In the past few weeks I found myself involved in various discussions centered around the challenges that enterprises face today when they want to deploy their application on the cloud.
These discussions were very timely as they gave me some interesting ideas for my talk next week Practical Guide for Developing Enterprise Application on the Cloud, taking place on Monday, April 20 at 11:00 am Eastern Daylight Time (GMT -04:00, New York), at the CloudSlam online conference.
I thought that some of those discussions are worth sharing. In this post I'll try to summarize the main highlights from those discussions.
I’ll start by pointing to the following discussion thread on the cloud mailing list: Challenges faced by developers and architects when moving to the cloud:
Kent Langley responded by listing the following challenges in his response:
- Dealing with a lack of persistence in some cases Dealing with distributed programming models (prob. one of the most important ones imho) Having to think about the whole stack. Not just the code.
- Caching considerations
- Messaging
- Using Memory Data Grids
- Understanding configuration management tools that might be involved Working more closely with the operations group in some case
Kevin Apte added the following comment:
There are many challenges- There is no out-of-the-box infrastructure for
hosting the typical J2EE and SOA Stack in the cloud. There is no Weblogic,
WebSphere, ALBPM, Message Bus like Tibco available in the cloud.A development team could certainly move all of this into the cloud, but the
configuration, licensing issues etc. are all something the team would have
to solve on its own. This is far too bleeding edge for many people.
Robert Hankel added other challenges
.. The problem of adding additional resources dynamically (e.g. more WebCache instances, or WebLogic servers) requires sophisticated distributed system management infrastructure where the entity being managed is no longer a physical or virtual box, but rather an array of boxes acting collectively as a single system..
Grig Gheorghiu posted an interesting write-up Experiences deploying a large-scale infrastructure in Amazon EC2, where he provide a very insightful summary of his lessons with deploying a large scale application on EC2.
Below are the main takeaways from Grig's summary, which I found relevant for this discussion:
1) Deploy multiple Web servers
2) Deploy multiple load balancers
3) Deploy several database servers.
4) Another way of dealing with databases is to not use them
Challenges summary?
It's easy to see that there is a common theme behind all those comments. Taking existing enterprise applications to the cloud can be very difficult simply because a) most of today's enterprise applications were built using frameworks and technologies not yet supported as first class citizen by cloud providers and b) most of those applications were not designed to take advantage of the cloud's elasticity.
Rather then pointing to my direct response to each of those challenges i thought that it would be better to provide a short summary of the main possible solutions that came through this discussion.
Does it have to be that difficult?
No. Below are two main approaches to those challenges.
- Packaging static images
The simplest approach would obviously be to package your local IT environment into images that could be easily ported to the cloud in the exact same way they run in your local IT environment; right? Well, yes, you can package anything in an image bundle and host your virtual machines in a reserved mode with fixed IP configuration. However, being able to technically do that doesn't mean that it makes sense. I would question what's the difference between this environment and any other hosting environment, and what do you expect to get by moving to such a hosted environment vs. running it in your local IT environment.
If you are going to try deploying your existing IT application on the cloud using static images, then most likely you'll end up "porting" not just the application but also the problems you were facing in your local IT environment; i.e. your application will be over-provisioned based on the peak load and you’ll end up with poorly utilized environment.
- Fully elastic application
The main driver for moving to cloud based environment in the first place was to be able to grow as you need and pay for what you use.
The question is whether you can deploy your application without changes to the application while at the same time leverage the elasticity that cloud brings.
Sounds impossible? Well, a good example that does just that is Storage. With storage you can take your existing application, run it with your local (static) disk, and then plug in a network storage and run the same application on that network device, without changing the application. In that world, instead of taking your existing local disk and virutalize it, you are taking the application and plug it to another device that has visualization built-in.
We can use the same approach as with storage; i.e. move your existing application code and run on top of a different underlying implementation that will enable you to capture the elasticity of the cloud without forcing you to re-write your entire application. If you're running in a JEE environment, it should be fairly easy. If your application has strong "ties" to back-end systems, you can use the hybrid model where your application front-end is running on the cloud, while being connected to the back-end system, using secured communication channel.
Is it really that simple?
Yes. My experience with the integration work that we had done between our Cloud Computing Framework (CCF) and EC2 showed me that getting a production-ready JEE application, including a load-balancer, self healing, auto scaling, security, database and even data grid capabilities plugged in, is actually much simpler than with any other environment I'm aware of. This is due to the built-in automation, predefined images, and the fact that I don't need to download and setup anything to get the entire system up and running. In fact, it's so simple that we decided to built our entire Demo as a Service framework around it; a framework that is used quite successfully and constantly, with customers, prospects, and now with partners.
Are there any production references?
Yes. Read Jim Liddle's blog post and see a good example of an Enterprise JEE deployment that is already running on EC2, in production, on top of our new Cloud Computing Framework. In his post, Jim describes how this Telco operator were able to address some of the common challanges that were mentioned above such as security,flexability, cost, development complexity and lock-in as well as high-avliability and scalability in a relatively simple manner .
What next?
In my discussion in the CloudSlam event I’ll try to provide a more detailed practical guide that demonstrating how you can take a step by step approach for porting existing JEE application to the cloud. I hope to get lots of questions and feedback during the discussion so that I can share those with you in one of my follow-up posts.
