Software as a Service (SaaS) for forums - Is it really a problem?

Hi,

When Internet Brands hinted at vB5 being possibly a SaaS solution, many customers voiced their displeasure and that was almost over 2 years ago. Since then the Cloud Computing revolution has continued to gain momentum. Companies like Google, Salesforce.com and Facebook have continued to grow and be successful, offering their versions of "community" in Social Networks and Social Business on cloud computing foundations. I have openly said, I want to start a community solution based on cloud computing. It will be a form of hosting service. So each individual site (not just a forum), will be its own "place" in the internet. However, there is still this negative reaction to SaaS. (and in our case, we are actually going to offer a PaaS solution)

I personally think cloud computing is the future just because humans are inherently lazy. And, please, please, please, I do not mean that negatively. I mean that absolutely positively. Humans just want things easy, efficient and as convenient as possible. They don't often want to bother with details. Those that do are actually special. In most cases, humans just don't want to constantly be bothered with the nitty gritty, pain-in-the-ass stuff like running their own servers or even the issues of finding a web space and constantly updating software. Most will say, it isn't a big deal and for sure, often it isn't. But then again, often it is. And no matter what, deep down, almost everyone says in someway to themselves, maybe even unconsciously, "There must be an easier way than this!" It is in our genes. And ta-da! Cloud computing is born. It didn't just happen because it's a stupid idea. It is very much happening, because it solves the problem of people having to worry about the nitty gritty, pain-in-the-ass details of service up a forum or community as a service. SaaS built on cloud computing is simply much more convenient, efficient and easy.

But that is my opinion.

What do you think are the advantageous and/ or disadvantages to a SaaS or PaaS solution for forums or community software on a cloud computing platform and, if possible, also explain why. What does it mean for the user? What does it mean for the admin? What does it mean for the developer?

Scott

 

Personally, for me as an end user/forum owner, it's all about the level of control you have over your entire operations. I'm a control freak

• For instance, if this week I want to run my community on an Apache, PHP, Oracle MySQL 5.5 linux stack but next week I decide I want to change it up and swap out Oracle MySQL 5.5 for better performing Percona 5.5 or MariaDB 5.5 MySQL forks - I can do so on my own schedule.
• If my custom written server performance metric stats show MySQL can do with some more performance tuning, then I can immediately change a few MySQL settings and restart services when I want
• The week after that, I decide I'd want to swap Apache for better performing Litespeed web server, I can do so.
• The week after that, I decide I want to add Varnish caching proxying in front of Litespeed web server, I can do so.
• The week after that, I get a great deal on new server hardware/hosting, so decide to change web hosts ASAP, I can do so on my own schedule.
• Week after that, I find out a more updated Linux kernel offers better performance than the one I am currently running, so I look at options to upgrade my Linux kernel when I want.

Freedom & Control

Convenience (laziness) vs Control
While, I agree humans are inherently lazy when it comes to some tasks, you have to remember humans also fear loss of control as well - less options less control vs more options and more control. How well you balance these two factors, will be key and up to each service provider.

Example in cloud hosting, I've tried various cloud hosting providers, Amazon EC2/Rightscale, GoGrid, Rackspace Cloud, and my fav now PhoenixNap Secured Cloud and each one has different levels of complexity when it comes to initially setting up new cloud based servers.
Amazon EC2 most complex and time consuming so being the lazy human that I am, I don't use much of these days. Rackspace cloud has a much easier setup process but most expensive bandwidth pricing and lowest server hardware performance of all cloud hosting providers I've tested as they still prefer to deploy slower AMD based processors most of the time. Then I stumble upon PhoenixNap Secured Cloud with ease of setup like Rackspace Cloud but pricing comparable or cheaper than Amazon EC2 and added benefit they use the latest and fastest Intel Xeon E5 processors !

Self hosted vs Cloud hosted
So you ask yourself why use dedicated/self hosting versus cloud hosting etc ? Still comes back to the amount of control you have. Alot of cloud hosts bandwidth costs are higher without set quota allocation and there's a level of unpredictability when it comes to bandwidth costs if you traffic peaks are much higher than you budgeted for. Take for example with typical dedicated server you'd get 10TB = 10,000GB of bandwidth - your monthly costs are fairly known and fixed. There's a level of control you have over your costs.

Ultimately, for most folks it will come down to price. Usually, for most things in life, convenience and ease usually come at a higher price - especially when it comes to cloud hosting/computing.

 

Hey George.

Thank for that post. Some interesting information.

Just some quick comments....

You are definitely someone I'd call special. Your specialization and expertise in the server world is and always has been impresive to me.

