I have heard many experts state that the use of a CDN has no impact on SEO, but I am not convinced. Clearly speeding up a website WILL have a positive impact but my primary area of concern is with regard to PageRank.
My understanding of pagerank is that it is calculated on a per domain basis and that each sub-domain is considered a separate domain.
IF that is true - surely offloading static content from your primary site 'myexample,com' to a sub-domain at 'static.mydomain.com' ( which is CNAMED to the CDN) would REDUCE the pagerank of 'myexample.com' because the 'juice' from all that static content would accrue to the sub-domain and not the main site which is where you want it.
I am currently having to make a decision about using a CDN and balancing the many benefits against the potential loss of PR juice.
I would appreciate any insights or experiences with this.
Comments
Issue
http://drupal.org/node/1060358
Interesting
Interesting discussion which I was aware of. However the focus was on duplicate content and I am more concerned with understanding the issues around pagerank ( if there are any ). I guess at the very heart of the matter I am trying to get an answer to this simple, but vital question.
If you put all your static content on a sub-domain, is the pagerank value associated with all those files lost to the root domain?
Suddenly all those links that previously self-referenced... are now effectively pointing to another site... is that going to have a negative impact by draining valuable PR from the root domain? . I have heard stories of pagerank being negatively impacted by using a CDN.
html cdn
I think that is mainly an issue if your hosting html on a CDN domain.
What about the Image Backlinks?
But what about all those images? The root domain must still link to them and those back-links are going in the wrong direction! I've googled extensively to see if google has anything to say about it and not come up with anything... yet.
Isn't PageRank for pages?
In your original post, you stated:
It was my impression that PageRank was calculated on a per page basis. The home page of a website will have a PageRank, since it is technically a page, like any other in this respect, though some confuse the PageRank of the home page as the "domain PageRank" for the domain.
Here is a statistics heavy article on PageRank from SEOmoz supporting this:
http://www.seomoz.org/blog/what-is-pagerank-good-for-anyway-statistics-galore
"homepage PageRank is sometimes used as a proxy for a hypothetical “domain PageRank.” While technically inaccurate, this study supports the idea that the PR of a website's homepage provides information about the domain as a whole."
If you move all of the static content to a different domain, you could preserve most of the PageRank by creating 301 redirects. These pages would likely do as well as they had been in SERP.
Without knowing anything about your website, the question is what content do you expect to show up in SERP, and what content will get incoming links, would help determine a recommended course of action.
David Minton, Managing Partner, DesignHammer | Durham, NC, USA
Re: Isn't PageRank for pages?
Well Yes! Pagerank is for pages but what determines a given pagerank is determined by a number of factors amonst these being the flow of PR juice through the links to and from a page. There have for example been many reports of positive PR for hotlinked images. The incoming links attracting an improved PR, so in my mind "resources" do attract some PR also, and similarly - in my view, may also give it away if linking to external sites...
For me the question is really one of understanding - so that I can make an informed decision about the implementation of a CDN. My current view is that moving static images or resources ( not html ) to a CDN does potentially have serious PR implications, no matter what kind of site is being contemplated - the question then becomes, how can this be mitigated against?
I was very interested in your comment about using 301 redirects, surely a SE would see straight through that and just assign PR to the resource on the domain ( or sub-domain ) it actually exists on. Could you point me to some sources on that?
The site that I am contemplating this issue over - will have a large number of content pages, each of these pages will also have several images associated with them. These images/resources are the nub of this discussion... I would like to off-load these to a CDN but am concerned that this will have a negative impact
Re: Isn't PageRank for pages?
I have not heard about images having PageRank, so I would be curious to read about it, if you can point me toward some articles. None of the PageRank tools I have seem to be able to detect PageRank in any images that I have checked, so I don't know if they don't have PageRank, or if I'm not using the correct tool.
Be that as it may, do you expect people to link to "static images or resources ( not html )" that you are considering moving to the CDN? If not, why would they get any PageRank that you fear losing?
Also, my comment about using 301 redirects is for cases of moving content that has PageRank that you want to preserve. There is nothing to "see through," since wherever the redirect points, is where the content can be accessed.
