Home > Technical & Creative Skills > Tracking Campaigns

Self hosted vs cloud hosted speed? (22)

03-20-2019 11:04 PM #1 chris r (Member)
Self hosted vs cloud hosted speed?

Hello all,

I tried doing a search before making this thread but couldn't come up with any definitive answers.

I'm curious if speed is an actual issue with self hosted tracking setups?

I've seen some people say that the tracking server should be close to where the offers are running or else speed will really suffer.

I've also seen people say that it's really not an issue.

Roman at Binom seems really passionate about his tracker, and I've heard nothing but great things regarding their support. My main concerns are speed and being responsible for server maintenance.

Are my concerns valid or am I over thinking things?

What do you guys think?

Thanks fellas,

-Chris

03-21-2019 05:14 PM #2 Mr Baffoe (Veteran Member)

It depends on your linking method.

For direct linking you want to reduce latency so one strategy is to have servers as close to the user as possible. However server location latency is just one part of the equation. Eg a cloud solution could connect faster but take more time processing the click so that the speed benefits are lost.

When using a landing page there’s usually no redirects from ad to lp, most trackers now support using a landing page pixel. So speed is not affected by your self hosted or cloud hosted tracker.

The bottle neck for performance is the redirect link from the landing apge to the offer.

If you have Prosper202 with purlink technology, there's no redirect link from lp to offer. So it's able to out perform every tracker (including cloud hosted) in speed tests.

Apart from the speed issue, the other factors are what kind of control you want over your data and if you are ok with server maintenance.

1. Every cloud solution I've seen has restrictions on how much historical data you have access to. Usually 3-12 months. is this a deal breakers? The self hosted is a better option.
2. With a saas solution, you don't deal with any server or database issues. This is a huge benefit for many because you can focus on marketing. A good self-hosted solution should have people on staff who can fix or help with tracker related server issues if they come up, that's assuming your hosting provider can't fix it.

Overall, your biggest performance drop is often not caused by tracker redirects, it's the affiliate offer redirects and slow bloated landing pages on both the affiliate and offer side. There are pros and cons to saas and self-hosted in relation to server maintenance and data access. There's really no definitive one size fits all answer to that question

03-22-2019 04:29 PM #3 voluum (Veteran Member)

@Mr Baffoe good summary, you've really exhausted the subject here.

@chris r some time ago, we've prepared a comparison self-hosted vs cloud-hosted, you can always take a look (here - sorry if it sounds too salesy, use it to have different points to consider while deciding between self- and cloud-hosted solutions )

Another thing came to my mind while reading this thread. Everyone's just talking about speed tests having in mind redirect speed. While that's fairly important, it would be good to do like a complete test of all the trackers, but with speed of doing things in mind - do a report for last 30 days where you have same amount of visits, create 30 offers one by one, replace one lander in every campaign that uses it, etc. just to see how the reporting speed and UX compares.

Karolina

03-25-2019 04:29 PM #4 chris r (Member)

Thank you guys so much for the great info. It really gives me a lot to think about.

-Chris

03-25-2019 08:59 PM #5 matuloo (Legendary Moderator)

There is one more thing to consider that wasn't mentioned yet, unless I missed it. Cloud hosted trackers usually bill per event, which can be a problem in case you plan to work with POPs, as those come in 10s/
100s of 1000s and exhaust the event limits really fast. So the final cost of tracking can be way higher than the initial plan.

I personally prefer cloudhosted trackers, simply because I don't want to mess with server stuff, and since I run banner traffic, the events count is not an issue. To be fair, I've run on self hosted too and didn't really face any issues.

03-25-2019 11:07 PM #6 thedudeabides (Moderator)

The round-trip latency is almost always going to be the biggest bottle-neck in terms of redirects, moreso than server processing time.

Any tracker that's worth its salt will be caching redirects in memory for fastest possible redirects over making slow database lookups, so you should be seeing speeds there in the low single or double digit milliseconds.