I'd say only a smaller portion of humans are control freaks and less than convenience freaks. If that weren't the case, the iPhone would have never made it big. And, you are absolutely right too. The trick to success will be in the mix of making enough convenience lovers and control freaks happy and I believe that is exactly where Apple gets alot right. And I think the trick is a smart platform that lets you get a lot done, but is easy and efficient to use.

Thanks for the information on cloud hosting. Are you still sure about Rackspace? They were rated top cloud hoster by Gartner in 2011. I don't think the criteria you mentioned was part of the comparison, but still. And I like their style and support. I'm going to look into PheonixNap.

And as for pricing. A SaaS or PaaS offering must have a Success based pricing in the mix. That is the only way the service can work. So pricing will fluctuate as a site gains momentum, stagnates or even loses momentum. I see that as a huge advantage of SaaS, because as a site owner, if you had built up a large server farm due to some short term success, but then you have a drought period, you are stuck with the most likely higher costs of running the whole farm and you'll be locked into some contract too, if the servers are hosted. That could cause some financial issues. With a Saas offering, this shouldn't be an issue.

Scott

P.s. I wrote all of the above on my IPad, so please excuse the quality of writing....

 

Two words which are the key to success or failure: Instant Gratification
It's what end users want in any product. I want what I want and I want it now.
The more a user can do and the more simply it can be done, the better.
Your average end user simply isn't all that bright. Cater to the simple-minded.
Or more politically correctly, to the least educated consumer.
Personally, I've read enough of your vBulletin posts to know that you can make anything work.
I'd certainly be interested in such a solution.

 

Thanks Scott, always still learning more as each day passes - it's neverending

Add another point, with the aspect of control for self hosted, some folks who don't even have the expertise still prefer control in self hosted environment and rather pay someone with the expertise to do what they want i.e. software upgrades, server optimisation etc etc.

That situation leads to demand for server administration and management companies or hired system admins or managed server web hosts. You don't necessarily need to be a server specialist or server guru to choose the self hosted path. Cloud hosting providers are also competing with that mix of alternative options as well the other end of the spectrum of extreme control, that of co-located server hardware hosting.

True - nice gui based interfaces, one click actions to do certain tasks etc. All make life easier. Simplicity hiding complexity - just look at Android as an OS (linux) compared to normal linux OS distribution popularity

Yeah pretty sure, Rackspace's strengths are in the top notch support not their hardware they use But then again it could be going back to what I said about their easy to use control panel for their Cloud offerings - Rackspace's way better than Amazon EC2 in terms of ease of use and setup.

I've been hired in the past to consult for folks moving their forums into the cloud with Amazon EC2 and Rackspace and a few other cloud providers and in most cases their server hardware - particularly the disk I/O performance is a real let down bottleneck. The clients, 100% of the time had better performance moving off the cloud and back to self hosted servers. Even had one client's cloud server disk I/O being slower than a laptop 5400rpm disk in terms of random seek times !

But for me PhoenixNap Secured Cloud is my preferred choice for testing in the cloud these days - especially ease of use, bandwidth pricing and fast Intel Xeon E5 based hardware with good disk I/O. Still early days but liking it

Yes cloud hosting properly done, should be priced well (particularly for the unpredictable and variable cost of bandwidth $$$) and allow customers to take advantage of the flexibility that cloud can offer in terms of scaling up and down when needed. But there are limits to how much you can scale vertically.

Most cloud providers max out at 8 cpu threads/cores and 64GB of memory per server. If you want more than 8 cpu threads and/or 64GB of memory, you have to scale out horizontally by adding more cloud/virtual servers to the mix. Goes back to the factor of control vs convenience (flexibility). Of course you could setup cloud hosting to allocate an entire cloud node cpu cores to one client, but then the costs would be approaching the approximate costs of going self hosted instead.

Why would you want more cpu threads/cores ? Explanation maybe too long, but short explanation is with multi-threaded tools and apps you can improve performance dramatically.

For instance, multi-threaded mysql database backup tool mydumper can backup databases upto 3x-10x times faster than mysqldump. Pigz multi-threaded gzip compression, can compress files up to 33x times faster than gzip. Combine mydumper + pigz and just imagine how much faster mysql database backup process becomes as your server supports more cpu core/threads

 

indeed goes back to what i referred to above as " Simplicity hiding complexity"

 

@ProSportsForums - thanks. It's going to be a long haul, as we are starting very much with nothing. But I want it badly. We'll see what happens.

@George - I see what you mean with the difference in price for bandwidth between Rackspace and PheonixNAP. 5c to 18c per Gig is quite a difference. They are generally overall less expensive.

