|By Peter Lubbers, Frank Greco||
|March 11, 2010 10:15 AM EST||
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.
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
HTTP/1.1 101 WebSocket Protocol Handshake\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.
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.
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
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:184.108.40.206) Gecko/20091102 Firefox/3.5.5
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
Server: Sun Java System Application Server 9.1_02
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.
- HTML5 Web Sockets Specification: http://dev.w3.org/html5/websockets/
- WebSockets.org: http://www.websockets.org
- Kaazing corporation: http://www.kaazing.com
- Download the Kaazing WebSocket Gateway: http://www.kaazing.com/download
- Skills Matter HTML5 Communication Training Courses: http://skillsmatter.com/expert-profile/java-jee/peter-lubbers
- Book: Pro HTML5 Programming: Powerful APIs for Richer Internet Application Development (Link)
|PeterLubbers 03/15/10 03:55:00 PM EDT|
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.
I hope that we will see WebSocket support in all production mainstream browsers very soon.
SYS-CON Events announced today that Isomorphic Software will exhibit at DevOps Summit at 19th International Cloud Expo, which will take place on November 1–3, 2016, at the Santa Clara Convention Center in Santa Clara, CA. Isomorphic Software provides the SmartClient HTML5/AJAX platform, the most advanced technology for building rich, cutting-edge enterprise web applications for desktop and mobile. SmartClient combines the productivity and performance of traditional desktop software with the simp...
Jul. 30, 2016 09:45 PM EDT Reads: 1,317
Sharding has become a popular means of achieving scalability in application architectures in which read/write data separation is not only possible, but desirable to achieve new heights of concurrency. The premise is that by splitting up read and write duties, it is possible to get better overall performance at the cost of a slight delay in consistency. That is, it takes a bit of time to replicate changes initiated by a "write" to the read-only master database. It's eventually consistent, and it'...
Jul. 30, 2016 08:45 PM EDT Reads: 2,400
Jul. 30, 2016 07:30 PM EDT Reads: 4,015
Internet of @ThingsExpo, taking place November 1-3, 2016, at the Santa Clara Convention Center in Santa Clara, CA, is co-located with the 19th International Cloud Expo and will feature technical sessions from a rock star conference faculty and the leading industry players in the world and ThingsExpo Silicon Valley Call for Papers is now open.
Jul. 30, 2016 07:00 PM EDT Reads: 2,775
There's a lot of things we do to improve the performance of web and mobile applications. We use caching. We use compression. We offload security (SSL and TLS) to a proxy with greater compute capacity. We apply image optimization and minification to content. We do all that because performance is king. Failure to perform can be, for many businesses, equivalent to an outage with increased abandonment rates and angry customers taking to the Internet to express their extreme displeasure.
Jul. 30, 2016 04:15 PM EDT Reads: 1,720
Ovum, a leading technology analyst firm, has published an in-depth report, Ovum Decision Matrix: Selecting a DevOps Release Management Solution, 2016–17. The report focuses on the automation aspects of DevOps, Release Management and compares solutions from the leading vendors.
Jul. 30, 2016 01:00 PM EDT Reads: 1,865
No matter how well-built your applications are, countless issues can cause performance problems, putting the platforms they are running on under scrutiny. If you've moved to Node.js to power your applications, you may be at risk of these issues calling your choice into question. How do you identify vulnerabilities and mitigate risk to take the focus off troubleshooting the technology and back where it belongs, on innovation? There is no doubt that Node.js is one of today's leading platforms of ...
Jul. 30, 2016 12:45 PM EDT Reads: 808
Adding public cloud resources to an existing application can be a daunting process. The tools that you currently use to manage the software and hardware outside the cloud aren’t always the best tools to efficiently grow into the cloud. All of the major configuration management tools have cloud orchestration plugins that can be leveraged, but there are also cloud-native tools that can dramatically improve the efficiency of managing your application lifecycle. In his session at 18th Cloud Expo, ...
Jul. 30, 2016 12:00 PM EDT Reads: 1,373
SYS-CON Events announced today that LeaseWeb USA, a cloud Infrastructure-as-a-Service (IaaS) provider, will exhibit at the 19th International Cloud Expo, which will take place on November 1–3, 2016, at the Santa Clara Convention Center in Santa Clara, CA. LeaseWeb is one of the world's largest hosting brands. The company helps customers define, develop and deploy IT infrastructure tailored to their exact business needs, by combining various kinds cloud solutions.
