Microservices Expo Authors: Mehdi Daoudi, Pat Romanski, Elizabeth White, Steve Wilson, Stackify Blog

Related Topics: Microservices Expo, Java IoT, Machine Learning

Microservices Expo: Article

HTML5 Web Sockets: A Quantum Leap in Scalability for the Web

An enormous step forward in the scalability of the real-time web

Lately there has been a lot of buzz around HTML5 Web Sockets, which defines a full-duplex communication channel that operates through a single socket over the Web. HTML5 Web Sockets is not just another incremental enhancement to conventional HTTP communications; it represents a colossal advance, especially for real-time, event-driven web applications.

HTML5 Web Sockets provides such a dramatic improvement from the old, convoluted "hacks" that are used to simulate a full-duplex connection in a browser that it prompted Google's Ian Hickson - the HTML5 specification lead - to say:

"Reducing kilobytes of data to 2 bytes...and reducing latency from 150ms to 50ms is far more than marginal. In fact, these two factors alone are enough to make Web Sockets seriously interesting to Google."

Let's look at how HTML5 Web Sockets can offer such an incredibly dramatic reduction of unnecessary network traffic and latency by comparing it to conventional solutions.

Polling, Long-Polling, and Streaming - Headache 2.0
Normally when a browser visits a web page, an HTTP request is sent to the web server that hosts that page. The web server acknowledges this request and sends back the response. In many cases - for example, for stock prices, news reports, ticket sales, traffic patterns, medical device readings, and so on - the response could be stale by the time the browser renders the page. If you want to get the most up-to-date "real-time" information, you can constantly refresh that page manually, but that's obviously not a great solution.

Current attempts to provide real-time web applications largely revolve around polling and other server-side push technologies, the most notable of which is Comet, which delays the completion of an HTTP response to deliver messages to the client. Comet-based push is generally implemented in JavaScript and uses connection strategies such as long-polling or streaming.

With polling, the browser sends HTTP requests at regular intervals and immediately receives a response. This technique was the first attempt for the browser to deliver real-time information. Obviously, this is a good solution if the exact interval of message delivery is known, because you can synchronize the client request to occur only when information is available on the server. However, real-time data is often not that predictable, making unnecessary requests inevitable and, as a result, many connections are opened and closed needlessly in low-message-rate situations.

With long-polling, the browser sends a request to the server and the server keeps the request open for a set period. If a notification is received within that period, a response containing the message is sent to the client. If a notification is not received within the set time period, the server sends a response to terminate the open request. It is important to understand, however, that when you have a high message volume, long-polling does not provide any substantial performance improvements over traditional polling. In fact, it could be worse, because the long-polling might spin out of control into an unthrottled, continuous loop of immediate polls.

With streaming, the browser sends a complete request, but the server sends and maintains an open response that is continuously updated and kept open indefinitely (or for a set period of time). The response is then updated whenever a message is ready to be sent, but the server never signals to complete the response, thus keeping the connection open to deliver future messages. However, since streaming is still encapsulated in HTTP, intervening firewalls and proxy servers may choose to buffer the response, increasing the latency of the message delivery. Therefore, many streaming Comet solutions fall back to long-polling in case a buffering proxy server is detected. Alternatively, TLS (SSL) connections can be used to shield the response from being buffered, but in that case the setup and tear down of each connection taxes the available server resources more heavily.

Ultimately, all of these methods for providing real-time data involve HTTP request and response headers, which contain lots of additional, unnecessary header data and introduce latency. On top of that, full-duplex connectivity requires more than just the downstream connection from server to client. In an effort to simulate full-duplex communication over half-duplex HTTP, many of today's solutions use two connections: one for the downstream and one for the upstream. The maintenance and coordination of these two connections introduces significant overhead in terms of resource consumption and adds lots of complexity. Simply put, HTTP wasn't designed for real-time, full-duplex communication as you can see in Figure 1, which shows the complexities associated with building a Comet web application that displays real-time data from a back-end data source using a publish/subscribe model over half-duplex HTTP.

