Software AG / Terracotta at Process World 2011
At Software AG's Process World in Orlando, which ran from June 27-29 this year, I had a chance to speak with a number of executives at Terracotta, as well as Software AG, which recently acquired the company. During my conversations throughout the day, I gained some insight into why Software AG made the acquisition, why Terracotta found Software AG an attractive suitor, and what their combined strategy is for now and for the future. A lot of this is interesting from the Enterprise Java perspective, so I decided to blog about it here.
I mainly asked questions about Terracotta's products for Java developers (Ehcache, BigMemory, Ehcache Search), but I also asked about Software AG's BPM stack, and its recent acquisition of Metismo, a mobile middleware company that offers tools to target multiple mobile devices with one source code base.
These questions and answers sprang from conversations at a press roundtable, as well as one-on-one interviews with:
- Wolfram Jost, CTO of Software AG
- Amit Pandey, CEO of Terracotta
- Ari Zilka, CTO of Terracotta
- Lynn Vojvodich, CMO of Terracotta
Q: Wolfram, can you give some insight into Software AG's acquisition of Terracotta; and Amit, can you talk about Terracotta's perspective?
Wolfram: From Software AG's perspective, the acquisition was attractive for the following reasons:
- Scalability and performance continues to be critical for BPM, as data volumes increase continually. Terraccotta technology is key here.
- Cloud enablement: In-memory and distributed caching was a key need here that Ehcache solves head-on.
- We believe that the database-centric architecture will fade away, and an in-memory strategy will be required going forward.
Amit: Terracotta turned down other acquisition offers over the years. But Software AG was attractive because they were excited by the open-source, in-memory aspect, and their vision to replace the database [as the central point in enterprise architecture]. Just as importantly, however, their goal is to allow Terracotta to run their business and treat Software AG as their biggest customer. This was in contrast to others who wanted to absorb us completely. We thought this was too attractive to pass up. In the end, the Software AG acquisition will help Terracotta go mainstream, but also help us to expand beyond Java, to other types of developers.
Q: Specifically, what are the strengths of Terracotta Ehcache that made it an attractive acquisition for Software AG?
Amit , Ari, and Wolfram: An in-memory approach to data leads to less risk, as it reduces the need for hardware. Terracotta applies caching in a way that you can cache more of your data in a standard way than most people can usually do on their own in a custom way.
Other companies and products (i.e., Akamai) have not been able to kill Ehcache because they use a static caching approach, whereas Ehcache is a dynamic cache. We're also talking with SSD vendors to come up with a hybrid in-memory/off disk but not in-memory, solution.
We have very aggressive customers, i.e. online gaming vendors love Terracotta. This is one area that's very innovative and greenfield, and very demanding. Travel and leisure is another one. Beginning this year, we're going strong after financial services. Basically, Ehcache appeals to anyone who needs performance at any cost, regardless of company size. ESPN.com is built on Ehcache, and they require only eight servers to serve the entire Web as a result. Insurance companies like Aetna and MetLife are two more examples of companies with extreme data needs.
Q: When BigMemory for Enterprise Ehcache was first launched, I heard a lot about latency and how it improves that and predictability. Is this still a focus?
Ari: Yes. In fact, Terracotta began as a company by asking wall-street companies if they'd be interested in a standard, de-facto, caching solution. The overwhelming answer was “Yes!” At the time, the data sets to cache were small enough because they weren't able to cache it all effectively. In many cases, this is still true, but with Ehcache usage on the rise, these companies are quickly caching more and more data. Although the data demands continue to grow, BigMemory has been meeting the increasing needs.
Q: Regarding Software AG's acquisition of Metismo (a company that offers tools for write-once-deploy-many in the smart phone space), what level of quality does the write-once/deploy-all approach really achieve? What gets lost?
I get that question often. Some things, like touch, get lost, but we're targeting enterprise applications, not games or utilities. So while we may lose 10% of a device's specific features, we don't think this is an issue for enterprise users. Also, quality is not an issue at all. BPM enablement is about access to data, and this is what Metismo's technology gives us for the mobile space.
