xenforo.com server

[Anton_Bodryachkom]

I have question , here is info from litespeed site "

• APC (Alternative Opcode Cache) – Appears to be discontinued. Supports PHP up to version 5.4, but has been reported buggy with PHP 5.4. May be a good solution for PHP 5.3. APC has opcode cache as well as user/variable cache. Since PHP 5.5 and PHP 5.6 are unsupported by APC, xCache or Zend Opcache can be alternatives for op-code cache while APCu could be alternative to APC's user cache."

You guys would recommend xcache instead APC , xcache supports 5.6

 

[Marcus]

Memcache is supported by xf and generally recommended. There are different client variants available. You would have to install Libmemcache.

 

[Marcus]

I tried linode and ordered a dedicated cpu core. I got different - mixed - results from that server for the same php cpu test (which is included within the php distribution) multiple times. I came to the conclusion that the cpu core was not totally dedicated to me.

 

[Anton_Bodryachkom]

hmm thnx for sharing

 

[Moscato]

I have question , here is info from litespeed site "

• APC (Alternative Opcode Cache) – Appears to be discontinued. Supports PHP up to version 5.4, but has been reported buggy with PHP 5.4. May be a good solution for PHP 5.3. APC has opcode cache as well as user/variable cache. Since PHP 5.5 and PHP 5.6 are unsupported by APC, xCache or Zend Opcache can be alternatives for op-code cache while APCu could be alternative to APC's user cache."

You guys would recommend xcache instead APC , xcache supports 5.6

xcache is slower than zend opcache for PHP processing.

zend opcache is literally built into php 5.6 core, without being necessarily an addon, in my case, it was removing one # from a configuration file

apcu, (and I believe xcache as well) has some issues with session cache hash slamming, so I'd recommend memcached over them

libmemcached is faster than memcached, and is functionally the same

I actually personally use disk based xenforo caching, on tmpfs, since the minimum memcached size is 64MB, while going on tmpfs, I'm using under 5MB of ram. It's not entirely optimal for most use cases, but the forum is just a small subset of my website, and I need to conserve memory for the rest.

 

[Moscato]

I tried linode and ordered a dedicated cpu core. I got different - mixed - results from that server for the same php cpu test (which is included within the php distribution) multiple times. I came to the conclusion that the cpu core was not totally dedicated to me.

It's not 100% dedicated to you. Sometimes you can burst higher than the resources you're allowed, but you never, ever go below a certain minimum level, due to how xen resource allocation works.

Reasonably, you're being promised half of a hyperthreaded physical core. Sometimes you get more. You never get less.

 

[Marcus]

I rented 1 CPU core so I really think it should be dedicated to me. I did hear stories where some guys rented dedicated servers that were not really dedicated to them

I am with another provider now and it offered me one cpu core and I am the only one using it. The php benchmark test is always giving me the same results, unlike linode.

I think linode might be a very nice solution if you are start your forum and are not so experienced with the cloud. With linode you can put everything on your single VPS like database, computing, memcache, load balancing, firewall, storage. I split everything up into its own service, connected these and could not be happier.

 

[Moscato]

I rented 1 CPU core so I really think it should be dedicated to me. I did hear stories where some guys rented dedicated servers that were not really dedicated to them

I am with another provider now and it offered me one cpu core and I am the only one using it. The php benchmark test is always giving me the same results, unlike linode.

I think linode might be a very nice solution if you are start your forum and are not so experienced with the cloud. With linode you can put everything on your single VPS like database, computing, memcache, load balancing, firewall, storage. I split everything up into its own service, connected these and could not be happier.

The reality of it, if you want dedicated resources, you need to have a dedicated box. There is no cloud provider in the world that is offering what you're asking.

Unless you go with a company with terrible support (like OVH), that's about 80$ starting for low grade crap.

