 Hello everyone and welcome back to the state of the web. My guest today is Paul Calvano He's a senior web performance architect at Akamai and we're going to be talking about web transparency In previous episodes we looked at tools like the HTTP archive and Chrome UX report and Paul is here to tell us about How he leverages these tools to not just understand the web, but actually make it better. Let's get started So Paul, thank you for being here. Thanks, Rick. It's great to be here Tell us about what you do at Akamai. Sure. So at Akamai, I work with a lot of our customers on some of their strategic performance initiatives Often I'm brought in to help do a performance assessment of a customer's website Try to help them understand where where maybe they're They're experiencing some of their performance delays what we can do to help them What maybe their developers can do to optimize optimize things as well and Akamai has a two different Product lines major product lines. Can you tell us about those? We have a couple of different a couple of different offerings. So Akamai is most people know us as As is the largest content delivery network and it certainly is we had we deliver about 20 to 30 percent of the world's web traffic It's a pretty massive network 240,000 servers and 2500 locations But we consider ourselves to be a cloud delivery platform So we have we have we have the delivery aspect of it where we're delivering We're basically delivering content across the across the web to end users But then we also have the ability to accelerate that content using a variety of different performance optimization strategies and then secure that traffic by by you know implementing Like WAF and security filtering technologies as well as bot mitigation technologies and then beyond that We also have some some value-added services one of them is is impulse Which is a real user measurement service that that we use to Essentially track the performance of actual end users across across sites now as one of HHP archives biggest power users You can get some interesting ways of mashing up both real user data with synthetic Can you tell us about a couple of ways that you've done that sure? So one of the first ways I've I went in and used HGP archive data When I when I first learned about the ability to query it on big queries. I was wondering We have like like what types of compression like liquid what types of content are people compressing on the web So we have a feature on Akamai called last mile acceleration and this feature jizip compresses resources for customers if they did not already jizip compress them it's kind of like a failsafe for customers that didn't compress content and We historically have done it based on based on a couple of different content types CSS text slash HTML and An application slash x dash JavaScript when I saw this Amazing amount of data that we had available to us. I was wondering what can I you know? what other types of content are being compressed and When I was able to run that query the results were a lot more than what we were actually using as our defaults So we went up expanding the defaults in our property manager Behavior so that when a user implements last mile acceleration They are they automatically start off with this this longer list of defaults based on what we what we captured in the HGP archive Gotcha, so using HGP archive. You can see what should be compressed correct and then from Akamai side You can ensure that those types of resources are compressed. Yep. Absolutely now You've also done things with security right a couple of months ago Doug Sillers wrote a really great post on the HGP archive discussion forums about Cryptocurrency mining and its performance impact on the web and if you visit a website that has a cryptocurrency minor You'll definitely notice that not only does the performance of that page drag But your entire computer just starts slowing down to a crawl because it's it's just using as much of your CPU as possible So when I saw that we started working on I know that you and I and Doug started working on Basically building out that that query so that we were able to detect a variety of different cryptocurrency miners What I've been doing since then is I'll actually run that query with every new HGP archive release But look at look specifically for Akamai customers that have cryptocurrency miners enabled on their site A lot of times I'm finding like one or two sites that have a cryptocurrency minor And then I go and I reach out to them and I let them know Hey, do you know you've got this on your site and you know the performance implications and you know how it got there because that may be a security implication and it's it's you know provided you know some really great insight into into you know Into the content and the content that maybe third parties might be putting on our customers websites and It's just yeah, and it's I found it to be very very very useful to have that insight to share with customers Another thing on the security side is we have another service called enterprise threat protector that enter enterprises can use to To secure their their sites internally so that let's say someone that loads a web page doesn't you know load on like Doesn't load content that had that that's known to serve malware I would actually use the enterprise threat protector data to parse the HGP archive and and look for sites that That we would have blocked with enterprise threat protector And then I found actually found one that was an Akamai customer but wasn't but the site was actually like a typosquadding domain and It was injecting malware onto the site So I went up reaching out to them and said and said hey, you know do you know about this? And they didn't and they were able