Announcement

Collapse
No announcement yet.

Time to First Byte

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Time to First Byte

    Since Google is placing more emphasis on fast loading sites, we have been testing our site for time to first byte.

    Yes, we have tested with Google Page Speed and all is good -- BUT Google is very forgiving.

    We have another well respected site WebPageTest.org (https://www.webpagetest.org) that will test web site performance. As this is an non-profit, everyone who uses it speaks very highly of it. They are very accurate and complete in their testing.

    Every time we test our site we receive either an "F" or "D" for time to first byte. All other metrics receive an "A".

    The explanation of time to first byte by our web designer is:
    Time To First Byte (TTFB) is a measurement used as an indication of the responsiveness of a webserver or other network resource. TTFB measures the duration from the user or client making an HTTP request to the first byte of the page being received by the client's browser.

    I writing this post for 2 reasons:

    1. Would all of you Miva users check your (or your client's) website on WebPageTest and see if you also receive an "F" or "D" on time to first byte. What are your results?

    2. No one from Miva has truly explained to us WHY our time to first byte score is so poor. Perhaps at the conference, this is something that might be discussed as it is very important for a website. There must be a reason our score is so poor.

    Thank you

    #2
    It basically comes down to Google acting intelligently and differentiating between dynamic content sites and static sites, while webpagetest simply considers ALL websites as comparable from a speed perspective. The reason there is such a difference is because it's nearly effortless for a web server to return static content, while there can be quite a bit of processing time to return dynamic content. Specifically, if you have an index.html as your site's home page for example, your TTFB is probably going to be in the 20 to 40 millisecond range, because all the web server has to do is read the file and serve it back. If it's served frequently, the file is likely already in memory on the operating system side, so it doesn't even need reading off disk again.

    In contrast, if your home page is your storefront (screen code SFNT), then that same request is going to trigger execution of the Miva software (underlying engine plus Miva Merchant store software), it will have to perform some amount of work (parse the template data, do things that may be special if the customer is logged in, if conditionals are in use, etc.) and then return the data. Depending on how any given copy of Merchant is configured, the TTFB will typically be somewhere in the 300 ms (0.3 sec) to 1 sec range; obviously heavy optimization sometimes involves feature tradeoffs, so a good webmaster/designer who knows Merchant and what can be turned on/off can be quite useful.

    Google likely realizes that shopping sites, in particular, often have some 'work to do' in serving a page, and may not hold TTFB in as high a regard; at least that's my theory.

    Now, you obviously want the lowest TTFB regardless of whether it's static/dynamic. There are various things we're working on behind the scenes that are targeting reducing that. An example; the 9.56 update that is coming out today has some optimizations that specifically relate to sites using our ReadyTheme templates, and in some cases may chop several tenths of a second off, to even a second or more on some sites configured in less than optimal ways that have really slow pages. Another one is running our engine software in a manner that a new process doesn't have to start and exit for each request; that should be possible soon as well. There are many other things; I'm happy to talk to you about it more at the conf if you're here, just find me, I'll be around all days and events. I might be able to get a better idea of what your site is doing specifically, and get you with someone who can take a look at possible optimizations, or just make some recommendations on things you could try.
    David Hubbard
    CIO
    Miva
    [email protected]
    http://www.miva.com

    Comment


      #3
      Thank you for your thoughtful reply.

      One thing that I did not mention is that we used to get either "B" or "C" on TTFB on our dynamic site. Then all of a sudden it dropped consistently to the D's and F's. We don't know what would have caused it - and as I said these lower results have been consistently low..

      I still would like to ask the other Miva users to test their sites on WebPageTest.org. It would be interesting to know if others are have the same low scores.

      Comment


        #4
        If the drop was sharp but your site's performance didn't seemingly change, they may have just revised their grading scale. Definitely worth investigating either way though.
        David Hubbard
        CIO
        Miva
        [email protected]
        http://www.miva.com

        Comment


          #5
          I'll gladly proxy for the site owner while I'm here at the conference.
          Leslie Kirk
          Miva Certified Developer
          Miva Merchant Specialist since 1997
          Previously of Webs Your Way
          (aka Leslie Nord leslienord)

          Email me: [email protected]
          www.lesliekirk.com

          Follow me: Twitter | Facebook | FourSquare | Pinterest | Flickr

          Comment


            #6
            Thank you Leslie. From personal experience working with Leslie -- she is an amazing (and extremely knowledgeable) designer. In a word. . .she is remarkable. YES! Please be my proxy!

            Comment


              #7
              I am still hoping that some Miva people will run the WebPageTest on their sites to see what results they get on their TTFB.

              Comment


                #8
                Hey WeLoveOurDogs,

                I ran the webpagetest on 4 live production Miva sites - for First Byte Time score:

                2 sites scored F
                2 sites scored A

                The 2 that scored the F are substantially more complex sites, running things like Search Spring and lots of conversion tracking scripts. The 2 that scored A are very simple Miva sites without much external scripts running. So it seems the underlying Miva core processor can be fast under the right circumstances (whatever those are).
                Joe McDonald
                patrickwebby.com

                Comment


                  #9
                  I just ran the test for my site for two different pages--a very simple non-Miva page that is generated with php and a much more complicated product page from my store that is generated by Miva.