Figure 1: The complexity of Comet applications

It gets even worse when you try to scale out those Comet solutions to the masses. Simulating bi-directional browser communication over HTTP is error-prone and complex and all that complexity does not scale. Even though your end users might be enjoying something that looks like a real-time web application, this "real-time" experience has an outrageously high price tag. It's a price that you will pay in additional latency, unnecessary network traffic, and a drag on CPU performance.

HTML5 Web Sockets to the Rescue!
Defined in the Communications section of the HTML5 specification, HTML5 Web Sockets represent the next evolution of web communications - a full-duplex, bidirectional communications channel that operates through a single socket over the Web. HTML5 Web Sockets provides a true standard that you can use to build scalable, real-time web applications. In addition, since it provides a socket that is native to the browser, it eliminates many of the problems Comet solutions are prone to. Web Sockets remove the overhead and dramatically reduce complexity.

To establish a WebSocket connection, the client and server upgrade from the HTTP protocol to the WebSocket protocol during their initial handshake, as shown in the following example:

Example 1: The WebSocket handshake (browser request and server response)

GET /text HTTP/1.1\r\n
Upgrade: WebSocket\r\n
Connection: Upgrade\r\n
Host: www.websocket.org\r\n

HTTP/1.1 101 WebSocket Protocol Handshake\r\n
Upgrade: WebSocket\r\n
Connection: Upgrade\r\n

Once established, WebSocket data frames can be sent back and forth between the client and the server in full-duplex mode. Both text and binary frames can be sent full-duplex, in either direction at the same time. The data is minimally framed with just two bytes. In the case of text frames, each frame starts with a 0x00 byte, ends with a 0xFF byte, and contains UTF-8 data in between. WebSocket text frames use a terminator, while binary frames use a length prefix.

Note: Although the Web Sockets protocol is ready to support a diverse set of clients, it cannot deliver raw binary data to JavaScript, because JavaScript does not support a byte type. Therefore, binary data is ignored if the client is JavaScript, but it can be delivered to other clients that support it.

The Showdown: Comet vs HTML5 Web Sockets
How dramatic is that reduction in unnecessary network traffic and latency? Let's compare a polling application with a WebSocket application.

For the polling example, I created a simple web application in which a web page requests real-time stock data from a RabbitMQ message broker using a traditional publish/subscribe model. It does this by polling a Java servlet that is hosted on a web server. The RabbitMQ message broker receives data from a fictitious stock price feed with continuously updating prices. The web page connects and subscribes to a specific stock channel (a topic on the message broker) and uses an XMLHttpRequest to poll for updates once per second. When updates are received, some calculations are performed and the stock data is shown in a table as shown in Figure 2.

Figure 2: A JavaScript stock ticker application

Note: The back-end stock feed actually produces a lot of stock price updates per second, so using polling at one-second intervals is actually more prudent than using a Comet long-polling solution, which would result in a series of continuous polls. Polling effectively throttles the incoming updates here.

It all looks great, but a look under the hood reveals there are some serious issues with this application. For example, in Mozilla Firefox with Firebug (a Firefox add-on that allows you to debug web pages and monitor the time it takes to load pages and execute scripts), you can see that GET requests hammer the server at one-second intervals. Turning on Live HTTP Headers (another Firefox add-on that shows live HTTP header traffic) reveals the shocking amount of header overhead that is associated with each request. The following two examples show the HTTP header data for just a single request and response:

Example 2: HTTP request header

GET /PollingStock//PollingStock HTTP/1.1
Host: localhost:8080
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv: Gecko/20091102 Firefox/3.5.5
Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
Accept-Language: en-us
Accept-Encoding: gzip,deflate
Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
Keep-Alive: 300
Connection: keep-alive
Referer: http://www.example.com/PollingStock/
Cookie: showInheritedConstant=false; showInheritedProtectedConstant=false; showInheritedProperty=false; showInheritedProtectedProperty=false; showInheritedMethod=false; showInheritedProtectedMethod=false; showInheritedEvent=false; showInheritedStyle=false; showInheritedEffect=false

Example 3: HTTP response header

HTTP/1.x 200 OK
X-Powered-By: Servlet/2.5
Server: Sun Java System Application Server 9.1_02
Content-Type: text/html;charset=UTF-8
Content-Length: 21
Date: Sat, 07 Nov 2009 00:32:46 GMT

Just for fun, I counted all the characters. The total HTTP request and response header information overhead contains 871 bytes and that doesn't even include any data. Of course, this is just an example and you can have less than 871 bytes of header data, but I have also seen cases where the header data exceeded 2,000 bytes. In this example application, the data for a typical stock topic message is only about 20 characters long. As you can see, it's effectively drowned out by the excessive header information, which wasn't even required in the first place.