What do you think about the "Cloud Files" offering and being able to serve up files in an Akamai CDN?

Scott

 

Yes huge difference, for 10TB per month the bandwidth price is $500/month at PhoenixNap vs $1,800/month at Rackspace. Of course Rackspace has a more diverse network presence especially if you're looking at edge presence in Europe or Asia/Oceania while PhoenixNap is mainly within USA and Amsterdam.

But remember to weight up bandwidth costs. If you know where to look, you can get a very high end performing dedicated server with 10TB included bandwidth for $79-199/month (bandwidth included) I've helped clients of mine save $1,000s or $10,000s each year in web hosting costs

For cloud/cdn files, again comes down to price and performance (target visitor audience determines the benefits of cdn). See my 3 page article at http://vbtechsupport.com/1419/ and response times of various CDN providers. At one end you have best performance at up to 2x to 10x times the price Akamai, CachyFly and Edgecast CDN and then other end value for money MaxCDN/NetDNA and CDN77 (wasn't around at time of article).

I mainly use maxcdn or cdn77 bandwidth costs can go as low as US$0.032/GB for decent performance/response time. With Rackspace cloud files US$0.10/GB and Akamai CDN at US$0.18/GB they are more expensive option that personally I've never see the need for me to take advantage of unless I was running a for-profit revenue earning web site/forum.

For instance MaxCDN deal in my inbox recently

• 5 Terabytes for $219 > Savings of $131
• 25 Terabytes for $899 > Savings of $741
• 50 Terabytes for $1699 > Savings of $1581
• 110 Terabytes for $3521 > Savings of $3060

All Bandwidth is Valid for 24 Months

Akamai CDN priced at US$0.18/GB would cost

• 5TB = $900
• 25TB = $4,500
• 50TB = $9,000
• 110TB = $19,800

At 110TB per month, I could literally buy a new car every month here in Australia with the money saved going with MaxCDN vs Akamai

I'm sure Akamai has discounts at higher volume levels too but the starting prices have me scared off already

As I stated before, cloud hosting benefits of flexibility and convenience will always come at a higher price compared with self hosted solution.

 

There are also a some intangibles involved too. Being I believe in complete transparancy, telling your customers your service is built on a hosting company like Rackspace with a solid and very reputable name or a hosting company like PhoenixNAP (although they are good too) can make a huge difference, simply because of the name or reference. Remember, reaching a high level of trust is the ultimate goal in a PaaS offering. So there is also a price to pay for something that is simply more trustworthy upfront. It might get you a number of customers more just on that first leap of faith they need to make. That is why a no-name, although I could save a ton of money, just doesn't come into the equation. I want the best for my customers.

Also, I am certain, should a PaaS offering ever get to the point of 110TB of traffic, then there is so much involved, the owners will be talking with the hoster's top brass about costs and can deal with them for better pricing. The PaaS might also have to start thinking about other options than a purely cloud hosting system too. Probably a private cloud system.

Also another consideration. Being it is PaaS, let's say you have a really successful customer and let's say they are using 10TB of traffic themselves. If they were to host it themselves, then they would also need to pay a lot more for servers and also have people running them. A good server admin can run around 80k a year, if not more. Sure, that admin may not be needed all the time, but you do need support, whenever something crashes and quickly, that all costs money and nerves too. With a (trustworthy) PaaS system, you have costs yes, but should have no worries. It just works! Again, it is like you say. You have to pay for convenience.

The thing of it is, the cloud computing paradigm also lets someone with little resources simply do a whole lot more, faster. And that also goes as an advantage to the service for the customers who do SaaS on the platform themselves. If their site skyrockets with users for some reason, a well done PaaS should just meet the need simply and with nothing other than increased costs. I do realize, this kind of scaling isn't unlimited, but it should cover most expectations, IF it is done right.

Scott

 

Indeed, I agree cloud allows someone to with little resources to do a lot more faster.

I'm mainly using cloud hosting at phoenixnap just for test servers to try new tweaks and tuning out, testing my centminmod project's new releases and more so for testing large cluster based setups. i.e. multiple load balanced web servers, mysql replication, tungsten based mysql replication and going to try out galera and mysql cluster based setups as well. Will only cost me a few dollars with cloud hosting and just destroy the cloud VM servers after I have done testing

One thing you'd have to be aware of, server hardware is getting more powerful these days Intel E3-12xx V2 Ivy Bridge and E5-16xx and E5-26xx Sandy Bridge-EP servers are way more powerful than previous generations and priced alot more competitively, so going self hosted or co-located isn't as expensive to run a large forum/site as it was in the past. And when I refer to large forums, I'm speaking of 20 - 60 million post range, 10-60K users online etc.

i.e.