They run 20 core boxes with 40 logical cores due to hyperthreading. Putting 40 vcores worth of customers makes sense, else they can't fully utilize their hardware (the hyperthreading)

You're also sharing large raid 10 SSD IOPS, and memory bandwidth with other customers. Linode does not make you share outgoing networking at all, since everyone gets their own slice. Incoming bandwidth is shared, but it's from a 160gbit pool, so nobody cares.

That's just how cloud works.

Given that, linode has excellent support (I had some issues with ipv6, and they walked me through the fix even though it was a config issue on my end), and I've had no downtime at all, except for one prescheduled and notified 4am reboot with clean shutdown so they could do a critical security update

While linode is not as fast as dedicated resources, it's definitely fast for what you're paying.

 

[MattW]

I've had services with Linode for almost 2 years now, and I've never seen them offer dedicated CPU. Before their package upgrades last year, all servers had equal share of the available CPU cores, allowing for you to burst higher than your package. I had to request them to move one of VPS with them to a different node, because the one it was one was very laggy.
Now, you get an allocation of cores which again are still shared, but a lot better managed (they only put the same plans on the same nodes). I've not had any issues since their package restructure.

Anyone needing more than their $80 package could start looking at dedicated servers for the price of the next step up.

 

[Marcus]

It states 1 CPU core and also 1 GB RAM. I would be really disappointed if they do not give me the full core, or the full RAM. It's like I need to use 800 MB but I only have access to 200 MB as the other clients grabs my RAM.

So from my tests, linodes does not give you a real cpu core, you also state that. And maybe you also do not get the full RAM and Data Transfer, I have no idea. So for me, I can not really recommend them. But they might be nice for others !

 

[MattW]

And maybe you also do not get the full RAM

Yes they do

[root@host ~]# free -m
             total       used       free     shared    buffers     cached
Mem:          8017       4823       3194         50        197       4127
-/+ buffers/cache:        498       7519
Swap:          511          0        511
[root@host ~]#

That is an 8GB plan
You also get "access" to 6 CPU cores, not 6 dedicated cores.

[root@host ~]# cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 62
model name      : Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
stepping        : 4
microcode       : 0x416
cpu MHz         : 2800.044
cache size      : 25600 KB
physical id     : 0
siblings        : 6
core id         : 3
cpu cores       : 1
apicid          : 7
initial apicid  : 7
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu de tsc msr pae cx8 apic sep cmov pat clflush mmx fxsr sse sse2 ss ht syscall nx lm constant_tsc rep_good nopl nonstop_tsc eagerfpu pni pclmulqdq ssse3 cx16 sse4_1 sse4_2 popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm ida arat epb pln pts dtherm fsgsbase erms xsaveopt
bugs            :
bogomips        : 5602.42
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

processor       : 1
vendor_id       : GenuineIntel
cpu family      : 6
model           : 62
model name      : Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
stepping        : 4
microcode       : 0x416
cpu MHz         : 2800.044
cache size      : 25600 KB
physical id     : 0
siblings        : 6
core id         : 3
cpu cores       : 1
apicid          : 7
initial apicid  : 7
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu de tsc msr pae cx8 apic sep cmov pat clflush mmx fxsr sse sse2 ss ht syscall nx lm constant_tsc rep_good nopl nonstop_tsc eagerfpu pni pclmulqdq ssse3 cx16 sse4_1 sse4_2 popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm ida arat epb pln pts dtherm fsgsbase erms xsaveopt
bugs            :
bogomips        : 5602.42
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

processor       : 2
vendor_id       : GenuineIntel
cpu family      : 6
model           : 62
model name      : Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
stepping        : 4
microcode       : 0x416
cpu MHz         : 2800.044
cache size      : 25600 KB
physical id     : 0
siblings        : 6
core id         : 3
cpu cores       : 1
apicid          : 7
initial apicid  : 7
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu de tsc msr pae cx8 apic sep cmov pat clflush mmx fxsr sse sse2 ss ht syscall nx lm constant_tsc rep_good nopl nonstop_tsc eagerfpu pni pclmulqdq ssse3 cx16 sse4_1 sse4_2 popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm ida arat epb pln pts dtherm fsgsbase erms xsaveopt
bugs            :
bogomips        : 5602.42
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