What happens when you deploy this application to a large number of users? Let's take a look at the network throughput for just the HTTP request and response header data associated with this polling application in three different use cases.

  • Use case A: 1,000 clients polling every second:
    Network throughput is (871 x 1,000) = 871,000 bytes = 6,968,000 bits per second (6.6 Mbps)
  • Use case B: 10,000 clients polling every second:
    Network throughput is (871 x 10,000) = 8,710,000 bytes = 69,680,000 bits per second (66 Mbps)
  • Use case C: 100,000 clients polling every 1 second:
    Network throughput is (871 x 100,000) = 87,100,000 bytes = 696,800,000 bits per second (665 Mbps)

That's an enormous amount of unnecessary network throughput. If only we could just get the essential data over the wire. Well, guess what? You can with HTML5 Web Sockets. I rebuilt the application to use HTML5 Web Sockets, adding an event handler to the web page to asynchronously listen for stock update messages from the message broker (check out the many how-tos and tutorials on www.tech.kaazing.com/documentation for more information on how to build a WebSocket application). Each of these messages is a WebSocket frame that has just two bytes of overhead (instead of 871). Take a look at how that affects the network throughput overhead in our three use cases.

  • Use case A: 1,000 clients receive 1 message per second:
    Network throughput is (2 x 1,000) = 2,000 bytes = 16,000 bits per second (0.015 Mbps)
  • Use case B: 10,000 clients receive 1 message per second:
    Network throughput is (2 x 10,000) = 20,000 bytes = 160,000 bits per second (0.153 Mbps)
  • Use case C: 100,000 clients receive 1 message per second:
    Network throughput is (2 x 100,000) = 200,000 bytes = 1,600,000 bits per second (1.526 Mbps)

As you can see in Figure 3, HTML5 Web Sockets provide a dramatic reduction of unnecessary network traffic compared to the polling solution.

Figure 3: Comparison of the unnecessary network throughput overhead between the polling and the WebSocket applications

What about the reduction in latency? Take a look at Figure 4. In the top half, you can see the latency of the half-duplex polling solution. If we assume, for this example, that it takes 50 milliseconds for a message to travel from the server to the browser, then the polling application introduces a lot of extra latency, because a new request has to be sent to the server when the response is complete. This new request takes another 50ms and during this time the server cannot send any messages to the browser, resulting in additional server memory consumption.

In the bottom half of the figure, you see the reduction in latency provided by the WebSocket solution. Once the connection is upgraded to WebSocket, messages can flow from the server to the browser the moment they arrive. It still takes 50 ms for messages to travel from the server to the browser, but the WebSocket connection remains open so there is no need to send another request to the server.

Figure 4: Latency comparison between the polling and WebSocket applications

HTML5 Web Sockets and the Kaazing WebSocket Gateway
Today, only Google's Chrome browser supports HTML5 Web Sockets natively, but other browsers will soon follow. To work around that limitation, however, Kaazing WebSocket Gateway provides complete WebSocket emulation for all the older browsers (I.E. 5.5+, Firefox 1.5+, Safari 3.0+, and Opera 9.5+), so you can start using the HTML5 WebSocket APIs today.

WebSocket is great, but what you can do once you have a full-duplex socket connection available in your browser is even greater. To leverage the full power of HTML5 Web Sockets, Kaazing provides a ByteSocket library for binary communication and higher-level libraries for protocols like Stomp, AMQP, XMPP, IRC and more, built on top of WebSocket.

For example, if you use a high-level library for protocols such as Stomp or AMQP (supplied with the Kaazing Gateway), you can talk directly to back-end message brokers like the RabbitMQ broker described in the example. By having a direct connection to services, there is no need for additional application server logic to translate those bi-directional, full-duplex TCP backend protocols to uni-directional, half-duplex HTTP connections; the browser can simply understand these protocols natively.