to go and go and you know work through their legal to see what they could do about it It's amazing how you can actually like Take this data which is almost like emotionless and like not directly tied to a person or an experience But you can take that back to Akamai and then start to reach out to your customers Actually see some real improvements on the web making it safer faster. It's great to see Yeah it's really it's really cool to be able to take that data source and take the output of that and Kind of correlate it with some of our data sources that give us the ability to reach out to customers I did that also recently with we saw there was a zero-day exploit on a popular on a on a popular publishing platform and As soon as I heard about that the morning that it was it was released to the to the public I I said, you know, I wonder if I can query Against the HGP archive to find out who's actually using this particular this particular server software And then I looked for the ones that were Akamai customers I very quickly joined that to a table of contacts and we were able to go out and proactively reach out to people and say Hey, we believe that you're running the software. Are you aware that there's an exploit and It also enabled us to go and offer the the ability to proactively patch the software for them via a web application Firewall before the customer would actually was it for some customers are actually able to patch it on their end It's really really cool ability to kind of correlate that to customer data So let's pivot to another data source. Google releases the chrome user experience report or crux for short where it's another real world user measurement data set could how have you been using this data and Making Akamai customers experiences better. Sure So so Akamai offers a service called impulse which is a real user measurement service And this allows our customers to see the performance of all of their users across all browsers and locations and experiences and pages And so forth. We also use it to correlate performance to business metrics I know Tim Cadillac was on here a few weeks ago He runs a site called WPO stats and on that site You can actually see a lot of use cases for how performance benefits the Performance benefits a company from a business perspective a lot of version rates and ad revenue Yeah, a lot of companies have talked about how they've improved performance and they have seen They've seen advertising revenue increase. They've made they've made more money be a higher conversion rates Impulse is designed to help our customers achieve that achieve those objectives by correlating the performance to to the conversions and revenue not every customer uses real user measurement service right now So a lot of customers I found use, you know, variety of different different techniques to understand the performance of their applications I've actually been showing customers the Akamai customers the Some Chrome user experience report data just to help them understand what their performance actually looks like to real users because as you know that you're actually collecting those real experiences from from from Chrome browsers and And aggregating that so that we can see the you know histograms of the performance some customers have actually implemented a real user measurement solution because like impulse because of that and So it's actually kind of sometimes change the direction of how people will monitor performance because they've seen how different It is from synthetics. And so I know that one of the things that we worked on Together was actually correlating the Chrome user experience report data to impulse data to try to see. Hey, is this is this data Consistent and what we found was that it was very closely aligned. Do you make any of this public through like a report or anything? So the impulse data is is not currently Published externally. So our customers each have access to their To that to their performance data and they use that to analyze their sites performance We used to have a report that we published called the state of the internet report which sounds a lot like the state of the web Video this report was a quarterly report that Akamai published that talked about connectivity across the web So what was the bit rate and in various different countries with percentage of traffic? We're seeing across the web and so forth This was something that has been on hiatus for a couple of a couple of quarters And we've actually been hearing from from people that they they want it back So we have an initiative within Akamai to kind of reboot this and one of the things that we're planning on doing is adding some Some impulse rum data to it so that we can talk about how you know how the web performs as well as how it's being used Interesting. So what kinds of things have you learned? Even if it wasn't published externally Digging into this data about how the web is built the first time I actually started looking into the impulse rum data as as the whole was around November when firefox released the Version of their browser called firefox quantum There was a very strong marketing push about how performance was one of the newest features of the browser Which is actually really great to hear that performance is a feature. I upgraded the new version I was like wow This is much faster than the previous version that I looked at But I wanted to quantify exactly how much performance improvement there was so I started looking at the at the data and I went up Summarizing the performance in histogram and that histogram showed the the range of the experiences that the users Had had received for a specific firefox browser And then I looked at the previous one and the and the one before that and so forth I was