Enterprise scalability?

[vincent]

Hi folks,

Has TYPOlight been proven reliable in an "enterprise" environment?

I work in the IT department of a small Canadian university. Earlier this year, my team was asked to review a number of open-source content management systems and select one for a limited pilot project. We settled on TYPOlight; finding it to have the best mix of features, flexibility and security out-of-the-box. The original plan was to put TL though its paces and eventually offer it to the campus community as the IT department's recommended and supported CMS solution. Usage of TL was to be opt-in.

This week, the CMS project as we knew it was changed dramatically by the university's senior administration. Now, my team has been told that use of the CMS will be mandatory campus-wide at the administrative and academic departmental levels.

With that in mind, I quickly crunched some numbers based on recent web stats and existing infrastructure which is almost entirely comprised of static sites and one-off custom code at the moment.

This is a rough estimate of the volume we're looking at with a mandated campus-wide product:
~ 1 million page views per month
~ 250,000 visitors per month
~ 175 unique websites (subdomains/subdirectories)
~ 350 content managers

So... To reiterate my initial question, is TL capable of scaling up to meet our projected needs? We've already invested a lot of time in evaluation, templating and support for our original pilot participants. Thus far, the general impression of TL from both end-users and developers has been positive, so we'd like to continue with the product assuming it can handle the load potentially burdened upon it.

--V

[scare]

I was thinking about this scalability of TYPOlight a lot lately and I concluded that it is not "enterprise ready" (yet). The biggest problem I see is with custom module development. Imagine that somebody creates a custom module in system/modules folder and uploads some file with syntax error in it. TL scans all the folders and includes files from them and this would immediately break all your 170 sites, because there is no way to decide which site loads what modules.

However I believe that no other open source CMS is any better than TL in this regard and I think that TL can be adapted to your needs with reasonable effort. I would do it this way:

1. create one TL installation per site, i.e. one database and one directory with TL files (put symbolic links to good use here) - a shell script may be created that would automate this. This has 2 advantages - you may create custom modules that don't affect other sites and you may distribute your sites to as many computers as required to handle requests.
2. use checkCredentials hook to authenticate users against some global database
3. the biggest problem is content sharing. Maybe you can live without it. My idea was to create something like content aggregator - a module that could query all databases of all sites (a good cache is required here indeed) and list available content. You then select any content you like and content aggregator fetches the content from selected database, caches it locally and then displays it.

Well, this are my thoughts, I don't really see any fundamental problem here, but I may be wrong

[vincent]

1. create one TL installation per site, i.e. one database and one directory with TL files (put symbolic links to good use here) - a shell script may be created that would automate this. This has 2 advantages - you may create custom modules that don't affect other sites and you may distribute your sites to as many computers as required to handle requests.
2. use checkCredentials hook to authenticate users against some global database
3. the biggest problem is content sharing. Maybe you can live without it. My idea was to create something like content aggregator - a module that could query all databases of all sites (a good cache is required here indeed) and list available content. You then select any content you like and content aggregator fetches the content from selected database, caches it locally and then displays it.

Per the points raised:

1) We're hoping to go with multiple site roots on a base of, perhaps, three TL installs--one each for academic groups, administrative groups and "everyone else".

2) Network credentials for staff, faculty and students are centralized, so tying TL to Active Directory is a given. I've already prototyped an AD/LDAP authentication extension which is being used with our two pilot sites.

3) Shared content hasn't been raised as a "nice to have" feature for our web trustees on campus (...yet). We'll cross that bridge when we come to it. Fortunately, virtually every site that will be ported to the CMS is unique and serves its own select audience. Our marketing/communications department already has a mechanism in place to publish and distribute generic university content which can be tied to a TL *.tpl with just a few lines of PHP.

--V

[FloB]

This is a rough estimate of the volume we're looking at with a mandated campus-wide product:
~ 1 million page views per month
~ 250,000 visitors per month
~ 175 unique websites (subdomains/subdirectories)
~ 350 content managers

Interesting note on Twitter:

Typolight isn't only for small websites... I serve ~250'000 pages for approx 450gig of traffic with it every months #typolight #typo3

[FloB]

The biggest problem I see is with custom module development. Imagine that somebody creates a custom module in system/modules folder and uploads some file with syntax error in it. TL scans all the folders and includes files from them and this would immediately break all your 170 sites, because there is no way to decide which site loads what modules.

Well, you shouldn't let anybody upload a TL module with a syntax error – you wouldn't develop on a running system, either, would you? Only allow the maintainers to upload working extensions!

[scare]

Only allow the maintainers to upload working extensions!