Figure 5: Kaazing WebSocket Gateway extends TCP-based messaging to the browser with ultra high performance

HTML5 Web Sockets provides an enormous step forward in the scalability of the real-time web. As you have seen in this article, HTML5 Web Sockets can provide a 500:1 or - depending on the size of the HTTP headers - even a 1000:1 reduction in unnecessary HTTP header traffic and 3:1 reduction in latency. That is not just an incremental improvement; that is a revolutionary jump - a quantum leap.

Kaazing WebSocket Gateway makes HTML5 WebSocket code work in all the browsers today, while providing additional protocol libraries that allow you to harness the full power of the full-duplex socket connection that HTML5 Web Sockets provides and communicate directly to back-end services. For more information about Kaazing WebSocket Gateway, visit www.kaazing.com and the Kaazing technology network at tech.kaazing.com.


More Stories By Peter Lubbers

Peter Lubbers is the Director of Documentation and Training at Kaazing where he oversees all aspects of documentation and training. He is the co-author of the Apress book Pro HTML5 Programming and teaches HTML5 training courses. An HTML5 and WebSocket enthusiast, Peter frequently speaks at international events.

Prior to joining Kaazing, Peter worked as an information architect at Oracle, where he wrote many books. He also develops documentation automation solutions and two of his inventions are patented.

A native of the Netherlands, Peter served as a Special Forces commando in the Royal Dutch Green Berets. In his spare time (ha!) Peter likes to run ultra-marathons. He is the 2007 and 2009 ultrarunner.net series champion and three-time winner of the Tahoe Super Triple marathon. Peter lives on the edge of the Tahoe National Forest and loves to run in the Sierra Nevada foothills and around Lake Tahoe (preferably in one go!).

More Stories By Frank Greco

Frank Greco is the founder of the New York Java Special Interest Group (NYJavaSIG), one of the largest Java Users Groups (JUGs) on the planet with over 8,000 active members in the local Java community. The NYJavaSIG has had some of the most famous Java luminaries speak at their meetings, including Java Champions Brian Goetz, Rod Johnson, Doug Lea and Josh Bloch. Their members are very enthusiastic and, like most New Yorkers, don't hesitate to ask tough questions of their monthly speakers.

Frank has a long history as a "Champion" of the Java Platform; he taught a developer track session at the very first Java Day back in September 1995 in New York and started the NYJavaSIG that afternoon. Frank has been involved with software development for over 10 years and has worked on sophisticated architectures, innovative user interfaces, mobile computing and next-generation collaborative financial systems. Frank is both a Java community leader as well as a luminary technologist specializing in cloud services and enterprise architectures.

Comments (2) View Comments

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.

Most Recent Comments
PeterLubbers 03/15/10 03:55:00 PM EDT

Thanks Tam!
Sorry about that error in use case B and C. You're right it should be in Mbps. I'll try to get the article updated. Thanks for pointing that out.

An additional problem with streaming is that you can get unexpected results if there are intermediaries such as buffering proxy servers in the way. Streaming solutions therefore often fall back to long-polling if they detect such intermediaries (if they are smart enough to detect them at all). With WebSocket and WebSocket secure you have a better chance of traversing proxy servers and firewalls.

tbayes 03/11/10 10:11:00 AM EST

Hi Peter and Frank,

Thank you for this clear and well-written article, which does much to effectively explain the benefits of the WebSocket protocol.

If I may point out one minor correction: In the WebSocket section of Example 3, Use case B and C should show overhead rates of 0.153Mbps and 1.526Mbps respectively, as opposed to 0.153Kbps and 1.526Kbps.

Perhaps a more natural comparison would be between Web Sockets and Comet-streaming (via, for example, the "Forever IFrame" technique), as I feel it a little unsporting to compare Web Sockets directly to polling or long-polling! At around 20 bytes per message, the overhead associated with clients receiving messages via Comet-style streaming is still more than WebSocket's 2 bytes per message, but is still one to two orders of magnitude less than the overhead associated with polling techniques. Similarly, latency whether receiving messages using Comet-streaming or WebSocket is the same. Of course, clients sending messages will gain the advantages you describe only with Web Sockets, as all other approaches do indeed require new HTTP requests.