Example #1

Single Intel Xeon E3-1230 V2 (4 cpu cores/8 cpu threads) is as fast as dual Xeon E5620 (2x (4 cpu cores/8threads)). But you can get an Intel E3-1230 V2 server for as low as US$79/month to US$199/month depending on configuration. My Intel Xeon E3-1270, 16GB 2x 1TB SATA Raid 1, 10TB server only costs US$99/month Had one client update to new Xeons from 2yr old 3x Xeon Nehalem based server config I setup for him and he has a new 3x server setup (E3-1230 Sandy Bridge based) and saved US$1,200/month (US$14,400/yr saved !) and got 6x times better performance

Example #2
For rental for 3x dedicated server setup consisting E3-12xx V2 x2 web servers with E5-16xx or E5-26xx db server could run around US$600-1200/month with 20-30TB of bandwidth or US$400-900/month with E3-12xx V2 db server and same 20-30TB bandwidth. Such a setup can handle easily 400,000-600,000 unique ip visitors/day. Of course depends on traffic patterns too and will be more expensive if fully managed.

Of course if your cloud hosting starts off with the same Intel processors mentioned above, then you're already on the right track for scalability and performance

 

Thanks for the tips George.

Yeah, going with a pure server rental or even a co-hosting system will need to be considered, just on the price savings alone. Like I said, a private cloud offering is also something to be considered and will have to be, once a certain size is reached for sure.

In the end, what needs to happen though is absolutely no user of the system can be left with a slow loading web page. That MUST be a main goal of any PaaS or SaaS. It's so very much a key responsibility, because success of the whole concept depends on it.

Scott

 

You're welcome and yes fast loading pages is a must, but for cloud based hosting other key features to justify the higher price would be high availability, redundancy, very good disaster recovery and reliability - all factors that involve alot of expensive infrastructure investments on the cloud provider's end - so working out that price charged vs feature offered matrix will involve alot of thought

 

Yup, and that is the advantage a customer of a PaaS system has. All that thought is continuously being made by the PaaS provider, so the customer doesn't have to. In the end, it is a win-win situation. The PaaS provider, if their services are good, gets more customers and thus gets more "weight" with his IaaS provider (if the system isn't already self-hosted like Google or Facebook) to deal with them about their costs. So theoretically, over time and as the system grows, either prices will stay stable or they might even drop.

Scott

 

Just a small update, especially for George: I brought up the pricing issue to Rackspace in their forum, actually wanting to hear from other customers about their experience and what had happened and either Rackspace already new about their pricing being an issue and had a plan up and almost ready to make public or they actually reacted very fast. (I doubt they did this just for me). At any rate, I even got a reply from their CEO in their forum and they now have a fairly fair tiered pricing in place for their bandwidth. It is competitive. It was an interesting conversation and experience and goes to show, customer service does really start at the top and RS believes in their support being very important to their success.



Scott

 

Nice http://www.rackspace.com/blog/lower-open-cloud-pricing/ Open Cloud bandwidth first 1TB at US$0.10/GB

Definitely much better than what they had before i.e. on 300TB bandwidth previous cost would be US$54,000 vs now at US$0.08/GB = US$24,000 and just a tad off from PhoenixNap at US$0.05/GB = US$15,000

Still US$24K vs US$15K per month is US$9K difference, and that can easily equate to 12x to 20x additional dedicated web servers you can use the difference for heh

Also curious are your SaaS plans going to adequately survive on Open Cloud's restricted internal network bandwidth http://www.rackspace.com/cloud/servers/techdetails/ ? highest plan at US$1.20/hr only going to give you 300Mbps public / 600Mbps internal bandwidth as opposed to a dedicated 1000Mbps private link ?

600Mbp after over heads equates to around capped 60MB/s max transfer speeds. Any private traffic backups between servers will only go as fast as 60MB/s on highest US$1.20/hr plan or on 2GB US$0.12/hr plan 12MB/s or 4GB US$0.24/hr plan at 20MB/s.

