Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Two VMS vs one Dedi
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Two VMS vs one Dedi

I was always curious, what should be done once resources are reached.

I currently use a CX-51 from Hetzner's cloud, but during spikes, it seems that my website needs more resources.

Since I do love cloud computing and its a big pros for me (almost everything redudant) - I am worried if its better to move to a Dedi, or to create a load balancer scenario from nginx, using a second CX-51.

Whats your opinion guys?
Running a simple LNMP there.

Comments

  • msg7086msg7086 Member
    edited July 2018

    Either should work. A dedi (usually) gives you more resource, while 2 VMs gives you better redundancy while being a bit complicated to set up. (While better redundancy, it still creates more failure points.)

    So it's really YMMV question. If you can bear with occasional hardware failures (a dedi is not that fragile anyway) I'd go with a dedi -- as long as the resource is actually enough for your usage.

  • MasonRMasonR Community Contributor
    edited July 2018

    Throwing more resources at something is hardly a solution if you don't understand the underlying issue. What resources are you maxing out (RAM, CPU, bandwidth)? Any enhancements (i.e. caching, optimizations, less DB queries, etc.) that can be done to lessen the resource burden?

    Once you understand the issue you're having a little more, then that should steer you in the direction to take. For instance, if you're maxing out memory on a 32GB VPS, then you won't want to buy a 32GB dedi as you've gained nothing from doing that. Likewise, if you're maxing out a 1Gbps connection, you won't want to get a dedi with a 1Gbps connection.

    Thanked by 1blackjack4494
  • jsgjsg Member, Resident Benchmarker

    @MasonR said:
    Throwing more resources at something is hardly a solution if you don't understand the underlying issue. What resources are you maxing out (RAM, CPU, bandwidth)? Any enhancements (i.e. caching, optimizations, less DB queries, etc.) that can be done to lessen the resource burden?

    Once you understand the issue you're having a little more, then that should steer you in the direction to take. For instance, if you're maxing out memory on a 32GB VPS, then you won't want to buy a 32GB dedi as you've gained nothing from doing that. Likewise, if you're maxing out a 1Gbps connection, you won't want to get a dedi with a 1Gbps connection.

    Analysing and understanding ones situation? How dare you suggest that?!

    @msg7086
    My favourite colour is 42.

  • NanoG6NanoG6 Member

    The situation is similar to me weeks ago, when I decided to split my VM into two VMs equally. One for web server and the other for mysql server. Turns out it was the web server which hit the most and need more resource, while mysql server load was just around 0.2. So I put back everything into one server, easier to manage.

  • Guys, of course, analyzing is #1 factor. Im not asking about my specific situation though, Im asking which would be the best practise generally.

    But I understand that a generic answer cant be given.
    I also understand that cloud servers are based on nodes and resources are not completely owned, so its easier to have dependencies, failures, etc.

    Im mostly worried why Hetzner stops cloud server packages at this total number of resources, rather anything other :)

    I mean, why they do not offer a CX-61? Do they believe that if CX-51 exceeds its resources, its time for a dedi? And if yes, what about ensuring the redudancy of cloud?

  • There is no such thing as "resources" it's likely something specific that needs improvement. This could be the code, but even if you want to throw more hardware and keep your code as it is, you have to know which resource you'd need more.

    Is IO wait your problem? Is that you don't have enough RAM or that you need a better CPU? Depending on what if your problem the solutions available to you won't be the same: your question is currently too generic...

  • Im not searching for a solution to a problem - but thanks anyway!
    Sharing mostly my thoughts regarding limits of cloud services.

  • jsgjsg Member, Resident Benchmarker
    edited July 2018

    OK, I tell you my opinion.

    • Read your own OP again
    • Read your last post
    • Activate your brain. Then actually use it.
    • Extend your knowledge.
    • Focus and learn more on "what's your favourite [whatever] threads" before opening another one like this.

    42!

  • I asked for opinions that are related to the topic.
    Your bulletlist is triggering copyscape duplicate content from lifecoachwannabe.com.

  • @punkstar69 - Difficult to say, any idea where the bottleneck is at the moment?

    Whilst upgrading to a server with more resources would probably solve it, it's more of a temporary fix than anything else - If you're growing, you'll eventually hit a bottleneck with that server.

    Scaling using multiple VMs will probably be the better long-term solution overall, but the downside is a more complex solution.

    The more information we have, the better people can advise and assist.

  • @ElliotJ said:
    @punkstar69 - Difficult to say, any idea where the bottleneck is at the moment?