I completely agree, however, that WebSocket is a definite improvement over the Comet IFrame-approaches, which have an air of string and glue about them.

You mention your usage of various JavaScript APIs built on top of Web Sockets - such as your AmqpClient, or StompClient. This is understandable, since while the WebSocket API is commendably simple, it lacks many useful features. These APIs provide additional functionality, and do not explicitly expose the raw, underlying WebSocket API. We agree with this approach, and having implemented WebSocket interfaces in our Nirvana messaging server, we have opted to retain an existing, feature-rich and robust messaging JavaScript API unchanged except for its support for underlying WebSocket communication. See this link for more details.

I hope that we will see WebSocket support in all production mainstream browsers very soon.


@MicroservicesExpo Stories
In his session at 20th Cloud Expo, Mike Johnston, an infrastructure engineer at Supergiant.io, discussed how to use Kubernetes to set up a SaaS infrastructure for your business. Mike Johnston is an infrastructure engineer at Supergiant.io with over 12 years of experience designing, deploying, and maintaining server and workstation infrastructure at all scales. He has experience with brick and mortar data centers as well as cloud providers like Digital Ocean, Amazon Web Services, and Rackspace. H...
When you focus on a journey from up-close, you look at your own technical and cultural history and how you changed it for the benefit of the customer. This was our starting point: too many integration issues, 13 SWP days and very long cycles. It was evident that in this fast-paced industry we could no longer afford this reality. We needed something that would take us beyond reducing the development lifecycles, CI and Agile methodologies. We made a fundamental difference, even changed our culture...
As many know, the first generation of Cloud Management Platform (CMP) solutions were designed for managing virtual infrastructure (IaaS) and traditional applications. But that’s no longer enough to satisfy evolving and complex business requirements. In his session at 21st Cloud Expo, Scott Davis, Embotics CTO, will explore how next-generation CMPs ensure organizations can manage cloud-native and microservice-based application architectures, while also facilitating agile DevOps methodology. He wi...
Docker is sweeping across startups and enterprises alike, changing the way we build and ship applications. It's the most prominent and widely known software container platform, and it's particularly useful for eliminating common challenges when collaborating on code (like the "it works on my machine" phenomenon that most devs know all too well). With Docker, you can run and manage apps side-by-side - in isolated containers - resulting in better compute density. It's something that many developer...
These days, change is the only constant. In order to adapt and thrive in an ever-advancing and sometimes chaotic workforce, companies must leverage intelligent tools to streamline operations. While we're only at the dawn of machine intelligence, using a workflow manager will benefit your company in both the short and long term. Think: reduced errors, improved efficiency and more empowered employees-and that's just the start. Here are five other reasons workflow automation is leading a revolution...
We have Continuous Integration and we have Continuous Deployment, but what’s continuous across all of what we do is people. Even when tasks are automated, someone wrote the automation. So, Jayne Groll evangelizes about Continuous Everyone. Jayne is the CEO of the DevOps Institute and the author of Agile Service Management Guide. She talked about Continuous Everyone at the 2016 All Day DevOps conference. She describes it as "about people, culture, and collaboration mapped into your value streams....
Cloud adoption is often driven by a desire to increase efficiency, boost agility and save money. All too often, however, the reality involves unpredictable cost spikes and lack of oversight due to resource limitations. In his session at 20th Cloud Expo, Joe Kinsella, CTO and Founder of CloudHealth Technologies, tackled the question: “How do you build a fully optimized cloud?” He will examine: Why TCO is critical to achieving cloud success – and why attendees should be thinking holistically ab...
Docker is on a roll. In the last few years, this container management service has become immensely popular in development, especially given the great fit with agile-based projects and continuous delivery. In this article, I want to take a brief look at how you can use Docker to accelerate and streamline the software development lifecycle (SDLC) process.
We define Hybrid IT as a management approach in which organizations create a workload-centric and value-driven integrated technology stack that may include legacy infrastructure, web-scale architectures, private cloud implementations along with public cloud platforms ranging from Infrastructure-as-a-Service to Software-as-a-Service.