David Minton, Managing Partner, DesignHammer | Durham, NC, USA
Pagerank for Images
As stated I have heard of reports of an improvement in SERPS from people who have experienced the dubious benefits of having images hotlinked - I cannot re-call the article nor vouch for its accuracy however this article on SearchEngineWatch http://blog.searchenginewatch.com/080428-095720 about google's "Visual Rank" clearly indicates that google at least places "value" or "Rank" on images, how or if this plays into the pagerank algorithm - can only be guessed at but I cannot but believe that if Google is going to this length to evaluate images - the results of that evaluation is likely to feed into the PR algorithm.
Your response does however raise a very interesting point. Whilst I am personally convinced that images do add "value" to a site, and that that value can influence its position in the SERPS - I have no factual evidence that this is so - and if it does - is that weighting part of Pagerank? I would love a clear answer! It clearly follows that IF images themselves attact no PR ( or other weighting benefit ) then using a CDN would not present a problems - I just want someone to convince me!
With regard to linking to images - I do not expect people to link to these images - the issue is that the primary domain does. An important part of getting a good pagerank is having a well designed internal linking structure that moves the PR to the pages you most want it. PR travels through the internal link structure of the site. My instinct tells me that images play a part in this linkage structure and links to effectively external images may draw PR away from the primary domain.
Pagerank for Images
That was an interesting article, but it only applied to image search, and not the general SERP, at that time. It also was geared toward identifying the relevancy of the images. I don't know what type of images you would move to the CDN, but if they are UI related, I don't see why anyone would be searching for content based on them, and therefore should not have any effect on your general SERP, if this technology ever was used for general SERP.
For example, if you had a page about iPods, and you had had a photo of an iPod, and people linked to that photo, maybe its popularity would help the page in SERP for searches related to iPods. You noted you don't expect people to link to the images on the CDN, so I don't think this would apply. I'm also not clear of what types of images your would host on the CDN, so they might not be relevant to any search anyway. For example, if you hosted a "add to cart" button on your iPod page, the content of that button is not relevant to a search for iPods, and shouldn't provide any boost to SERP.
As far as internal link structure, I thought that only considered hypertext links (i.e. anchor tags), and not any other link, such as image tags. I amy well be very misguided on this, but I haven't read anything about image files playing any role in SERP. The anchor text does, but I don't recall the location of the image playing a role, that rather the page that has the anchor text get "credit for it, though this is something that should be testable.
Take a look at what major players who do well in search do. I took a quick look, and Apple and Sony, and both use subdomains for hosting images. Amazon, on the other hand, actually uses a completely different domain for images. It could be due to their scale they had no choice but to use a CDN, despite any SEO drawbacks (if their are any). What are your competitors with better SERP placement doing?
While you raise some interesting questions, I don't believe they are relevant to your decision to deploy a CDN. I may well be completely wrong on this topic, and would like to learn how things work if I am mistaken.
David Minton, Managing Partner, DesignHammer | Durham, NC, USA
Yep - Their all doing it!
Whist I still have some reservations, I am generally persuaded! Everyone is doing it including competitiors - so if there is any SEO impact it is likely to be small compared to the benefits of a CDN. Ultimately, if this is the way things are generally done on the web - I have no doubt that the SE's will adapt their algorithms to reflect the "real world".
I think you are right - that any SEO impact using a CDN is likely to be so small that it is not really a consideration! Of course it is impossible to be certain about anything without "insider knowledge".
"pagerank is determined by a
"pagerank is determined by a number of factors amonst these being the flow of PR juice through the links to and from a page."
Flow through is not a requirement. The only requirement is that the URI in question receive inbound links from URI that have PR.
PR and XML Sitemaps
Taking this further, PR is calculated on a URI basis, not on a domain or even page basis.
PR is a measure of the quantity and quality of backlinks pointing to a specific URI. If the image or video is embedded within the page and backlinks point to the URL of your "page", the resource would not receive the PR, the URL that the resource is embedded into would. However, the resource URI could also receive PR. The best bet is to look to see what the CDN is doing doing with their XML Sitemaps for video and images. If the are pointing to your URLs then you will benefit. It isn't just a matter of PR though. Google for example can display a thumb of the video in the SERPs if they can associate the URL with the resource. This could have a significant impact on click-through rate so is another reason to get sitemaps in order (this is true with YouTube and other embedded media as well).