However someone connecting to a server from across the globe could take upwards of a full second or more to get a response back. Chances are you're probably not doing that though and the response time should be negligible.

But this is why I see no good reason to be using a physical server over a CDN for landing pages if you're running whitehat campaigns and don't require PHP or some other server-side language.

Regarding tracking via redirects vs direct to landing page, I don't think it's worth it to sacrifice the ease that comes with using redirects to rotate landing pages by going direct to lander. Unless you're not allowed to by the source obviously.

This industry is all about testing and moving fast.

03-26-2019 01:04 AM #7 Mr Baffoe (Veteran Member)

Quote Originally Posted by thedudeabides View Post
The round-trip latency is almost always going to be the biggest bottle-neck in terms of redirects, moreso than server processing time.

Any tracker that's worth its salt will be caching redirects in memory for fastest possible redirects over making slow database lookups, so you should be seeing speeds there in the low single or double digit milliseconds.

However someone connecting to a server from across the globe could take upwards of a full second or more to get a response back. Chances are you're probably not doing that though and the response time should be negligible.

But this is why I see no good reason to be using a physical server over a CDN for landing pages if you're running whitehat campaigns and don't require PHP or some other server-side language.

Regarding tracking via redirects vs direct to landing page, I don't think it's worth it to sacrifice the ease that comes with using redirects to rotate landing pages by going direct to lander. Unless you're not allowed to by the source obviously.

This industry is all about testing and moving fast.
Great points.

I'll add that there have been a ton of improvements by CDN providers over the past few years, and the technology is here to allow you to fully manage split tests, rotations and a whole lot more stuff via CDNs that traditionally needed a physical server.

Before the end of the year, I predict the lines between self-hosted and cloud hosted will blur even more.

03-26-2019 05:15 AM #8 thedudeabides (Moderator)

Quote Originally Posted by Mr Baffoe View Post
Great points.

I'll add that there have been a ton of improvements by CDN providers over the past few years, and the technology is here to allow you to fully manage split tests, rotations and a whole lot more stuff via CDNs that traditionally needed a physical server.

Before the end of the year, I predict the lines between self-hosted and cloud hosted will blur even more.
Yeah there are some neat solutions for split-testing via CDN out there. Netlify makes it possible to do it via git branches. I still wouldn't go that route over just using a tracker redirect, cause it's a bit more involved to setup, but it's a cool option none the less.

And more trackers nowadays are adding multi-variate testing to their tracker so you can test on-page elements easily, Funnelflux and Binom being ones that come to mind there.

Hoping to see someone come out with a solution where you control and pause things from within the tracker itself, akin to Google Optimize/VWO, instead of having to manually edit code to set weights, text, etc.

03-26-2019 12:28 PM #9 roman binom (Member)

There are a lot of arguments regarding this issue. And for sure, cloud-based trackers would unanimously claim that they alone are worth using, while the rest do not function the way they are to function. But this is simply not so.
First of all, the problem of slow redirects is relevant only when it comes to certain types of traffic, such as cheap pops-related traffic (it is also important here, whether your backfix gets the traffic), push-notifications probably. As for Facebook traffic, it is not important here whether your landing page loads during 50 ms or 150 ms. The mechanism is totally different here.
Secondly, Binom is used by numerous top affiliates with enormous traffic volumes (guys from my support service dealt with trackers associated with dozens of billions clicks on a single tracker) across all the geos at once. Believe me, they are professionals. And they would never direct their traffic via a tracker which gives them only a half of their typical ROI after they start using it! Just take a look at our AdPlexity ranking after all. I believe we are leaders in terms of average traffic volume per customer. That is why I have always viewed such problems as unsubstantiated and exaggerated by cloud-based trackers, since it is their main advantage.
Third, just think of it. Where are the servers related to your traffic sources? Where are your publishers’ servers? And advertisers’ servers? Where is your offer hosted? How may redirects are there on the user’s way to the final page? So, this redirect from a source to the tracker is just one of many other redirects. And very few links of this chain use AWS.