30GB data file internal transfer would take

• 30,720 /100MB/s normal providers with 1000Mbps private link = 307 seconds
• 30,720 / 60MB/s at rackspace = 512 seconds
• 30,720 / 20MB/s at rackspace = 1,536 seconds
• 30,720 / 12MB/s at rackspace = 2,560 seconds

 

Um, there won't be anything needed to back-up on the servers themselves. That wouldn't work at all with our plans, because we would then be "limited" to the confines of the storage on each server or we'd need to use block storage, which we are going to use anyway. We will have SSD storage for process sensitive data like templates and template cache, customer extension programming, customized application programming, etc. and then we'll have normal storage for a private data file system. The servers are going to have to be redundant too, so we can scale horizontally and I believe storing data on them would kill that ability. We are also going to use the Cloud Files/ Akamai CDN for the public/static file system. Something I know (through their Service rep, who I talked with personally) and they also didn't mention in that blog post is the outbound bandwidth for the CDNs is also going to be a bit cheaper than even PhoenixNAS using large amounts of bandwidth. (I think it was above 500TB= $0.04/GB).

If I had the backing from the start, I'd definitely go a more dedicated route. This is also why I asked the question in their forum. At some point, if we ever get a fair number of customers (and I am thinking in the low to mid 10s of thousands), we are going to reach a point, where the cloud offering from RS just isn't going to be economical for the benefits or advantages it gives us as a start-up. So, sooner or later, we will need to migrate our services to more dedicated infrastructure.

Thanks for your help George. Your tips got me wondering and asking the right questions and I thank you for that.

Scott

 

You're welcome

Interesting plans you have ahead of you though

 

Well, one thing cloud computing should allow us to do, and is one big advantage, is scaling relatively quickly and still only having to pay for the compute power and storage we use.

Let's say we have a customer who just happens to go to Oprah and she gets national publicity for her start-up and her site is on our system and gets a big surge of traffic (the Oprah Effect). If we had a dedicated and limited system, because we have a system that is only a bit bigger than what we need or could afford at the time, then possibly her site would go down and even bring everyone else's down too. That would be a real no-go!

Our planned cloud system will have a robotic monitoring system. If traffic reaches a certain level, the system replicates server nodes and/or even load balancers by itself. That is the "fast" track. If the system reaches a certain "compute" level over a bit longer period of time, then the nodes will also scale vertically increasing CPUs and RAM.

I've timed it and building or rebuilding a node with our current set-up takes about 3 and a half minutes. I expect that to increase some as we build the system out more, but it shouldn't take more than 4 to 5 minutes to get new nodes online. Adding load balancers takes a bit less time. So we could add a node or change the flavor of a node or expand the front line to the Internet with load balancers relatively quickly to help cover even big spikes in traffic.

In order to be able to give that kind of service with a dedicated set-up from the beginning, we'd have to invest in a whole lot of infrastructure, which we might not even need. Also, with the cloud set-up, should the customer's site who got hit by the Oprah Effect cool down later, we could then also pull back nodes and load balancers and actually save money.

That is a pretty neat advantage. I think the possibilities are actually pretty cool, despite the extra costs of the cloud. They make it worth having, especially for a system, which will need to grow and shrink like a living organism.

Scott

 

My only three concerns with SaaS are:

1)How much is the 'stupid tax'? Meaning how much extra is it going to cost me to run on your managed platform vs a self hosted system like vBulletin or IPB. Updating forums is part of my business I don't mind an FTP upload and a few template edits here and there.

2)How easily can I get my data out (posts, members, attachments, avatars, etc) if I'm not happy.

3)Are you going to try and create a universal login that works across all forums running on your software? I always disliked that idea.

 

1) We are attempting to price the system, as if it were just a normal hosting plan, with on-premise software like vB or IPB. How expensive it may get should be in direct proportion to your success. We'd like to think of your success being ours. That is the only way a PaaS or SaaS can work. In the end, if you are successful in terms of traffic and usage, then your costs will increase, but fairly. If you stay humble and small, then your costs would stay humble and small.

2) We are planing to have a type of Data-Loader. This will be the data import and export system. It will also be a crowd sourcing part of the system, as people will be able to create mapping of their data to import into our system (and also export out) and they can give these mappings to others to use.

3) Yes, we will have a universal system for logging in and some other neat ideas that can only work over a network. However, you, as a site owner, will not have to be on that network, if you don't want to. Not being on that system though, will mean your users won't be able to take advantage of the free stuff that will get them back onto your site and reading and taking part in discussions and being good community members. The network is what will make your site stronger, get more users to your site faster and get those users back on your site more often.

One thing I want to definitely break is this "they are my users" mentality so many forum owners have. Nobody "owns" users. I wouldn't even consider the users on our system "ours". They could be a user on our system one day and not the next. The same goes for your system too. What makes them "our" or "your" users is loyalty and trust and possibly even devotion. You have to build these things. And even then, the users still don't belong to you. They want to belong to you. Quite a big difference.

Scott