FastView acceleration overview

Editor's Notes

• #3: Web site and web application performance are key to any online business success. Take for example media sites:(1) Those are interested in maximizing their page views. A study from Abardeen group showed that a 1 second delay in page load time decreased the number of page views on a site by 11% in average.(2) Any enterprise which provides services online – like Bing – MS’s search engine as an example. They have conducted a study over 6 weeks period and showed that customer satisfaction dropped deeper, the higher the page load time increased, up to almost 4%, with 2 seconds delay.The interesting part in that study was that those numbers increased as the study continued, meaning the effect of slower sites is accumulative. - And the most important KPI of all – revenues per user – in the same study, they saw that revenues per user also dropped as page load time increased – up to 4.3% per 2 seconds higher delay.(3) Amazon sow that reducing their site’s response time by 100ms increased their revenues by 1%. Hotmail had the oposite experience, where a service slowdown of 6 second resulted in 40M drop in ad impression per year, resulting in a $6M revenue drop per year from online ads.I could go on with examples from smaller and bigger online businesses and portals – all getting to the same conclusion:(4) There is a clear correlation between Website performance and online revenues, number of page views and customer satisfaction / page abandonment. This means that any online customer facing business has clear motivation to accelerate his website / web application performance, and monitor it to ensure it meets his business goals
• #4: Radware understands the importance of performance to its customer and developed FastView – a set of result drive acceleration capabilities, designed to maximize business impact for website and online web applications:Accelerate response time to maximize conversion rate, increase page views and improve customer satisfaction.(2) Understanding that end-users come today in various forms and shape, FastView was designed so that all acceleration capabilities will address ALL clients, whether desktop, laptop, tablet or mobile, regardless of which type of browser they are using.FastView is also focused at providing acceleration for users from the first time they access the site as well as improve response time for repeat views – which are basically repeat customers.(3) Being result oriented, we understand our customers must be able to measure and monitor their application’s performance as well as their actual users’ experience. This is why we are launching together with FastView an application performance monitoring (APM) tool which both gives intuitive visibility into users’ QoE and a powerful tool for troubleshooting application slowdowns & errors.
• #5: So what are the challenges that affect application performance?The first one is WAN latency. The time it takes for a request to reach the server, and for the reply to travle back to the user. This time is strongly impacted also by the protocols which carry those requests and responses like TCP and HTTP, which add their own overheads, and are far from being optimal for all of the WAN connection types being used by end users.The second challenge is the growing weight of web pages. Both the raw weignt in Killo Bytes per page and in the number of objects each page contain. Did you know for example, that just recently, the average home page weight of the 200,000 leading websites has just crossed 1MB? And the trend is still going upper (or heavier).Page rendering time is yet another challenge affecting QoE and overall page load time. There is more content per page which takes time to fetch and render, and which is not necesserally visible to the user, like JS like the Google Analytic JS – it is important to the site owner, but slowsdown the page load time and display, since the browser needs to process it as part of the page. In general, today, there is more complex content for browsers to render, which slows down page load time.
• #6: Traditionally, Radware’s ADC addressed the three aspects of User QoE – which include Security, Availability and Performance. (1) Now we are extending the performance acceleration capabilities of our Alteon ADC with FastView, delivering faster web application response time for any user (2)(2) Any user means that the acceleration is applicable to any client type, whether it is a windows / Linux based desktop or an IOS/Android typ mobile.It also means it is applicable for all types of websites, regardless of the CMS they are using.(3) It is available for any user – anywhere, as it is a datacenter centicasymetric solution – it doesn’t require the user to install anything on his client to enjoy this acceleration.(4) Anytime means acceleration capabilities are applicable for both pages and objects which are accessed for the first time – first views, as well as in repeated views – both will be loaded by the client faster(5) FastView is very simple to implement. Enjoying the acceleration is achiveable at a push of a button, no need to configure and teach the ADC how to achieve faster performance – it’s a simple and fast implementation, meaning fast time to market, and fast RoI.
• #7: We learn from the experts. Web Performance Optimization is a hot topic in the industry today. (Later this month there’s the Velocity show around solutions for Web acceleration). Steve Sounders, a world recognized specialist has published a book with 14 golden rules for accelrating website performance. Today, these rules are practically industry standard.FastView delivers and implements 8 of those 14 rules – through minimal investment and zero ongoing effort.Lets dive into some high level technology examples on what we’re doing.FastView delivers faster transport between the user and the web application server, by first minimizing the number of requests per page. I’ll show more details and examples to that in the next page.Just as important, we perform TCP protocol optimization, which also reduces the timerequired to complete object transfer between the server and the user.(2) The second part is reducing the amount of information transferred from the server to the user, either by smart dynamic caching, compression or other techniques of information minification.