Jul. 30, 2016 11:30 AM EDT Reads: 1,416
SYS-CON Events announced today that Venafi, the Immune System for the Internet™ and the leading provider of Next Generation Trust Protection, will exhibit at @DevOpsSummit at 19th International Cloud Expo, which will take place on November 1–3, 2016, at the Santa Clara Convention Center in Santa Clara, CA. Venafi is the Immune System for the Internet™ that protects the foundation of all cybersecurity – cryptographic keys and digital certificates – so they can’t be misused by bad guys in attacks...
Jul. 30, 2016 10:15 AM EDT Reads: 1,526
DevOps at Cloud Expo – being held November 1-3, 2016, at the Santa Clara Convention Center in Santa Clara, CA – announces that its Call for Papers is open. Born out of proven success in agile development, cloud computing, and process automation, DevOps is a macro trend you cannot afford to miss. From showcase success stories from early adopters and web-scale businesses, DevOps is expanding to organizations of all sizes, including the world's largest enterprises – and delivering real results. Am...
Jul. 30, 2016 05:45 AM EDT Reads: 2,403
The 19th International Cloud Expo has announced that its Call for Papers is open. Cloud Expo, to be held November 1-3, 2016, at the Santa Clara Convention Center in Santa Clara, CA, brings together Cloud Computing, Big Data, Internet of Things, DevOps, Digital Transformation, Microservices and WebRTC to one location. With cloud computing driving a higher percentage of enterprise IT budgets every year, it becomes increasingly important to plant your flag in this fast-expanding business opportuni...
Jul. 30, 2016 05:15 AM EDT Reads: 2,732
Let's just nip the conflation of these terms in the bud, shall we?
"MIcro" is big these days. Both microservices and microsegmentation are having and will continue to have an impact on data center architecture, but not necessarily for the same reasons. There's a growing trend in which folks - particularly those with a network background - conflate the two and use them to mean the same thing.
They are not.
One is about the application. The other, the network. T...
Jul. 30, 2016 05:15 AM EDT Reads: 3,773
If you are within a stones throw of the DevOps marketplace you have undoubtably noticed the growing trend in Microservices. Whether you have been staying up to date with the latest articles and blogs or you just read the definition for the first time, these 5 Microservices Resources You Need In Your Life will guide you through the ins and outs of Microservices in today’s world.
Jul. 30, 2016 04:30 AM EDT Reads: 4,207
Before becoming a developer, I was in the high school band. I played several brass instruments - including French horn and cornet - as well as keyboards in the jazz stage band. A musician and a nerd, what can I say? I even dabbled in writing music for the band. Okay, mostly I wrote arrangements of pop music, so the band could keep the crowd entertained during Friday night football games. What struck me then was that, to write parts for all the instruments - brass, woodwind, percussion, even k...
Jul. 30, 2016 01:45 AM EDT Reads: 2,426
This digest provides an overview of good resources that are well worth reading. We’ll be updating this page as new content becomes available, so I suggest you bookmark it. Also, expect more digests to come on different topics that make all of our IT-hearts go boom!
Jul. 30, 2016 12:45 AM EDT Reads: 3,797
Keeping pace with advancements in software delivery processes and tooling is taxing even for the most proficient organizations. Point tools, platforms, open source and the increasing adoption of private and public cloud services requires strong engineering rigor – all in the face of developer demands to use the tools of choice. As Agile has settled in as a mainstream practice, now DevOps has emerged as the next wave to improve software delivery speed and output. To make DevOps work, organization...
Jul. 30, 2016 12:30 AM EDT Reads: 2,370
Right off the bat, Newman advises that we should "think of microservices as a specific approach for SOA in the same way that XP or Scrum are specific approaches for Agile Software development". These analogies are very interesting because my expectation was that microservices is a pattern. So I might infer that microservices is a set of process techniques as opposed to an architectural approach. Yet in the book, Newman clearly includes some elements of concept model and architecture as well as p...
Jul. 29, 2016 02:45 PM EDT Reads: 9,848
This is a no-hype, pragmatic post about why I think you should consider architecting your next project the way SOA and/or microservices suggest. No matter if it’s a greenfield approach or if you’re in dire need of refactoring. Please note: considering still keeps open the option of not taking that approach. After reading this, you will have a better idea about whether building multiple small components instead of a single, large component makes sense for your project. This post assumes that you...
Jul. 29, 2016 04:15 AM EDT Reads: 4,302
In his session at @DevOpsSummit at 19th Cloud Expo, Yoseph Reuveni, Director of Software Engineering at Jet.com, will discuss Jet.com's journey into containerizing Microsoft-based technologies like C# and F# into Docker. He will talk about lessons learned and challenges faced, the Mono framework tryout and how they deployed everything into Azure cloud. Yoseph Reuveni is a technology leader with unique experience developing and running high throughput (over 1M tps) distributed systems with extre...
Jul. 28, 2016 10:15 PM EDT Reads: 2,271