Wistia for example points the sitemap listing to your URL. See http://wistia.com/product/video_seo and http://wistia.com/doc/video-seo.
Disclosure: I have no affiliation with Wistia so can't vouch for or against.
I completely agree with your
I completely agree with your point of "associating" your content with the resouce - getting images and videos attached to your listing in the serps or from the "image/video" searches is very powerful.
But going back to the CDN issue - the normal way that CDNs are implemeted is that images etc are simply dumped on the CDN and then referenced from the main site ( somtimes slightly obfuscated with redirects etc). These images have no way of pointing back at the main domain. The idea of creating a sitemap on the CDN that points back to the primary site is very interesting - it becomes a way of giving some PR love back. But would it work? the wistia example you gave works because the sitemap appears on another website and is effectively providing incoming links -but would'nt those links be going to the CDN if the resources were on a CDN? Is there a way in a sitemap to point to the resource on the CDN and also say "this resource is associated with this URI" which would point to the content it is embedded in?
I have had a quick look at some of the search results and all the images I have seen are based on the same domain as the page. It was a very quick look - it would be interesting to see if there are examples of a google listing with an image that is on a different subdomain or domain or CDN...
"it would be interesting to
"it would be interesting to see if there are examples of a google listing with an image that is on a different subdomain or domain or CDN..."
This occurs all the time. For example, the SERP listing will include a video thumbnail. The SERP points to a page on your site but the video is actually embedded on the page and is located on YouTube for example. What makes this work is the video sitemap indicating that a given page contains a XML video and a "page" XML sitemap references the page. Google sees that the two are linked via the video and may add the thumbnail to the SERP listing.
Thanks for clarifying - do
Thanks for clarifying - do you think rdf as implemented by drupal 7 could be utilized to enhance the linkage between video/images and content?
RDF will increase in
RDF will increase in importance as SEs implement more of its capabilities and adoption increases. Currently specific industry and data types benefit while others don't. It's only a matter of time...
Sitemaps
Wistia's articles state that I am to put a "sitemap" into my Robots.txt file. Doesn't that override my XMLSitemap? If that's the case, wouldn't my site's rankings plummet because Wistia doesn't include all my content.
Nancy Dru
You specify the location of
You specify the location of the sitemap file in robots.txt.
http://www.sitemaps.org/protocol.php#submit_robots
And...
"The" is singular, indicating that if I put it in robots.txt, the XMLSitemap (in a different location) for the whole site will be lost to the search engines.
Nancy Dru
Then I don't understand the
Then I don't understand the issue. You can specify multiple sitemaps in robots.txt.
Finally
"You can specify multiple sitemaps in robots.txt." There's the answer.
Nancy Dru
Does PR follow image src uri's?
One interesting question came out of the above discussion with @DavidMinton was whether PR juice follows the image src uri's. If anyone has an answer or insights into that little nugget I'd love to hear about it!
I don't know if its
I don't know if its measurable. Toolbar PR may not show for a resource say for example, a URI to an image file. However that is Toolbar PR, not true PR that we're not privy to. Image files do receive backlinks and they do rank in image search so it seems logical that G might include PR as a ranking factor. This is difficult to check. Even checking backlinks is challenging. Some of my tests on Open Site Explorer yielded no results. Same with Site Explorer and G doesn't reveal much in that regard either. However, Blekko provides a wealth of data (including linkage).
Some of my colleagues and I discussed this recently and some claim that files receive PR though I'm not certain they would be able to validate that either. I've seen nothing official published from G but that doesn't mean much. Personally I think the image search point signals to me that they are calculating it.
In regards to CDN, I think the question is not whether CDN is useful or not. The question is which ones smartly factor SEO into their offering. So, I'd of course want to know how good their service is (speed, reliability, etc) but I'd also want to know how they will be helping or hurting my site as it pertains to SEO. Wistia is a good example.
Edit: Added "Even checking backlinks is challenging." Corrected a few typos :-)