That would work. In an ideal world

[mymyhope]

i think typo3,typolight,modxcms is the three best cms in the open source php cms world

[FloB]

Only allow the maintainers to upload working extensions!

That would work. In an ideal world

Well, then make sure there are only very few people who can give the permission to upload an extension, the best would be only one person. Anyway, this shouldn't be the reason for not using TL .

[scare]

Well, then make sure there are only very few people who can give the permission to upload an extension, the best would be only one person. Anyway, this shouldn't be the reason for not using TL .

I agree, but I think it is the reason not to put all websites in a single TL install. Imagine that extension works, but slows down page significantly. This may not be detected in development and testing phase, but could slow down and (in high load environment) even disable all sites. I have bad experience with bugs that do not show up until the code is deployed on production server. I believe it is always better to set your system in a way that only small part is affected if something goes wrong, because eventually something will go wrong.

[vincent]

i think typo3,typolight,modxcms is the three best cms in the open source php cms world

For interest's sake, before settling on TYPOlight, we looked briefly at both Typo3 and MODx, as well as CMS Made Simple, Concrete5, Drupal, Joomla 1.5, Pivot, SilverStripe and SavaCMS.

With respect to users adding their own extensions, this isn't even an option in our application--Security and stability trump everything (even at the expense of added features, if necessary). Website "trustees" will be granted sufficient privileges to manage their content, but not tinker with system settings, user accounts nor extensions.

--V

[vincent]

As a followup to my original post 6 weeks ago, the decision has been made for my team to put its full effort into deploying TYPOlight as the exclusive CMS for departmental use on-campus.

We're now in the midst of putting finishing touches on the design template, backend user permissions and LDAP authentication. There are currently a dozen early adopters in the queue awaiting training and website rebuilds. Barring any major catastrophes between now and then, we estimate as many as 150 sites coming on board over the next 8-12 months; hosted in a multi-site configuration. (I suspect we'll have 3-5 instances of TYPOlight installed to distribute the load.)

Now if only I could convince the nay-sayers that there's more to the open-source CMS world than Joomla, Drupal and Typo3!!

[Ben]

Vincent - it's good to hear you decided to go with TYPOlight. Regarding your last sentence, I'm always shocked at how many people think WordPress / Joomla / Mambo / Typo3 / Drupal are so amazing. It seems like people don't do their own research and simply go with what their friend thinks they should use. :lol:

[matt590]

Vincent - it's good to hear you decided to go with TYPOlight. Regarding your last sentence, I'm always shocked at how many people think WordPress / Joomla / Mambo / Typo3 / Drupal are so amazing. It seems like people don't do their own research and simply go with what their friend thinks they should use. :lol:

What's wrong with those CMS ? especially Drupal...

[Ben]

Matt, there is nothing wrong with those systems. They just aren't TYPOlight.

[=qp=]

I don't think Joomla, WP, Drupal, etc. and Typo3 should be mentioned in one sentence, different leagues.

If you want to develop web-applications, Typo3 or Zope is the way to go IMHO. Didn't look close enough at ModX, but that could be a lightweight framework judging by 5 minutes on their site.

Then again, Vincent was talking about web-sites here, and I would be very interested to hear more about first-hand experience in enterprise-class deployment scenarios.

[Ben]

qp - I only mention those systems together because they have high popularity. That's about the only similarity they have other than an ability to manage content.

So how are things going so far, Vincent?

[vincent]

So how are things going so far, Vincent?

Our dedicated server with 1 TL installation is currently serving 66 sites/domains: 11 are live and 55 are being staged by their respective content managers.

I anticipate that next week will be the biggest test as we switch the university's main "www" domain over to live service under TL with 100-200K/mo page hits.

Otherwise, so far, so good. Aside from a few small hiccups, the response from content managers--technically proficient and raw novices alike--has been quite positive. Recent upgrade from 2.7.2 to 2.7.4 was completed inside 15 minutes and with only a blip of downtime. I've been dabbling more with custom-built extensions to address a few of our in-house concerns and continue to be thrilled with the speed at which a working prototype can be crafted.

--V

[lsanju]

Vincent,

I would appreciate if you can share your experience on the performance of the website that you deployed using Typolight. I'm planning to start a website which is a commercial in nature. My biggest concern is the PERFORMANCE and SEO. My target website will be similar to Amazon.com with lot of dynamic webpages and reviews. I'll appreciate if you can share your experience using Typolight.

[vincent]

I would appreciate if you can share your experience on the performance of the website that you deployed using Typolight.

Follow-up to initial post "Suggestions need from TypoLight Experts" ~ viewtopic.php?f=4&t=992#p3860

--V