processor       : 3
vendor_id       : GenuineIntel
cpu family      : 6
model           : 62
model name      : Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
stepping        : 4
microcode       : 0x416
cpu MHz         : 2800.044
cache size      : 25600 KB
physical id     : 0
siblings        : 6
core id         : 3
cpu cores       : 1
apicid          : 7
initial apicid  : 7
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu de tsc msr pae cx8 apic sep cmov pat clflush mmx fxsr sse sse2 ss ht syscall nx lm constant_tsc rep_good nopl nonstop_tsc eagerfpu pni pclmulqdq ssse3 cx16 sse4_1 sse4_2 popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm ida arat epb pln pts dtherm fsgsbase erms xsaveopt
bugs            :
bogomips        : 5602.42
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

processor       : 4
vendor_id       : GenuineIntel
cpu family      : 6
model           : 62
model name      : Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
stepping        : 4
microcode       : 0x416
cpu MHz         : 2800.044
cache size      : 25600 KB
physical id     : 0
siblings        : 6
core id         : 3
cpu cores       : 1
apicid          : 7
initial apicid  : 7
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu de tsc msr pae cx8 apic sep cmov pat clflush mmx fxsr sse sse2 ss ht syscall nx lm constant_tsc rep_good nopl nonstop_tsc eagerfpu pni pclmulqdq ssse3 cx16 sse4_1 sse4_2 popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm ida arat epb pln pts dtherm fsgsbase erms xsaveopt
bugs            :
bogomips        : 5602.42
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

processor       : 5
vendor_id       : GenuineIntel
cpu family      : 6
model           : 62
model name      : Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
stepping        : 4
microcode       : 0x416
cpu MHz         : 2800.044
cache size      : 25600 KB
physical id     : 0
siblings        : 6
core id         : 3
cpu cores       : 1
apicid          : 7
initial apicid  : 7
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu de tsc msr pae cx8 apic sep cmov pat clflush mmx fxsr sse sse2 ss ht syscall nx lm constant_tsc rep_good nopl nonstop_tsc eagerfpu pni pclmulqdq ssse3 cx16 sse4_1 sse4_2 popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm ida arat epb pln pts dtherm fsgsbase erms xsaveopt
bugs            :
bogomips        : 5602.42
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

[root@host ~]#

 

[Marcus]

Maybe linode also gives you access to 8GB Ram but you have to share it lol.

Seriously, I am with the largest cloud now and it provides me with dedicated resources like 1 cpu core, 1 gb ram. I don't have to share them, but these resources are somewhat "cloudy" meaning they might disappear from time to time It's a totally different system when you are in the cloud or rent your own server. I think both is really good. I had a nice server before, 64 GB RAM, 12 Cores and honestly this was totally overkill.

Edit: I think you want to say that linodes actually gives the customer one full cpu core. I read it first that linode gives you a core, but you might share it with another customer

 

[Anton_Bodryachkom]

I can tell this , we was with siteground Cloud server , 1 core 3 gigs ( 240$ month managed) . and performance its not good at all , idk why , tech guys from Sitegound saying its something with soft ( xenoforo ) , so now we switched to Linode for testing and successfully transfered all data to 2 core 2 ram ( for now ) and i can tell its already works faster . on Litespeed . here is some graphs when site on siteground with memcache . May be we just have badluck with them

 

[Moscato]

Maybe linode also gives you access to 8GB Ram but you have to share it lol.

Seriously, I am with the largest cloud now and it provides me with dedicated resources like 1 cpu core, 1 gb ram. I don't have to share them, but these resources are somewhat "cloudy" meaning they might disappear from time to time It's a totally different system when you are in the cloud or rent your own server. I think both is really good. I had a nice server before, 64 GB RAM, 12 Cores and honestly this was totally overkill.

