Building web server

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Jaguaras

New Member
Aug 27, 2020
12
0
1
Hi,

I am planning to get a server for Wordpress + Woocommerce = 100k product. And I decided that now a days motherboard must have at least socket 3647, so I could upgrade with much more RAM & additional CPU. I am looking for a DELL Tower T440, it has socket 3647, 4210 SILVER CPU 2.2Ghz, 16Gb + 16Gb RAM, 2x240GB SSD, Raid controller is H330.

And I have some questions:

1. Raid controller H330 is it really bad controller, should I buy a H730P version? I know that H330 has no cache.
2. Is this a good choice a T440 with low-end CPU?
3. I am confused about maximum RAM, it shows RDIMM 512GB, LRDIMM 1TB. What is a difference between them?

Overall what is your opinion about this server and what would change or you would buy different custom server?

Thanks in advance!
 

vladimir.mijatovic

New Member
Jan 12, 2019
27
15
3
We run a PHP-based SaaS and also host couple of Wordpress sites.
imho Xeon silver (and all Xeons with low base clock speeds) might be a very poor choice for a WooCommerce (or Magento) site as you will need a significant portion of non-cached performance (shop, cart, admin, stock...).

PHP is a single thread process and CPU clock speed matters a lot and in most SMB webshop cases matter much more that core count.
Just finished writing up for our blog on subject like this
Impact of CPU speed on websites running PHP / Development - Klik-MALL.com
and started creating a list of best CPUs for PHP
Which are the fastest CPUs for PHP websites like Wordpress and Magento / Development - Klik-MALL.com

Best CPU for PHP? i9-9900K and its close relatives. But utilizing an i9-9900K is a bit of a project itself.

Get a W2145, W-2245 or W-2255 if you need lots of cores, pcie lanes (though from your config this does not seem to be the case) and 128GB+ RAM.

I'd argue that for your project Xeon E-2246G or E-2278G could be a far better choice esp. since you are getting only 32GB RAM.
Get an Intel Optane for DB and put it into one slot. When your CDN/media folder outgrows Optane 480GB-1.5TB you can go with additional NVMe put into other pcie slot and have a great all-in-one setup for pretty much the same money.

WooCommerce with 100k products could use quite some "full page cache" storage. Having an Optane or ie Samsung NVMe for this would give you a lot more storage than Redis (RAM) you probably plan to have for a negliable difference in performance. And you keep your RAM for DB and cached queries.
We host Spletni vodič po Sloveniji - Slotrips.si (2-3k webpage requests/day) with FPCache on 970 EVO+ and getting 10-100ms Server response time via Google Analytics with the fact that we serve it inside PHP, not with a proxy like Varnish or NGINX server-fpc setup. TTFB via LAN is ~3ms :) Some visitors might happen to have a bigger part of that TTFB delay due to the distance from server (SI/EU) as the site also has users from AU, US, JPN... than by server making a response.


Potential considerations.
BAD:
- There is basically 0-reduncancy in this setup,
GOOD:
- it might run circles around your config,
- it's a simpler hardware and software hosting setup,
- I'd trust a single Intel Optane more than 2 SATA SSDs in RAID via Dell's entry level RAID card.
I have 2x H310 (even flashed one to IT mode for ZFS usage), 1xH710p - they all work great. But every time I see @Patrick or guys like Ally Malventano write about Optane it just feels much better.
If you decide to go with Dell Raid card - I'd get a H730p as vendors will always put some artificial limitations on lower end components about which you find out later (H310 had notoriously low QD ~30 while that same original LSI card/firmware had it at 600 - that was the reason for flashing it while having a spare in case something goes wrong)

From your config there is nothing keeping you from getting a 100% faster 6-8c CPU for PHP tasks instead of that Silver Xeon.
Your TTFB o uncached pages will be much better and that Silver provides almost zero added value in case of huge traffics.
 
  • Like
Reactions: Jaguaras

vladimir.mijatovic

New Member
Jan 12, 2019
27
15
3
any particular reason for this DIY and not use a hosted solution like WooCommerce Hosting - WordPress Ecommerce - Bluehost ?
Do you have N+1 Internet access, Power, Cooling, Physical security?
There are without a doubt many benefits of "quick deploy, cheap, managed, redundant power,..." hosting packages but you pay with a performance-penalty. How big is a case by case.

My guess is someone with 100k products WooCommerce store has already tried all the benefits of "Shared hostings" with "Unlimited Online Stores, Unmetered SSD Storage" without any indication what hardware it runs on. Or more probably is already on a small VPS or dedicated server, trying to improve with his online business blooming. @Jaguaras what is your case?