    Whilst upgrading to a server with more resources would probably solve it, it's more of a temporary fix than anything else - If you're growing, you'll eventually hit a bottleneck with that server.

    Scaling using multiple VMs will probably be the better long-term solution overall, but the downside is a more complex solution.

    The more information we have, the better people can advise and assist.

    Frontend is already optimized. Its a heavy WP website that receiving 1800-2200 concurrents, with 5gb~ mysql. I think that we have reached the limit of the available VMs on Hetzner's cloud.

    I assume that we are at a point we cannot further optimize the frontend and in the meantime, continuous backend work is being done, so having it crashed is not an option.

    To me, its obviously a resources run-out situation, working under PHP 7, MariaDB combined with Redis cache and Varnish for the frontend.

  • jsgjsg Member, Resident Benchmarker

    @punkstar69 said:
    I was always curious, what should be done once resources are reached.

    ... Whats your opinion guys?

    vs

    @punkstar69 said:
    Sharing mostly my thoughts regarding limits of cloud services.

    and

    @punkstar69 said:
    Guys, of course, analyzing is #1 factor. Im not asking about my specific situation though, Im asking which would be the best practise generally.

    There is no "best practise generally". Other than using ones brain and analysing ones situation.

    I mean, why they do not offer a CX-61?

    How would we know? Ask Hetzner. And being at it also ask for a CX-71 because some day the CX-61 won't suffice anymore.

    Do they believe that if CX-51 exceeds its resources, its time for a dedi? And if yes, what about ensuring the redudancy of cloud?

    Again: Ask Hetzner. Only they know. Looking at the CX-51 my guess is yes they supposedly (quite reasonably) think that beyond the CX-51 dedis should be used.

    As for redundancy, that can be achieved by multiple dedis, too.

  • punkstar69 said: Frontend is already optimized. Its a heavy WP website that receiving 1800-2200 concurrents, with 5gb~ mysql. I think that we have reached the limit of the available VMs on Hetzner's cloud.

    Ooft, fair enough. Considering the size of the DB, am I fair to assume it's heavy on the reads rather than writes? If so, that shouldn't be the limiting factor.

    If it's the case that it's read heavy, anonymous non-logged in users and whatnot, it might be an idea to have a look into nginx microcaching. Redis and varnish combined might be overthinking the issue, although, every situation is a lil' different.

  • punkstar69punkstar69 Member
    edited July 2018

    @ElliotJ said:

    punkstar69 said: Frontend is already optimized. Its a heavy WP website that receiving 1800-2200 concurrents, with 5gb~ mysql. I think that we have reached the limit of the available VMs on Hetzner's cloud.

    Ooft, fair enough. Considering the size of the DB, am I fair to assume it's heavy on the reads rather than writes? If so, that shouldn't be the limiting factor.

    If it's the case that it's read heavy, anonymous non-logged in users and whatnot, it might be an idea to have a look into nginx microcaching. Redis and varnish combined might be overthinking the issue, although, every situation is a lil' different.

    Well, cant tell - just because the nature of the website has no standard procedures. Can be 10 continuous hours of wp-admin posting things by the authors, but the next 10 might be a no-posting period. Overall data returns 56% read.

    We've implemented ESI as part of our tests in the past, but metrics showed that Varnish/Redis combination works better for us.

  • FHRFHR Member, Host Rep

    You could also split the services, e.g. run MySQL on VPS #1, run the rest on VPS #2.

  • @MasonR said:
    Throwing more resources at something is hardly a solution if you don't understand the underlying issue. What resources are you maxing out (RAM, CPU, bandwidth)? Any enhancements (i.e. caching, optimizations, less DB queries, etc.) that can be done to lessen the resource burden?

    That's it. You just need to know what resources you need. Depending on your case a dedi is more viable or two vms. Maybe you can have a master server and some minor slaves.

    @FHR said:
    You could also split the services, e.g. run MySQL on VPS #1, run the rest on VPS #2.

    This is another good point. Maybe you can split.

  • Sounds like,you need lots of RAM. MySql query cache and Varnish should fix all your problems.

  • @punkstar69 - Interesting! I'm fairly intrigued now, would you be happy to PM me with the website so I can have a little poke?

  • raindog308raindog308 Administrator, Veteran

    Two VMs, one dedi. I was just watching that on PornHub last night.

    Thanked by 1classy
  • @ElliotJ sure thing :)

Sign In or Register to comment.