Did you know that you can develop for mainframes in Java? Or that the testing and deployment can be automated across mobile to mainframe? In his session and demo at @DevOpsSummit at 21st Cloud Expo, Dana Boudreau, a Senior Director at CA Technologies, will discuss how increasingly teams are developing with agile methodologies, using modern development environments, and automating testing and deployments, mobile to mainframe.
As DevOps methodologies expand their reach across the enterprise, organizations face the daunting challenge of adapting related cloud strategies to ensure optimal alignment, from managing complexity to ensuring proper governance. How can culture, automation, legacy apps and even budget be reexamined to enable this ongoing shift within the modern software factory?
While some vendors scramble to create and sell you a fancy solution for monitoring your spanking new Amazon Lambdas, hear how you can do it on the cheap using just built-in Java APIs yourself. By exploiting a little-known fact that Lambdas aren’t exactly single-threaded, you can effectively identify hot spots in your serverless code. In his session at @DevOpsSummit at 21st Cloud Expo, Dave Martin, Product owner at CA Technologies, will give a live demonstration and code walkthrough, showing how ...
There are several reasons why businesses migrate their operations to the cloud. Scalability and price are among the most important factors determining this transition. Unlike legacy systems, cloud based businesses can scale on demand. The database and applications in the cloud are not rendered simply from one server located in your headquarters, but is instead distributed across several servers across the world. Such CDNs also bring about greater control in times of uncertainty. A database hack ...
@DevOpsSummit at Cloud Expo taking place Oct 31 - Nov 2, 2017, at the Santa Clara Convention Center, Santa Clara, CA, is co-located with the 21st International Cloud Expo and will feature technical sessions from a rock star conference faculty and the leading industry players in the world. The widespread success of cloud computing is driving the DevOps revolution in enterprise IT. Now as never before, development teams must communicate and collaborate in a dynamic, 24/7/365 environment. There is ...
API Security is complex! Vendors like Forum Systems, IBM, CA and Axway have invested almost 2 decades of engineering effort and significant capital in building API Security stacks to lockdown APIs. The API Security stack diagram shown below is a building block for rapidly locking down APIs. The four fundamental pillars of API Security - SSL, Identity, Content Validation and deployment architecture - are discussed in detail below.
In his session at 20th Cloud Expo, Scott Davis, CTO of Embotics, discussed how automation can provide the dynamic management required to cost-effectively deliver microservices and container solutions at scale. He also discussed how flexible automation is the key to effectively bridging and seamlessly coordinating both IT and developer needs for component orchestration across disparate clouds – an increasingly important requirement at today’s multi-cloud enterprise.
With Cloud Foundry you can easily deploy and use apps utilizing websocket technology, but not everybody realizes that scaling them out is not that trivial. In his session at 21st Cloud Expo, Roman Swoszowski, CTO and VP, Cloud Foundry Services, at Grape Up, will show you an example of how to deal with this issue. He will demonstrate a cloud-native Spring Boot app running in Cloud Foundry and communicating with clients over websocket protocol that can be easily scaled horizontally and coordinate...
DevOps at Cloud Expo, taking place October 31 - November 2, 2017, at the Santa Clara Convention Center in Santa Clara, CA, is co-located with 21st Cloud Expo and will feature technical sessions from a rock star conference faculty and the leading industry players in the world. The widespread success of cloud computing is driving the DevOps revolution in enterprise IT. Now as never before, development teams must communicate and collaborate in a dynamic, 24/7/365 environment. There is no time to w...
IT organizations are moving to the cloud in hopes to approve efficiency, increase agility and save money. Migrating workloads might seem like a simple task, but what many businesses don’t realize is that application migration criteria differs across organizations, making it difficult for architects to arrive at an accurate TCO number. In his session at 21st Cloud Expo, Joe Kinsella, CTO of CloudHealth Technologies, will offer a systematic approach to understanding the TCO of a cloud application...
Most companies are adopting or evaluating container technology - Docker in particular - to speed up application deployment, drive down cost, ease management and make application delivery more flexible overall. As with most new architectures, this dream takes a lot of work to become a reality. Even when you do get your application componentized enough and packaged properly, there are still challenges for DevOps teams to making the shift to continuous delivery and achieving that reduction in cost ...