Q: Wolfram, in your presentation this morning on business process management (BPM), your slides show applications below the processes (see Figure 1). I always envisioned it the other way around, with applications built on processes and capabilities. Can you explain this?
It's all about perspective. The process is a provider of process logic, and the application is a provider of business logic, and this leads to maximum flexibility. We put processes on top because the application is the enabler of the process. A process is what is to be executed, and the application is only the enabler of some feature set or functionality. We find that over time the applications may change, but the processes do not need to change.
Q: HTML5: myth or integration opportunity, and will HTML5 lead to a new level of caching?
Wolfram: We view HTML5 as an important area, and the Web UIs in our products and solutions start with HTML5 because it helps us target mobile devices. We find the main distinction between native mobile app versus an HTML5 app is connectivity. Does the application need to run and gather data even when there is no connection to the web? Another issue, HTML5 apps need to share space with the browser. Overall, however, we view the need to support both mobile apps and the mobile web (HTML5) equally important.
Amit: We take a layered approach to caching, but tend not to talk about how early the layers intercede in the model. Overall, we can get into it pretty early. For instance, with Ehcache Web which caches web data from web servers and other sources applications can get into the cache before the web server is invoked. It's effectively a Web filter before all the servers.
Q: With more enterprises looking to host with cloud vendors and PaaS providers are these vendors the new customers?
Wolfram: What is a cloud service provider? Do they really offer a PaaS? We believe it's simply about consumers and providers. It's a value chain with a many-to-many relationship. We have customers who already host Software AG's stack and now Terracotta within Amazon's cloud platform. We indeed need to dive in at a deeper level, so the Amazons and Googles of the world will offer our stack out of the box. So we're working with EC2 today, to do this.
Q: Ehcache Search allows you to perform queries against your cache, and avoid hitting the database in many cases. Do you have plans to use this as a launching pad to eliminate the database altogether?
Amit: Database vendors are an ecosystem, and we do not want to take on an ecosystem. Database interfaces, search, security, bulk load these are all core database technologies that are required. There are certainly other database features that enterprises will require, so it's probably not feasible to replace databases outright. We look at it more like Hibernate Grid, with JPA straight to Terracotta.
Q: Any interest in integrating a NoSQL solution into the stack, such as MongoDB with Ehcache?
Wolfram: Yes, but these are different products. Our process-modeling solution takes the first step, where we replace direct RDBMS access with Terracotta. In this use case, you can live without a traditional database in the back end. However, other areas such as reporting are an issue. Solutions need to be found that solve these problems in the NoSQL space. Databases are key there. But for applications? Yes, over time, we see enterprises moving more towards NoSQL and so will we.
Q: Any desire to offer hardware appliances for those who want in-house solutions, similar to the new Oracle approach?
Wolfram: Customers have been asking us the same question. We need to figure out what it means to Software AG, and where it belongs in our product offering. Process intelligence could be a good place. It's standardized and highly customizable. But it could lead to new integration issues all over again; the same sort of issues that BPM aimed to resolve in the first place. And we certainly don't want to lose flexibility. In general, it only makes sense if there is tremendous value in doing it. Optimization has to lead the customer to greater value. I can see us combining hybrid SSD, and physical memory, with Terracotta BigMemory and Ehcache, and delivering the layers with I/O self tuning in some sort of appliance.
Amit: Early on, we looked at offering an appliance...We were going to market a "Grid Server." What we found was that the value proposition had to be so high that it may have been impossible to attain. The answer for us was "no." Our opinion is that the success of Oracle's Exadata integration is a result of the fact that it's difficult to configure their software product.
Q: Some financial companies put firewalls between the application and DB layer. Any security concerns amongst customers for an in-memory cache in front of the firewall?
Ari: Yes, but most are less secure than that, outside of banks, retailers, PayPal, and government regulations, or anything to do with credit card data. With Ehcache, data is encrypted at every stage, i.e., on the wire, in memory, across API calls, and so on. The NoSQL guys are still focused on core algorithms, but Terracotta has more experience. Yes, there are always security challenges, but it's not a deal breaker at all. We may have lost one deal early on due to security, but that doesn't happen anymore.