• #8: HTTP Archive is a site that monitors over 200K sites / URL twice a month, recording many characteristics and trends of how content is evolvong. Lets take the page size trend as an example. There’s a clear ongoing trend of page weight growth, per HTTP archive which just recently has crossed the 1Mb bar. Another interesting trend is the number of objects constracting a web page, which is also on the rise and is now standing on about 80.Out of those 80, 19 of those are of JS and CSSs.FastView provides CSS andJS combining, which allows to transfer all JSs and CSSs as one big object instead of separate objects.The result is a 20-30% reduction in the number of requests per page,Saving about 17 times the round trip delay between the user and the server – which can amount to a significant faster page load time.
• #9: Here is another interesting fact from HTTP archive: 58% of the cacheable content on the internet is not flagged properly for effective caching. So a lot of the content needs to be accessed by the user in the web server origin (even CDNs won’t be effective here).FastView provides a smart Dynamic Caching engine, which analyzes the content changes to decide whether to cache it or not, regardless of the caching flags in the contet’s header.By using smart caching, we can first cache the content in the ADC, so that access to those objects will be served directly by the ADC, saving the time it takes the server to create a connection fetch and deliver the object – applicable even for content accessed for the first time (first view), after the first user has accessed it of course…Those flagged also trigger in browser caching, allowing the user to cache those objects in his local browser cach, resulting in a much faster object load time. This is applicable per object accessed in repeat view, even if the page requesting it is accessed for the first time.Whether served from local ADC or Browser cache, Smart Caching significantly offloads the servers, without changing a single thing in the application of the caching flafs of the objects it uses.
• #10: So we can accelerate response time. But our customers want also to see how their application actually perform. The want to ensure their application performance can meet their business goals. A business goal can be an SLA their end users get – the QoE they get on various transactions. And of course to ensure your application is performing you need also a tool that will allow you to be proactive in maitainingperofrmance and track down root cause for application slowdowns.Our APM tool can actually extract information on end-user QoE from the client itself – wherever he is.The information is collected in distributed manner from all client, all ADCs, and all applications we want to monitor, whether in the same DC or across DC or in the cloud.This information is then sent to an analytic engine which produces reports that are readable to the user, and reflect the actual user experience and the actual application performance in real time – with a bird’s eye view which covers all applications, transactions, users, datacenters, and can dreal down to the user level or even application transaction level.Lets see how it looks.
• #11: (1) There’s the historical view which gives you average response time along the day. This can also be displayed in correlation with transaction volume, to see how the infrastructure can handle load, and whether it meets our business goals in terms of how much users we want to serve concurrently.The response time history is provided with a breakdown of where in the application delivery chain most load time was spent, at any point of time. How much time was spent on tendering, network or in the DC. This can also give an idea of which type of optimization would have the greater affect.One can also drill down and see those historical measurement at the application level or even at the transaction type or URL level. That way, each application Admin can get his own personalized view and anlyze information relevant to their specific application.
• #12: The APM reports provided by the Sharepath analytical engine also provide an SLA view. This SLA is user defined and may include the minimum avg. response time an application should provide, the % of errors user experience and much more.(2) SLA can also be borken down to the geo location level, so administrators can understand if there’s a problem with users from a specific region, and maybe they need to add a specific application to another DC to serve that region better.(3) It is also very easy to track down which app/region or even specific transaction is out of SLA target, showing clearly both at the table report or the graphical target board, with the out of SLA item name.(4) The SLA report also provides information about the amount of errors in the application per all of the report level mentioned. This is important to both be proactive about errors that occur in real time, and also to drill down and find the application – transaction that caused the errors.(5) Needless to say all of the measurements are based on real-life traffic from actual user transactions and not from synthetic scripts running from predefined locations. It represent the actual QoE the users are experiencing and not assessments, and all this without loading the website infrastructure with synthetic traffic.Radware’s APM is an advanced tool for monitoring application performance and user QoE, providing easy to read reports which lets organizations proactively react to any change in application’s SLA levels – keeping their application performance optimized at all times.
• #13: To Summarize: with the understanding how web application performance affects online businesses, Radware FastView – a result driven acceleration technology embedded in all or our Alteon ADCs, designed to deliver maximized business impact. (2) It means all acceleration capabilities we deliver are applicable for all types users, clients and browsers types, as well as relevant for all types of CMSs. We understand that users can connect to their online services / applications from anywhere with different types of devices at different times and acceleration should be consistent across the various user combinations.(3) To measure application performance and how fast it becomes, we also provide a new APM tool which enables full visibility on application performance and user QoE, through simple and intuitive reports. This is a powerful and valuable tool which enables real time and proactive approach to maintaining application performance optimized at all times.(4) Both FV and the APM solution are simple to implement enabling fast time to market of the value they bring. This is what “Result Driven Acceleration” is all about – because every second count!