This is how just a CPU affects WooCommerce on a shared hosting.
It's best seen trough TTFB timings, esp. the part "server response time".
For PHP/WooCommerce you often need instant un-cached performance (producing results with TTFB < 1.5sec as recommended by Google and that already seems a bit laggy).
- shared hosting providers prefer many cores, low power CPUs and this results in CPUs running at (very) low frequency. Your website could easily take 0.5+ sec more in TTFB than with a better-suited CPU based on just their turbo-speeds.
- it will probably take another 0.5+ longer as a shared CPU will be serving your PHP at base freq. due to constant load while a dedicated one might have bursts serving them at or close to turbo freqs.
- you will probably be in some sort of a virtualized environment. Our tests showed that even in best case scenario (LXC, privileged, no-mitigations, server in performance mode) loss of PHP performance was ~10-15% per PHPBench.
- being on a shared hosting means your requests will wait in queue for CPU and disk resources along with (many) other requests.
- Other people's PHP code may not even be optimized or you having "noisy neighbours" means more waiting time for CPU to free up and take your webpage request.

These things add up.
Add bottlenecks due to disk and RAM and you will be looking at 1-2 sec of extra TTFB waiting time per each click as a wishful scenario.
1 sec extra is noticable on UX, SEO...
2+ of extra waiting time (ie 3.5 instead of 1.5) after each click starts driving part of your customers away.
Mind you that was all just "server response time" as part of TTFB, not a fully loaded page that fits "Time to Interactive" kpi.
Will not quote Amazon how much they loose with each extra sec of loading time as nobody is close to them but in 2020 "nobody likes a slow page".

How bad can it get?
- I was witnessing a Magento frontend dev waiting for 10sec+ after every press on SAVE button to cloudEdit/cssTemplateFile. Then waited 5+ sec to be able to refresh and check changes he'd made. It was still a dev project, on Amazon cloud, one of cheaper packages, with a promise from project-leader "when we go production and if needed, we spend more". We literally had mini conversations during loading times.
- recently got an inquiry by a WooCommerce owner stating that their /cart page would too often take up to 13 secs to load.
- This comment is a really nice example how this particular CPU/PHP issue often looks in real life:


If you'd say "Want a car, for min. money, as long as it's moving"
You kow what will you get? A bus ticket. You will get to yoor destination and not have to worry about gas, tires or a parking spot.
That is enough for many.

Are shared hostings good for most WP blogs that can serve cached pages? Yes.
Is there a difference between workloads for Wordpress blog and WooCommerce shop? Yes.
Is there a significant difference between shared and a dedicated server? Yes.
Can selected components significantly affect performance of my dedicated server running PHP? Yes.

There are not many shortcuts to a fast PHP/Woo webshop.
 

Jaguaras

New Member
Aug 27, 2020
12
0
1
Hi, thank you for such informative reply!

I learned from your articles so much. Now I understand that it is very important Ghz / Core number.

We are considering for buying from Dell for better warranty / quality parts. But Dell Server motherboards have no socket 2066, so we can't get W series CPU's.

We have 2 options Xenon 5222 or AMD EPYC 7262, I personally think AMD would be near W2145 model. But Intel 5222 has better performance on CPU Single thread test, but just 4 cores. AMD has 23% less performance on CPU Single thread test but has 8 cores. Would be great to get your opinion about that. But I assume you would be more on Intel 5222 correct?

I'd argue that for your project Xeon E-2246G or E-2278G could be a far better choice esp. since you are getting only 32GB RAM.
Get an Intel Optane for DB and put it into one slot. When your CDN/media folder outgrows Optane 480GB-1.5TB you can go with additional NVMe put into other pcie slot and have a great all-in-one setup for pretty much the same money.
About disk space, we don't have much media, at the moment it only stores 10Gb of media. So does it needs intel optane or am I understand wrong about this beast? I thought that Boss controller + 2 x M.2 240Gb with S140 RAID1 would be best for performance. S140 this is software RAID on DELL. And only for start to have a 32Gb of RAM.

WooCommerce with 100k products could use quite some "full page cache" storage. Having an Optane or ie Samsung NVMe for this would give you a lot more storage than Redis (RAM) you probably plan to have for a negliable difference in performance. And you keep your RAM for DB and cached queries.
Yes you are right, we want to get best performance for WP+WOO website. So you mean if we would have 2x M.2 SSD it would be better then RAM Redis storage?
My guess is someone with 100k products WooCommerce store has already tried all the benefits of "Shared hostings" with "Unlimited Online Stores, Unmetered SSD Storage" without any indication what hardware it runs on. Or more probably is already on a small VPS or dedicated server, trying to improve with his online business blooming. @Jaguaras what is your case?
Yes we tried Cloud hosting where 16Gb of RAM's, 6 virtual core CPU, and APACHE :) After we moved to Litespeed webserver we saw a good performance so it means we have pretty normal coding, but we need to make server just for WP+WOO combination for the best performance, and server would be pay off after 3 yrs. And we want to test for many variantions like Nginx web server, objective cashing, Litespeed webserver + WP plugin. What would be your suggestion about best software for WP+WOO combination?

Thank you in advance!
 
