Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

When a small online company is expecting drastically increased demand for a short period of time, how do they counteract this?

(First of all, I am a not a server/network techie and this post is purely for informative purposes.)

 

When a small online business is running and they release a new product that has somehow been waiting in anticipation for years and lets say they get increased traffic to their network by 1,000-2,500% for 3-5 days, how would an online business counteract this?

 

I know you can just 'rent' more servers, but how easy and quick is it to set-up that process. Just curious. 

On 11/5/2020 at 10:12 PM, dizmo said:

AMD is, frankly, pathetic.

CPU: AMD Ryzen 9 16-core 5950X

CPU Cooler: Artic Freezer 2 AIO 360mm Radiator

Motherboard: Asus ROG Strix X570-F Gaming

Memory: 32GB (2x16GB) G.Skill Trident Z Royal 3600 MHz CL16

GPU: Nvidia Founders Edition RTX 3090

Storage OS: 500GB Samsung 980 Pro Gen4 M.2 NVme SSD

Storage Games: 2TB Corsair MP600 Gen4 M.2 NVme SSD + 2TB Samsung 860 Evo SSD + 500GB Samsung 850 Evo SSD

Storage Misc: 3x 2TB Seagate Barracuda Compute 7200 RPM

PSU: Corsair HX Platinum 1000W 80+

Case: Fractal Design Meshify S2 ATX Mid Tower

Monitor: Asus PG258Q 240Hz 1ms 1080p TN panel 24.5"

Link to post
Share on other sites

Well that mostly depends on how the website is architected and where the load demand is highest or if increasing front end web capacity will just overload other components like a database server.

 

If the load is mostly just people viewing products pages etc then that is a fair decent use case for putting that out to a CDN.

Link to post
Share on other sites

Most businesses rent solutions that scale on demand along with some sort of content delivery network solution (CDN).

 

By the time CDN propagation is finished and handling the majority of demand, the only thing that really hits their actual databases is stuff like account creations, orders, etc. When one server handling those things is fully loaded, their provider will automatically spin up an additional instance to handle the spillover and so on.

MacBook Pro 16

i9-9980HK - Radeon Pro 5500m 8GB - 32GB DDR4 - 2TB NVME - many dongles

Link to post
Share on other sites
31 minutes ago, leadeater said:

Well that mostly depends on how the website is architected and where the load demand is highest or if increasing front end web capacity will just overload other components like a database server.

 

If the load is mostly just people viewing products pages etc then that is a fair decent use case for putting that out to a CDN.

29 minutes ago, Vitamanic said:

Most businesses rent solutions that scale on demand along with some sort of content delivery network solution (CDN).

 

By the time CDN propagation is finished and handling the majority of demand, the only thing that really hits their actual databases is stuff like account creations, orders, etc. When one server handling those things is fully loaded, their provider will automatically spin up an additional instance to handle the spillover and so on.

Ahi I see. So in layman terms, it is sort of a snowballing system where the bigger the demand, it automatically gets upgraded to more server space etc.

 

If that's the case though, why do some websites still crash when receiving high demand? Even some bluechip websites I have seen go down when demand is high.

 

On 11/5/2020 at 10:12 PM, dizmo said:

AMD is, frankly, pathetic.

CPU: AMD Ryzen 9 16-core 5950X

CPU Cooler: Artic Freezer 2 AIO 360mm Radiator

Motherboard: Asus ROG Strix X570-F Gaming

Memory: 32GB (2x16GB) G.Skill Trident Z Royal 3600 MHz CL16

GPU: Nvidia Founders Edition RTX 3090

Storage OS: 500GB Samsung 980 Pro Gen4 M.2 NVme SSD

Storage Games: 2TB Corsair MP600 Gen4 M.2 NVme SSD + 2TB Samsung 860 Evo SSD + 500GB Samsung 850 Evo SSD

Storage Misc: 3x 2TB Seagate Barracuda Compute 7200 RPM

PSU: Corsair HX Platinum 1000W 80+

Case: Fractal Design Meshify S2 ATX Mid Tower

Monitor: Asus PG258Q 240Hz 1ms 1080p TN panel 24.5"

Link to post
Share on other sites
1 minute ago, Actual_Criminal said:

Ahi I see. So in layman terms, it is sort of a snowballing system where the bigger the demand, it automatically gets upgraded to more server space etc.

 

If that's the case though, why do some websites still crash when receiving high demand? Even some bluechip websites I have seen go down when demand is high.

 

It depends on the load. Say there's a crazy pre-order going on and you have tens of thousands of people trying to purchase something. When a site like that stops responding, their database servers are likely going kaput. The additional resources being allocated can't communicate fast enough with each other to keep up or their provider simply ran out of available hardware.

 

If you see a site just go down because hordes of people are viewing the same page, it's usually because the specific CDN a user is connecting to can't physically resolve the incoming addresses fast enough, even if they still may have plenty of headroom available to deliver actual content after the connection is made. Meanwhile, someone across the country connecting to another instance (in another physical location) from the same CDN may NOT be overloaded and experience no issue connecting at all. Cloudflare for instance has servers eeeeeverywhere. Users in one geographical location may be seeing an unresponsive site while others a few states over are having a normal experience.

 

Then you have stuff like DDoS attacks which just essentially flood servers with pings on such a massive scale that it affects real users trying to resolve their regular page requests.

MacBook Pro 16

i9-9980HK - Radeon Pro 5500m 8GB - 32GB DDR4 - 2TB NVME - many dongles

Link to post
Share on other sites
17 minutes ago, Actual_Criminal said:

Ahi I see. So in layman terms, it is sort of a snowballing system where the bigger the demand, it automatically gets upgraded to more server space etc.

 

If that's the case though, why do some websites still crash when receiving high demand? Even some bluechip websites I have seen go down when demand is high.

Not every site can actually easily dynamically scale, things like database caching services can only accept so many connections and automatically adding servers for that may not be possible or necessarily safe to do without oversite. The simpler the website the easier it is to just keep scaling out capacity, once you start adding more complex functions it gets harder and harder.

 

Auto scaling web services is more of a aspirational goal for most corporates, they don't actually have them or there are technical limitations to which parts and components of the site can actually be scaled out easily.

 

We have this same problem where we run a learning platform service called Moodle, we've add a lot of extra parts to the hosting of this to allow for the number of students that use it. While we can scale out the web frontends easily and update the front end load balancers to include these servers where we actually have performance problems is with Redis and the PostgreSQL database. We are in the middle of upgrading the database platform to support active-active write architecture, currently only one server node for the database platform is active and used for website load and the rest are passive and waiting in case the primary fails. With that type of architecture you have to keep making the server bigger and bigger resource to handle more load but there is a maximum possible and also a maximum effective possible.

 

How easily it is to scale out a website heavily depends on the website itself.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×