Quote Originally Posted by chris r View Post
Roman at Binom seems really passionate about his tracker, and I've heard nothing but great things regarding their support. My main concerns are speed and being responsible for server maintenance.
Thank you, Chris, you re right.

03-28-2019 02:33 PM #10 kintura (Member)

I've seen dns CNAME lookups in certain geos take 1200ms alone so this whole thread, technically and relatively speaking, should be about DNS. So if you *are* using a cloud-based tracker boasting ~1ms redirects yet they're sticking your custom domain behind a CNAME record: you've been duped, sorry. Also check their SSL implementation of your cert and ssllabs score (which should look like this)

Redirecting a user with no complex lookups in ~5ms is easy. You need to explore Pingdom which will give you a breakdown of where your bottleneck is if you have one. You also need to be checking your self-hosted/cloud tracker on Pingdom from whichever geo your camp(s) are running in. I've done consulting work for the largest self-hosted tracker, trust me when I say vertical scaling is a train wreck. In addition to that, I firmly believe session store should be a separate system entirely. Then the machine learning which you need to remain competitive as an affiliate...this just isn't a one-instance/one-server job anymore. Maybe 8 years ago it was.

A lot of self-hosted trackers are missing out on the power of distributed computing. Just our AI alone couldn't be run on one instance of any size but we're a little different in that our customers are essentially building their own personal facebook pixels in terms of machine learning.

03-28-2019 03:48 PM #11 Mr Baffoe (Veteran Member)

Quote Originally Posted by kintura View Post
I've seen dns CNAME lookups in certain geos take 1200ms alone so this whole thread, technically and relatively speaking, should be about DNS. So if you *are* using a cloud-based tracker boasting ~1ms redirects yet they're sticking your custom domain behind a CNAME record: you've been duped, sorry. Also check their SSL implementation of your cert and ssllabs score (which should look like this)

Redirecting a user with no complex lookups in ~5ms is easy. You need to explore Pingdom which will give you a breakdown of where your bottleneck is if you have one. You also need to be checking your self-hosted/cloud tracker on Pingdom from whichever geo your camp(s) are running in. I've done consulting work for the largest self-hosted tracker, trust me when I say vertical scaling is a train wreck. In addition to that, I firmly believe session store should be a separate system entirely. Then the machine learning which you need to remain competitive as an affiliate...this just isn't a one-instance/one-server job anymore. Maybe 8 years ago it was.

A lot of self-hosted trackers are missing out on the power of distributed computing. Just our AI alone couldn't be run on one instance of any size but we're a little different in that our customers are essentially building their own personal facebook pixels in terms of machine learning.
I don't think we've done a consulting project with you (sorry had to throw that line in)

But seriously yes, you are right.

1. There're many bottlenecks in the whole speed equation that have nothing to do with the tracker. CNAME is terrible in multiple ways, in addition to the latency issue, it's also a vector for link banning. You can't just CNAME to a banned url and expect the new url to be ok.

2. Scaling a self-hosted tracker is still a huge pain point for the average affiliate, it's a problem self-hosted trackers need to solve to be competitive. Work is being done on the Prosper202 side to automate and eliminate scaling issues, but it's much more difficult when dealing with a collection of heterogeneous computing environments. If only there was a way to... ding! wait a minute there is a way that just might work...

3. The nature of self-hosted trackers, also allow them to become the most powerful, biggest, cost effective, resilient distributed computing network. But that's all I'm going to say about that for now, because this is still a spark of an idea that's been evolving over the past year.

ps: I know we are "technically competing" but from the outside looking in, I'm a fan of your work.

03-28-2019 07:35 PM #12 kintura (Member)

I don't think we've done a consulting project with you (sorry had to throw that line in)
Ha awesome! I was hired for my "domain expertise" in the area of distributed computing in performance marketing and that is all I can say (to legally clear myself haha). Maybe they weren't the largest self-hosted but...my god the traffic.

