Azul Zing Reviews

As a large scale high performance trading platform serving the high frequency trading community, latency has always been critical for our clients.

Before using Azul/Zing JVM, we had to spend a lot of effort tuning our JVM GC and profiling our application to minimize latency, which is sometimes very unpredictable due to Java GC. Azul/Zing solved our Java GC problem.

With minimal effort, Zing has helped us improve our latency performance dramatically. There are no more jitters or pauses that used to require hours or days of research to resolve in our production environment. Zing has freed our development resources to focus on our core business.

Like every other great product, there is always something that can be improved. Although Azul claims to have a fast warmup, we still experience higher latency for the first trade when our system starts up on Sundays. It would be extremely helpful if Zing JVM can learn to warmup that code path or print some insights of the path that lead to a solution. The other feature we find very useful, but would like to see some improvement on, is the Azul’s Java profiling tool.

The pauseless garbage collection is the primary feature. ReadyNow is becoming more and more important.

As we provide real-time, low-latency trading solutions, any 'pause' in the application flow can cause serious financial risks to our customers. We used to spend numerous hours tuning each release, and then each customer site, in order to minimize the length of pauses in the standard Java runtime. With Zing, we've now been able to re-allocate those tuning resources to general product development.

As a large scale high performance trading platform serving the high frequency trading community, latency has always been critical for our clients. Before using Azul/Zing JVM, we have had to spend a lot of effort tuning our JVM GC and profiling our application to minimize latency, which is sometimes very unpredictable due to Java GC. Azul/Zing solved our Java GC problem. With minimal effort, Zing has helped us improve our latency performance dramatically. There are no more jitters or pauses that used to require hours or days of research to resolve in our production environment. Zing has freed our development resources to focus on our core business.

Like every other great product, there is always something that can be improved. Although Azul claims to have a fast warmup, we still experience higher latency for the first trade when our system starts up on Sundays. It would be extremely helpful if Zing JVM can learn to warmup that code path or print some insights of the path that lead to a solution. The other feature we find very useful but would like to see some improvement on is the Azul’s Java profiling tool.

The most important Zing feature for LMAX Exchange is the C4 garbage collector. C4 effectively does continuous garbage collection and for LMAXExchange that means less latency jitter. Latency is a big deal for us and while low latency is important, predictable latency is also a requirement for our customers. Zing doesn't need to force periodic safepoints during normal operation, so assuming that you don't have any other sources of jitter in your program, you'll get a flatter latency profile with out-of-the-box behaviour.

Zing has completely solved the garbage collection problem for us. We no longer need to concern ourselves with tuning the garbage collector or coding around it. This has saved us an enormous amount in terms of potential opportunity cost and allows us to spend more time on the business problems that really matter. Not having to butcher our code to make it garbage free has probably also helped us to retain a high level of productivity and a low level of defects.

Pause-less JVM garbage collection on JVMs with heaps in the hundreds of gigabytes is the feature I find most valuable.

We use the Zing JVM to deliver Solr search capabilities to our clients. Without it, we would need a team of full-time engineers to continually shard and re-shard the indices so that small heaps are used to avoid pausing during garbage collection.

The non-stop garbage collector (C4 - Continuously Concurrent Compacting Collector) is the most valuable feature.

It has made it possible to spend time on developing new functions rather than optimizing old ones.

• GC Pause times are maintained in low milliseconds. This is beneficial for applications with high volume, high memory allocation rate, and a high memory footprint, (I would say >2.5gb of max heap). It also increases throughput by approximately 20-25% as compared to Oracle Java Development Kit, based on some tests that I have done with real applications
• Azul engineering cleans up Oracle Java Development Kit bugs before GA release of the product, so you will get less bugs
• The engineering support is very qualified, so they can help you diagnose code level issues and provide recommendations. Zing has free tools to get you to the root of many applications problems, like memory leaks, thread blockades, system limitations, etc.
• Support lifecycle is almost unlimited. Oracle Java Development Kit's lifecycle is three years, after that you have to buy a “commercial license” which is expensive

We use it on highly visible applications, with application characteristics described above.