Edit: I think you want to say that linodes actually gives the customer one full cpu core. I read it first that linode gives you a core, but you might share it with another customer

You probably should learn how virtualization software works.

You aren't getting 1 full core.

The virtualization software just doesn't work like that.

Your VPS creates threads.

Those threads are placed on a vcore.

The underlying virtualization platform sees the whole vcore as a thread, and uses the cpu scheduler to place the various threads on whatever core it feels is most useful for the job based off load on other threads.

The reason why you don't lock is because cores are hyperthreaded. Because of hyperthreading, if they do lock, they're losing 1/3rd of the total capacity of the box.

There are 40 logical cores on a linode box, with 20 physical cores.

If they only allowed 20 vcores, the other 20 logical cores would be wasted.

It's not financially reasonable. They have to maintain a network (hundreds of gigabit here), have support staff, have a security team, and you want a 100% dedicated to you core.

Not gonna work at low price tags.

http://ark.intel.com/products/75277/Intel-Xeon-Processor-E5-2680-v2-25M-Cache-2_80-GHz

They have 2 of these per box.

A linode server probably costs about 8000$ (including 40gbit adapters), and that doesn't even count the rest of the datacenter. You want them to charge 200 a month per box? Not gonna work.

Until you need to exceed their 40$ plan, linode is the fastest host around for the money

 

[WSWD]

View attachment 101341
It states 1 CPU core and also 1 GB RAM. I would be really disappointed if they do not give me the full core, or the full RAM.

You're not understanding how it works. You aren't going to get a dedicated core at those prices.

To make things simple, let's say the host rents a E3 server (4 cores) for VPSs. It costs the host $200/mo. If you want 1 dedicated core, it's going to cost you at least $50/mo., assuming the host makes no profit. Same with the 20 core servers.

 

[Moscato]

20 core servers actually cost more per core, which is why so many crappy hosting companies use e3 chips

thing is, they're more resistant to cpu hogging, which is why they're better

 

[WSWD]

20 core servers actually cost more per core, which is why so many crappy hosting companies use e3 chips

thing is, they're more resistant to cpu hogging, which is why they're better

We actually like the E3's a lot for SSD applications, because we can provide much better disk i/o. We can run a very limited number of containers on those systems and still make money, where as on a 20+ core box, you're going to have to have dozens of containers to make a profit. Dozens of containers kills disk i/o.

 

[Moscato]

We actually like the E3's a lot for SSD applications, because we can provide much better disk i/o. We can run a very limited number of containers on those systems and still make money, where as on a 20+ core box, you're going to have to have dozens of containers to make a profit. Dozens of containers kills disk i/o.

Couldn't you just run raid 10 SSD with more disks?

Though I suppose that has limits to your controller

e3 are cheaper per unit of cpu computation as well

 

[Marcus]

Moscato thanks for your detailled inside look into how servers are setup. It is really interesting!

My current hoster charges 16.000 USD / year for one single server with 18 full core E5-2666 v3 (36 vcore), and this price is without bandwidth, service, contract. I guess it looks like a sustainable business model. But yeah its not linode I tested their cores where hyperthreading is disabled and they are really dedicated - well at least they were the same speed everytime I tested them, linode was very volatile

 

[Moscato]

Moscato thanks for your detailled inside look into how servers are setup. It is really interesting!

My current hoster charges 16.000 USD / year for one single server with 18 full core E5-2666 v3 (36 vcore), and this price is without bandwidth, service, contract. I guess it looks like a sustainable business model. But yeah its not linode I tested their cores where hyperthreading is disabled and they are really dedicated - well at least they were the same speed everytime I tested them, linode was very volatile

It's entirely possible that they have a cpu timeshare limit on your vcores that they don't allow you to exceed at all

It'll seem 100% consistent, but actually be worse than the slightly more volatile conditions linode offers

Linode will offer an equal or better baseline, it just will have spikes of performance exceeding the other options.