I know we are "technically competing" but from the outside looking in, I'm a fan of your work.
Same to you, truly. Also we're neighbors and still need to meet up in SF one of these days so I can get you drunk and steal all of your brilliant ideas!

04-05-2019 08:10 PM #13 simputer (Member)

Never SaaS your money. Your data is your money. Once you SaaS it someone else has it.

04-22-2019 11:49 PM #14 chris r (Member)

Geez louise guys! I totally forgot about this thread! I ran into it last night doing a search and come to find out, it's full of awesome replies. Thank you all so much!


Quote Originally Posted by matuloo View Post
There is one more thing to consider that wasn't mentioned yet, unless I missed it. Cloud hosted trackers usually bill per event, which can be a problem in case you plan to work with POPs, as those come in 10s/
100s of 1000s and exhaust the event limits really fast. So the final cost of tracking can be way higher than the initial plan.

I personally prefer cloudhosted trackers, simply because I don't want to mess with server stuff, and since I run banner traffic, the events count is not an issue. To be fair, I've run on self hosted too and didn't really face any issues.
That's a great point Matej! I don't plan on staying with pops any longer than I have to, but if I did, this would be a great advantage over cloud hosted trackers.

I do plan to start taking push more seriously though. Wouldn't push be similar to banners in that respect?


Quote Originally Posted by roman Binom View Post
There are a lot of arguments regarding this issue. And for sure, cloud-based trackers would unanimously claim that they alone are worth using, while the rest do not function the way they are to function. But this is simply not so.
First of all, the problem of slow redirects is relevant only when it comes to certain types of traffic, such as cheap pops-related traffic (it is also important here, whether your backfix gets the traffic), push-notifications probably. As for Facebook traffic, it is not important here whether your landing page loads during 50 ms or 150 ms. The mechanism is totally different here.
Secondly, Binom is used by numerous top affiliates with enormous traffic volumes (guys from my support service dealt with trackers associated with dozens of billions clicks on a single tracker) across all the geos at once. Believe me, they are professionals. And they would never direct their traffic via a tracker which gives them only a half of their typical ROI after they start using it! Just take a look at our AdPlexity ranking after all. I believe we are leaders in terms of average traffic volume per customer. That is why I have always viewed such problems as unsubstantiated and exaggerated by cloud-based trackers, since it is their main advantage.
Third, just think of it. Where are the servers related to your traffic sources? Where are your publishers’ servers? And advertisers’ servers? Where is your offer hosted? How may redirects are there on the user’s way to the final page? So, this redirect from a source to the tracker is just one of many other redirects. And very few links of this chain use AWS.



Thank you, Chris, you re right.
That's great info Roman! There is no doubt in my mind that your tracker really kicks ass. I'm just not sure if a self hosted tracker in general, would be the best solution for pops or push. Like you said, if I was doing FB or something like that, then it wouldn't be a problem.

You're very welcome!


Quote Originally Posted by thedudeabides View Post
The round-trip latency is almost always going to be the biggest bottle-neck in terms of redirects, moreso than server processing time.

But this is why I see no good reason to be using a physical server over a CDN for landing pages if you're running whitehat campaigns and don't require PHP or some other server-side language.
That's interesting. I hadn't considered the whole round-trip being the biggest issue.

It's funny you mention the lack of needing a physical server anymore. I've been wondering the same thing as of lately. The more I've been reading and doing homework, the more it sounds like it's not really needed. Unless of course you're running PHP landers like you had mentioned.


Quote Originally Posted by Mr Baffoe View Post
1. There're many bottlenecks in the whole speed equation that have nothing to do with the tracker. CNAME is terrible in multiple ways, in addition to the latency issue, it's also a vector for link banning. You can't just CNAME to a banned url and expect the new url to be ok.
Can you explain this more? I think I understand what you're saying but not totally sure.

My lander domain just got boned (flagged) by the great Googal, and your statement has me worrying that just changing out my domain won't be enough to be in good standing with Google.



