ANTs: a strange name for a database
ANTs not pants
Comment ANTs Software, otherwise known simply as ANTs (which actually stands for Asynchronous Non-preemptive Tasks), is little known, yet it is a public company that has been around since the 80s. It started life working on parallel supercomputers but then faded into dormancy for the best part of a decade. It then re-emerged in 1999 as a project to implement its MPP (massively parallel processing) expertise within the database field. As a result the ANTs Data Server (ADS) was launched last November.
There are a number of interesting things to say about ADS. The first is that it contradicts almost everything that I used to write about databases back in the 90s. Then, we regularly used to castigate the likes of Sybase for being late in implementing row-level locking, which we regarded as being essential for OLTP (on-line transaction processing) environments. Well, ADS not only does not implement row-level locking it promises a lock-free environment altogether.
There are a number of facilities that allow this. First, it allows simultaneous updates to a row, provided that different fields are being updated. Secondly, it treats separate updates to the same field (say, adding three stock and deducting two) as arithmetic operations where these are commutative (that is, where operations can be performed in any sequence and produce mathematically consistent results). It is only when neither of these conditions applies that the software has to resort to traditional means: serialising the process of update based on user-defined priorities.
The second interesting thing about ADS is that ANTs claims that it typically runs five to 15 times faster than standard relational databases. The fact that it offers a lock-free environment is one reason for this. Another is that it uses a process called "preparation by compilation" which effectively means that SQL is compiled into binary code and then held in memory, where it can be (re-)executed as required. A third reason for its performance advantage is that ADS treats memory as its main area for processing. However, unlike in-memory databases such as TimesTen (now part of Oracle) ADS overflows to disk where necessary. Finally, a fourth reason is that ADS creates multiple threads when it is started but it never closes them - keeping them open and re-assigning them as needed, thus saving the overhead of starting up threads.
The third interesting thing about ADS is that its superior performance does not seem to have compromised scalability. For example, Nextel, one of its customers, is processing 10 million transactions per day using ADS.
The fourth interesting thing about ADS is that as standard it comes with support for Oracle schemas, stored procedures (PL/SQL), triggers and so on. In principle, you can take an Oracle database with all of its logic and implement it, without change, on ADS. What's more, this facility is not limited to Oracle but also applies to Sybase and SQL Server with support for Transact-SQL and T-SQL. In the next two releases, due before Christmas, this support will be extended to other well-known databases.
The last sentence may raise some eyebrows: two releases before Christmas! ANTs makes a big play of its combined R&D and QA capabilities. It has 500+ servers in its data centre and all new code is checked in and quality tested hourly. Typically, it claims to do 3,000 hours of automated QA testing every day, across all supported platforms (11 of them) - so that's why it reckons to be able to turn out new releases so quickly.
Finally, there is one more interesting thing about ADS: it is less expensive than any of the products it is aiming to replace. In fact, the company is not solely aiming at the replacement market but it is also addressing the high performance OLTP market in its own right, as well as the SMB sector. In this last case, its entry-level pricing is $1,995 per server per year for a maximum of 2 CPUs (which may be dual core) including, where appropriate, some migration services.
To conclude, ANTs may still be a strange name for a database company but it has a product that it is certainly worth watching.
Copyright © 2005, IT-Analysis.com
Sponsored: Benefits from the lessons learned in HPC