actually able to see that there was a 24 percent improvement in DOM content loaded time for the newest version of firefox Compared to a few versions before that and that there was actually a consistent improvement from from every every version release that was released in 2017 So there's a there's a blog post. I wrote about that on an acclimatized developer blog I also did another investigation using impulse data recently where I looked at there's a large power outage in Brazil So it affected millions of Customers and we're hearing that it was mostly northern Brazil But and there was some really great research from dine that was published That talked about the you know from like looking at trace routes where the You know which networks were failing to respond and were most likely part of that power outage What I was able to do is go and look at every city within Brazil that we had traffic from and I created like a Sparkline visualizations for all of them where I can see with a number of page views per minute work from every city in Brazil and I was very very quickly able to see that it was you know a couple of cities within the northeast eastern Brazil traffic Just dropped off instantly and started gradually coming back up and it was really cool to be able to correlate that with with the rum Data and provide some of those insights That's that that's the type of thing that we're looking to incorporate into the new state of the state of the internet report Well, that's great So one of the really cool things about the chrome user experience report is that you can compare performance across Websites even outside of the ones that you control what kinds of avenues does this open up for performance analysis? Especially as you start connecting it with impulse data So with impulse our customers have access to their their own data So they're monitoring their websites and they can see the performance of their websites One of the with synthetic measurements customers are able to look at the performance of not only their site But their competitor sites But they can't get the real user performance of their competitor sites because they don't have any instrumentation on those sites And one of the things we've been hearing from our customers is they really want to understand how their How their impulse data compares to their competitors? Obviously, we can't share that But the chrome user experience report opens up some new possibilities there So because this is a public data source that contains Performance for I think it's up to a million sites now three million three million Wow You're almost certain to find any website that you want to find in there And you'll be able to you know get the comparative performance information So one of the things that we're actually building into impulse It's not there yet But we're building it is the ability to select a couple of different sites that you want to compare your performance to And then it will actually use chrome user experience report data To to show that that that performance comparison so you can actually now see not only how your website performs with your data But see how your website performs with with data that is not your data competitors Right industry web as a whole correct and and yeah, I know that you recently presented a presentation titled not my rum and It's essentially exactly what that is is you know, you've got your rum data But then here's not your rum data, which is from Google cracks Yeah, check it out link in the description So I want to ask you about the future of the web What excites you about the direction that the web is moving? Oh, wow. So that's that's a pretty loaded question I've been doing work on the web for a very long time and one of the things that I Constantly see is that things are changing constantly, but a lot of the we're seeing a lot of similar trends You know if you think about the way mobiles evolved 20 years ago We were talking about like WAP pages to serve simple versions of web content to feature phones And then you know years later. We're talking about responsive web design We're talking about content adaptation to you know serve a desktop site that have a mobile version And now we've got amp things are constantly changing But we're constantly trying to go for the same the same goals Which are reducing the amount of bytes on a page reducing the complexity of the page for for mobile clients And I think that as as we're seeing mobile explode, you know upwards of 50% of traffic for most sites is There's so much more opportunity to you know optimize the performance of that and I feel like we're just getting started What kinds of things would you say to somebody who wants to get started, you know analyzing this kind of data? Some of my colleagues will let they'll nerd snipe me and they'll ask me a question And suddenly I'm doing six hours of research just because it's fun But a lot of that just comes from being extremely curious and trying to trying to ask questions and answer them And if you don't know how to how to answer them look for help and try to find the right data sources And I feel that a lot of a lot of it really just comes down to being curious and and just trying to problem-solve The data is there. There's there's community for analyzing the data and support for that And you know if you're looking to get started in that there's a there's a lot of support in the HGP archive discussion forums and You can you can do so much with it. Well Paul, thank you so much for being here. Thanks for having me So if you'd like to check out Paul's HGP archive presentation at fluent conference this year We have a link to it in the description or if you'd like to keep the discussion going leave a comment below Thanks for watching. We'll see you next time