Question for everyone:

This early in my AM career, I don't expect to be running a whole lot of traffic in Tier 1 geos anytime soon. If I did end up going with a self hosted tracker, would it be wise to host it on a VPS in Europe or Asia? I'm in the states btw.

Thanks again fellas!

-Chris

04-23-2019 07:24 PM #15 matuloo (Legendary Moderator)

That's a great point Matej! I don't plan on staying with pops any longer than I have to, but if I did, this would be a great advantage over cloud hosted trackers.

I do plan to start taking push more seriously though. Wouldn't push be similar to banners in that respect?
Push is somewhere in the middle, it doesn't reach the volumes of POPs in terms of the tracked events, but the numbers are higher than with banner clicks. So depends on how much you'd manage to scale.

This early in my AM career, I don't expect to be running a whole lot of traffic in Tier 1 geos anytime soon. If I did end up going with a self hosted tracker, would it be wise to host it on a VPS in Europe or Asia? I'm in the states btw.
I did some tests, comparing a US based VPS against a CDN, the GEO I was targeting was India I think. And didn't really spot a difference. The most important part is that the host has close to 100% uptime and works "fast", there is no point in chasing miliseconds.

04-24-2019 05:28 PM #16 chris r (Member)

Quote Originally Posted by matuloo View Post
Push is somewhere in the middle, it doesn't reach the volumes of POPs in terms of the tracked events, but the numbers are higher than with banner clicks. So depends on how much you'd manage to scale.
Ah okay. That's understandable.


Quote Originally Posted by matuloo View Post
I did some tests, comparing a US based VPS against a CDN, the GEO I was targeting was India I think. And didn't really spot a difference. The most important part is that the host has close to 100% uptime and works "fast", there is no point in chasing miliseconds.
Was this for tracking or for hosting landers?

-Chris

04-24-2019 07:27 PM #17 matuloo (Legendary Moderator)

Quote Originally Posted by chris r View Post
Was this for tracking or for hosting landers?

-Chris
This was for landers, it's true that a tracker would ad another redirect... so it could possibly have a bigger impact. Didn't test this, so don't want to give you wrong advice.

04-24-2019 09:39 PM #18 chris r (Member)

Quote Originally Posted by matuloo View Post
This was for landers, it's true that a tracker would ad another redirect... so it could possibly have a bigger impact. Didn't test this, so don't want to give you wrong advice.
Oh okay. I might just have to try a self hosted setup one day and see what I can determine.

-Chris

05-30-2019 07:24 PM #19 kintura (Member)

Quote Originally Posted by simputer View Post
Never SaaS your money. Your data is your money. Once you SaaS it someone else has it.
Oh yeah much better to use a self-hosted tracker encrypted with ioncube/zend guard so you can't see backdoor data access

The only solution to this is to do business with people you trust, period.

05-30-2019 10:32 PM #20 Mr Baffoe (Veteran Member)

Quote Originally Posted by kintura View Post
Oh yeah much better to use a self-hosted tracker encrypted with ioncube/zend guard so you can't see backdoor data access

The only solution to this is to do business with people you trust, period.
Agree trust is important.

But you can also just choose the only self hosted tracker that comes with 100% unencrypted source code

05-31-2019 02:28 AM #21 kintura (Member)

Quote Originally Posted by Mr Baffoe View Post
Agree trust is important.

But you can also just choose the only self hosted tracker that comes with 100% unencrypted source code
Very good point and I'm glad you said something!

06-18-2019 03:12 PM #22 redtrack (Member)

We often receive similar questions and we have a lot to say here, but let me share what our customer who switched from a self-hosted solution said. Yet another point to consider:

“The best case is proxy detection databases. Due to licensing and stuff, self-host trackers cannot provide it and their IP databases are revised when their tracker updates like once in
2-3 months. So the IP information is not precise.

Secondly, to run them we need high spec hardware which is not useful in case of low traffic”.​

Aksana

Home > Technical & Creative Skills > Tracking Campaigns