                  The php page scored an A with TTFB of 113 ms. The Miva page scored a C with a TTFB of 466 ms.

                  While a C is something I wouldn't stand for on my daughter's report card when I know she can get an A, a difference of .35 seconds in TTFB is not going to have a notable effect on user experience for someone visiting my site. I would worry more about the numerical value of TTFB and the impact on user experience than the arbitrarily defined letter grade returned as part of the test.
                  Todd Gibson
                  Oliver + S | Sewing Patterns for Kids and the Whole Family

                  Comment


                    #10
                    And, for what it's worth, I absolutely hate using Pinterest. I can't stand it because every single page takes ages to load. It's a horrible user experience because it's so slow. Because this topic was on my mind and I needed to go to Pinterest to view an image, I just ran the test for one random page on their site. The page scores five A's, but it take just shy of six seconds to first view.

                    My Miva product page, meanwhile, scores two C's and three A's but only takes 3.5 seconds to first view. And there is much more content on my page than on the Pinterest page.

                    While the test results in the waterfall view are useful to identify where you can optimize individual components of your page, the overall grades don't seem to align with what's of ultimate importance--time to first view and, by extension, user experience.
                    Todd Gibson
                    Oliver + S | Sewing Patterns for Kids and the Whole Family

                    Comment


                      #11
                      This is an issue that is not going away. I've had an SEO "guru" tell a store owner that he needs a "new host, DNS, and SSL certificate" and then recommends "Amazon AWS for as much of that as possible, and namecheap for the SSL if not (let's encrypt)". He had tested the store's static home page with the "Chrome - 3GFast" setting. The page got a C, A, A, B and F. I'm going to send the report he provided to Miva Support.
                      Leslie Kirk
                      Miva Certified Developer
                      Miva Merchant Specialist since 1997
                      Previously of Webs Your Way
                      (aka Leslie Nord leslienord)

                      Email me: [email protected]
                      www.lesliekirk.com

                      Follow me: Twitter | Facebook | FourSquare | Pinterest | Flickr

                      Comment


                        #12
                        I'd question the advice from anyone who suggests the brand of SSL cert plays an SEO role; with the obvious exception to certs from unrecognized authorities or that Google has chosen to distrust of course. DNS would be even more questionable. DNS too for that matter; DNS' critical metrics are response time and availability, and we use one of the largest DNS providers for our hosted sites' DNS for those reasons.

                        The most common issue I see with Miva Merchant sites scoring low, if the low grade is the result of page load times being 0.75sec or higher, relates to unoptimal template configuration, settings, or add-on modules. The wrong combination of any of those can kill an otherwise fast site, or, more importantly, can cause a site to be unable to handle high concurrency, so it may seem fine at times, but under heavy load falls apart. Our professional services group can do a thorough site audit that can identify those types of issues.
                        David Hubbard
                        CIO
                        Miva
                        [email protected]
                        http://www.miva.com

                        Comment


                          #13
                          I opened a support ticket on behalf of the store owner with the full screenshot. I almost want to publically mention this guy's name but will not. David, I seem to recall conversations you and I had about the overall impact of requiring everything to go through SSL. At the time, you were not a fan.

                          Here's what the "guru" is complaining about when it comes to SSL (his circles not mine). Keep in mind, this is a static HTML page and not one dynamically generated by Miva Merchant:

                          Leslie Kirk
                          Miva Certified Developer
                          Miva Merchant Specialist since 1997
                          Previously of Webs Your Way
                          (aka Leslie Nord leslienord)

                          Email me: [email protected]
                          www.lesliekirk.com

                          Follow me: Twitter | Facebook | FourSquare | Pinterest | Flickr

                          Comment


                            #14
                            A specific site's metrics are not really something that is well handled via forum discussion regardless of whether the site name is publicized. TTFB is heavily dependent on a huge variety of factors and without the necessary context, any discussion here would just be general theory, or with the site name, would delve into information about the given site's configuration that is not appropriate here. Dynamic, static, are rewrite rules being used, how many, is a CDN being used, which one, where's the test being run from, is it IPv4 vs IPv6, how is the DNS set up, so on and so forth. If you feel the Technical Assistance Center staff have made a mistake, or someone with the requisite knowledge has not been involved, I'm happy to re-route it and/or contribute if I can, just shoot me an email with the details.
                            Last edited by ILoveHostasaurus; 10-30-18, 08:50 AM.
                            David Hubbard
                            CIO
                            Miva
                            [email protected]
                            http://www.miva.com

                            Comment


                              #15
                              If you have a developer who's willing to look under the hood, it's definitely worth checking out the Miva debug log: https://www.miva.com/forums/forum/de...ript-debug-log

                              I've used it to pinpoint things that slow down sites I've worked on, such as slow modules, unnecessary mvt:items, etc.

                              Also worth noting that your TTFB can be different depending on which page template is being used. Your homepage (SFNT) might be 500 while your product pages (PROD) are 400 and category (CTGY) are 800, etc. So each page should be checked thoroughly
                              Looking for work as of March 2024! I've been a web developer for going on 20 years, with most of that time spent on Miva sites.

                              Comment

                              Working...
                              X