meet.
Cloud shows the way
The way IT managers are raising to the challenge is by adopting the same techniques as those used by the public cloud providers and large scale data center operators. The set of techniques is quite simple. You consolidate and standardise your servers, storage and networking equipment to create a shared pool of resources. You virtualize it so that resources can be allocated on demand, and, you automate it to the max to enable self-service. The difference is that, unlike public cloud operators who rent this virtualized capacity to anyone with a credit card, enterprise IT creates this shared resource pool, the cloud, solely for the benefit of their enterprise. The cloud resources are located in enterprise’s own data centers behind the firewalls and protection of their security processes and personel. So, while the result of this consolidation, virtualization and automation is what we would call a cloud, it is a private cloud.
Using virtualization does not mean you have a cloud
If private clouds are so easy, why isn’t everyone just cranking out a few a week? Many IT teams are well on the way of doing step one and two i.e. consolidating and virtualizing the infrastructure. It is extremely rare to come across an organization that has not had some experience with virtualization. Some will even go as far as to say “we use cloud computing because we use VMWare on our servers”. Wrong! Virtualization is important … delivering IT capacity via self-service is what makes a cloud. You must have automation or your benefits will stop at server consolidation which is a minor portion of the savings that cloud computing promises.
What is automation and self-service? In traditional IT, you provision and manage equipment for a project. Your IT team may be able to take on 10-20 projects a year. The rest are what the industry calls “application backlog” or what non-IT folks call “my stuff that will never be done”. For most companies, the application backlog is an order of magnitude larger than the number of projects that actually get funded and implemented. The promise of cloud computing is to greatly reduce or even eliminate the application backlog by removing the capital budget constraints and by greatly reducing the IT effort to provision and maintain the infrastructure. A fast rising gaming company is known to provision over 15,000 servers. You would think they have hundreds or even thousands of system administrators. They have eight. It is the magic of automation that makes this possible. In its’ simplest form, cloud automation provides a catalog of services and a way to deploy these services on to the cloud servers, storage, and network. And when done right, service requesters can describe what they want from the catalog and can specify how they want it deployed, all without involvement from system administration staff.
IT vending machine
Where do you get all this wonderful automation from? As with most things IT, you can build it or you can buy it. IBM happens to offer a solution called WebSphere CloudBurst Appliance (WCA) that delivers a great way to deploy complete WebSphere environments on private cloud infrastructure. Note the emphasis on complete; we will get to it soon enough. First, a really important point that many people miss. You do not deploy to the WCA appliance i.e. it is not a server on which you run your software. WCA is a distribution and management point, a catalog, not a run-time environment. So, while it helps you dispense software, it does NOT run any of your software. The software runs on the virtual machines that make up private cloud.
Think of WebSphere CloudBurst Appliance as a vending machine. However, instead of chips, cookies and soda, WebSphere CloudBurst Appliance is stocked with saved virtual machine images that have IBM middleware and your software. Instead of picking a couple of bags of potato chips and a cola for an afternoon snack, you get a cluster of WebSphere servers and a DB2 database server deployed in minutes on to the hardware that is your private cloud. This is why I said complete WebSphere environments in the previous paragraph. Current version (v2) of WebSphere CloudBurst Appliance is really focused on WebSphere environments. A typical WebSphere application would obviously have one or more WebSphere Application Servers (WAS) and it may have other components like messaging and queuing, load balancers etc. And in almost every case it would also have a persistent data store i.e. a database. This whole environment is called a pattern in the WCA world. The process of creating and deploying patterns replaces punching of the codes for various items on a typical vending machine. WebSphere CloudBurst provides a nifty component it calls “pattern editor” where you drag components from the catalog to define and describe the servers that comprise a pattern. Four WebSphere Application Servers, a load balancing server, and a cluster of DB2 servers in a high availability configuration … this may be a pattern that you would create for your next project. Not only can you specify the servers, you can also specify the connections between these servers. Drop a line between your WebSphere cluster and the DB2 HADR cluster and you have the WAS connection pool, properly configured for getting data out of your database. Do these patterns have to have WebSphere servers? Not at all, you can use WCA to provision patterns of just DB2 servers if you wish.
Where is DB2 Hypervisor Edition?
If WebSphere CloudBurst Appliance is a vending machine, how do you keep it stocked? How do you pay for the software it deploys? WCA has a catalog of virtual machine images with various IBM middleware products. While the images are in the catalog on the WebSphere CloudBurst Appliance, you still have to purchase the licenses. WebSphere Application Server machine images are known as “Hypervisor Edition” and you would be purchasing licenses for these special “Hypervisor Edition” products and entering these licenses in to the “License Management” tool that is a part of the WCA.
Current version of the WCA does not have DB2 images on the appliance. You download the images and add them to the appliance yourself. The process is very easy and takes a few minutes. As with all products, DB2 images contain configured and optimized virtual machines complete with the operating system (SLES) and DB2 Enterprise Edition v9.7 software but they do not contain licenses for either the operating system or for DB2.
Also notice that DB2 is NOT called “Hypervisor Edition”. What is the implication of that? This small detail provides a great deal of flexibility to our customers. This means you can purchase standard DB2 Enterprise Edition licenses and apply these licenses for DB2 running on “bare metal” hardware, or toward running on virtual servers in the private cloud, or toward virtual servers in the public cloud like Amazon EC2 or the IBM Cloud. Let’s say you bought a 1000 PVUs of DB2 Enterprise Edition. You could allocate 600 toward your DB2 servers for your existing hardware and put remaining 400 PVUs in to the WCA License Management tool. Whenever patterns containing DB2 servers are deployed to your private cloud by the WCA, provisioned license capacity is subtracted by the License Management tool. If any of these patterns are decommissioned (e.g. you finished your test cycle and no longer need the environment) these licenses are returned back in to the WCA license pool for other projects to use. If it turns out that you do not need all 600 PVUs for your physical servers and 400 PVUs will suffice, you can add 200 PVU licenses to your WCA to make available to whatever projects need the capacity. While this is very flexible and convenient licensing, many of the WCA customers expect products that are provisioned by the WebSphere CloudBurst Appliance to be called “Hypervisor Edition”. So, once in a while, you may actually see DB2 images being called “Hypervisor Edition” as well or we may refer to an images as “DB2 Server for WebSphere CloudBurst Appliance“. Just remember, the product you purchase from IBM or an IBM business partner is DB2 Enterprise Edition (and other editions in the future). We think the flexibility more than justifies a bit of awkwardness in the naming and we hope you will agree.