PHP Next Generation is outperforming Facebook’s HHVM

Last month Andi Gutmans, CEO and co-founder of Zend gave an exclusive interview to Antoine Crochet-Damais at Journal du Net in Paris.

Below is the transcript of his interview:

JDN – What is your vision today on the future of PHP ?

Andi Gutmans. PHP is being used by 40% of web sites worldwide. At the same time we are noticing an evolution in companies towards web technologies for their applications metiers, but also more and more projects are going mobile. Projects needing to innovate rapidly, by being attractive and client-oriented. So, PHP has always been good on the web; it’s also a productive language. Traditionally, DSI were oriented to Java and .net. Now they’re in the process of transforming their capacity to deploy applications, and increasingly use PHP in this way. For us, it’s the main current change in the PHP world that’s making this language become stronger today.

The PHP ecosystem is also extremely dynamic, and is growing with lots of innovations. They focus on frameworks and PHP runtime especially, but also new disruptive tools in terms of continuous deployment. I’m thinking about like Composer [which manages dependencies between versions of NDLR components]. All that ensures we are living in exciting times, probably much more hectic than in previous years.

PHP 6 hasn’t come out yet, like the native integration of Unicode. Is that still a goal ?

That’s an important question. UTF has become a standard for internationalization. And PHP has been slow on the uptake. We have made significant progress in this direction by building an internationalisation library for PHP. Thanks to that, the language is being used today in Japan, China, in lots of countries. In PHP 5.4 for instance we have integrated transcoding options for the Japanese market. As for native support of Unicode, we decided not to go there because of a drop in performance, and the problem of exceeding memory capacity. But it can be integrated with the components we have added since then.

What do you think of the Facebook HHVM project?

It’s a very interesting project. I know the Facebook team working on it and they’re doing good work. But I think that the question of runtime is just one question of many in terms of performance. There’s also the question of database. In a broader sense, with choosing a language there is also the question of community, ecosystem… remains a reference. I think that the PHP community should stay united around one language, and not multiply the declensions. Fragmentation isn’t a good thing.

Recently you have also worked on improving PHP performance. In its phpng declension, how does PHP stand in relation to HHVM ?

We have effectively done lots of work on PHP runtime. HHVM integrates a just-in-time compilation, making it faster on some benchmarks. Across the phpng range, or PHP Next Generation, we have refactored and revamped some basic points of Zend Engine, with really good results. This shows that performance can be boosted by aspects other than JIT – which we have not decided to put in place for the moment.

We have noticed almost the same performances on WordPress between HHVM and phpng, and it’s a fact that HHVM is slightly faster, by 10%. On SugarCRM, phpng is three times faster than HHVM. Overall, phpng is twice as fast as HHVM. And even on PHP 5.6, which is before phpng, we’ve seen better performance for some applications. Some published benchmarks were not quite right, because they were using versions without OPcache. But like I said, performance isn’t everything.

Fabien Potencier estimates that at any given moment SensioLabs has been able to be more innovative than Zend. He especially cites projects like Composer or Twig (read the interview). What do you think of that ?

All innovations emerging in the PHP ecosystem are good for PHP. And if they are good for PHP, they are also good for Zend. Our PHP server, Zend Server, can take on any PHP application and infrastructure. We have Zend Server clients using Symfony or Zend Framework, but also CMS like Drupal or Laravel, or even their own PHP frameworks. We have to get away from this tactical exchange between Symfony and Zend Framework on the question of the innovation. These discussions are contributing nothing to PHP community. It’s a short-sighted outlook.

Our vision is clearly more strategic. We are willing to respond to questions our clients are putting in terms of time-to-market, productivity of developments, runtime engine… Our aim on both the open source and the commercial level is to create a change in paradigm in IT value. How do we ensure that PHP continues to be the leader in Web development, and in mobile as well? What is it that will grow PHP market share ? How will PHP let companies succeed ?Those are our real aims.

We also want to keep a step ahead in terms of innovation. The open source project Apigility we launched in late 2013 is a good illustration here. The idea is that web architectures are evolving from HTML orientation to API. We have clients in the media and health sectors who are starting to use them. The idea is to take leadership. We chose to put forward a disruptive product, a game-changer in direct competition with other languages. It’s a real change, and will go further than incremental improvements made in Zend Framework or Symfony.

On the question of communities, we’ve got nothing else to prove. Just look at all the projects we have launched and contributed to: PHP, Zend Framework, Eclipse PDT… We also contributed to MySQL. Consequently we have strong investment capacity in and with communities. And that’s a very important point for us. But the success of a project is measured above all by the level of contributions from the community. If you take the PHP or Zend Framework projects, the number of contributions they bring together are majorities outside the Zend teams. It’s a sign of success.

Continued deployment seems to be as one of your focus this year …

When we see that Amazon is delivering an evolution every 9 seconds on AWS, we understand the issue of companies today in terms of speed of deployment of applications. This is especially the case in the mobile app arena. We want to help companies respond to this challenge by giving them agile solutions, DevOps and optimisation of code quality. And do this as we respond to the question of scaling, and reach across API management.

You have signed agreements with the intention to take your technologies to the cloud. That’s the case with IBM, Amazon and Microsoft. Do you already have clients using this type of solution ?

Some clients have acquired Zend licences and are deploying our technologies on AWS instances. Others are starting to subscribe to the offer of payment at the capacity we launched with Amazon a year ago. We co-designed this program with Amazon. It lets smaller-sized companies take advantage of Zend Server at a lower cost, and pay by usage. It’s a very elastic solution. Zend Server lets you manage 200 server farms. We also signed an agreement with IBM, both on the IaaS aspect with Softlayer and also PaaS with Bluemix – built as CloudFoundry. We also approached Microsoft when Satya Nadella became head of Server & Tools and cloud Azure, and when he promoted IaaS on Azure. More and more, the Back End of mobile applications and API are being hosted on clouds.

Are you considering this link between cloud and mobile in your strategy ?

It is evident that this link does exist. And our product policy is going this way. That’s the aim of porting Zend Server and Zend Framework on these different clouds, but it’s also the aim of Apigility and recent evolutions made in Zend Studio in terms of development of mobile apps.But the main challenge of a mobile project is not on the client side. It’s in your capacity to build cloud services that enable contextualization, personalisation and geolocation, managing connections to the existing client information system. All done with adequate scaling. It’s what the value of your application will do. So the idea is to propose a scalable PHP infrastructure, in cloud mode, with the option of using applications in API form from multiple channels.

Original interview published on Journal du Net

The following two tabs change content below.
  • Evert

    Are there any numbers available to back up this claim: “Overall phpng is twice as fast as hhvm”.

    • Paul Tarjan

      +1. I’m incredibly surprised by that statement and would love to help Andi dig into this if he can give his numbers.

      The SugarCRM numbers were an incorrect benchmark not comparing apples to apples.

  • Zend

    This Benchmarking post talks about the kind of impact PHPNG has on PHP’s performance.