Last edited:

BoredSysadmin

Not affiliated with Maxell
Mar 2, 2019
1,050
437
83
There are without a doubt many benefits of "quick deploy, cheap, managed, redundant power,..." hosting packages but you pay with a performance-penalty. How big is a case by case.

My guess is someone with 100k products WooCommerce store has already tried all the benefits of "Shared hostings" with "Unlimited Online Stores, Unmetered SSD Storage" without any indication what hardware it runs on. Or more probably is already on a small VPS or dedicated server, trying to improve with his online business blooming. @Jaguaras what is your case?

This is how just a CPU affects WooCommerce on a shared hosting.
It's best seen trough TTFB timings, esp. the part "server response time".
For PHP/WooCommerce you often need instant un-cached performance (producing results with TTFB < 1.5sec as recommended by Google and that already seems a bit laggy).
- shared hosting providers prefer many cores, low power CPUs and this results in CPUs running at (very) low frequency. Your website could easily take 0.5+ sec more in TTFB than with a better-suited CPU based on just their turbo-speeds.
- it will probably take another 0.5+ longer as a shared CPU will be serving your PHP at base freq. due to constant load while a dedicated one might have bursts serving them at or close to turbo freqs.
- you will probably be in some sort of a virtualized environment. Our tests showed that even in best case scenario (LXC, privileged, no-mitigations, server in performance mode) loss of PHP performance was ~10-15% per PHPBench.
- being on a shared hosting means your requests will wait in queue for CPU and disk resources along with (many) other requests.
- Other people's PHP code may not even be optimized or you having "noisy neighbours" means more waiting time for CPU to free up and take your webpage request.

These things add up.
Add bottlenecks due to disk and RAM and you will be looking at 1-2 sec of extra TTFB waiting time per each click as a wishful scenario.
1 sec extra is noticable on UX, SEO...
2+ of extra waiting time (ie 3.5 instead of 1.5) after each click starts driving part of your customers away.
Mind you that was all just "server response time" as part of TTFB, not a fully loaded page that fits "Time to Interactive" kpi.
Will not quote Amazon how much they loose with each extra sec of loading time as nobody is close to them but in 2020 "nobody likes a slow page".

How bad can it get?
- I was witnessing a Magento frontend dev waiting for 10sec+ after every press on SAVE button to cloudEdit/cssTemplateFile. Then waited 5+ sec to be able to refresh and check changes he'd made. It was still a dev project, on Amazon cloud, one of cheaper packages, with a promise from project-leader "when we go production and if needed, we spend more". We literally had mini conversations during loading times.
- recently got an inquiry by a WooCommerce owner stating that their /cart page would too often take up to 13 secs to load.
- This comment is a really nice example how this particular CPU/PHP issue often looks in real life:


If you'd say "Want a car, for min. money, as long as it's moving"
You kow what will you get? A bus ticket. You will get to yoor destination and not have to worry about gas, tires or a parking spot.
That is enough for many.

Are shared hostings good for most WP blogs that can serve cached pages? Yes.
Is there a difference between workloads for Wordpress blog and WooCommerce shop? Yes.
Is there a significant difference between shared and a dedicated server? Yes.
Can selected components significantly affect performance of my dedicated server running PHP? Yes.

There are not many shortcuts to a fast PHP/Woo webshop.
Valid points. My point is that running a production e-store on a single physical server seems to me (as IT engineer) is a silly idea, no matter how beefy it is. I really don't know much about Woo, but I do know what for best performance and security the (in-memory?) database should be isolated and secured. There should be multi-tiers and multiple servers to provide both HA and scale-out performance.
I 100% see you point what running medium to large e-store on share hosted (SaaS) environment isn't optimal, but there are still ways to engineer this on PaaS or even IaaS environments.
These articles share some ideas:
 
Last edited:

Jaguaras

New Member
Aug 27, 2020
12
0
1
any particular reason for this DIY and not use a hosted solution like WooCommerce Hosting - WordPress Ecommerce - Bluehost ?
Do you have N+1 Internet access, Power, Cooling, Physical security?
Our main customers are from one country and we don't have good hosting companies which provide good hardware + software combination for server. So we decided to get our own web server so it could payoff in 3 yrs. Could you explain about N+1 internet access? Power you mean UPS? Cooling I guess rack or tower cooling system would be fine for 1 server? Physical security what do you mean? It has alarm security. Where in 1-2 min arrives security guards :)
 

BoredSysadmin

Not affiliated with Maxell
Mar 2, 2019
1,050
437
83
N+1 means you have at least one more network path redundant in case the primary network fails. For power - I mean two independent (both UPS protected) power source. Cooling I mean providing sufficient cooling for the servers, and ideally have it redundant as well. Physical security, few layers of security (locked rack) locked room, locked outside the fence with security cameras, motion sensors, and last not least security guards.

I know that IT IS VERY expensive to provide all of these, so most people would use one of the existing colocation datacenters to locate the servers in it. Most colo will allow you to bring your own servers, but then you have pay provide all the networking and data security equipment (firewalls/IPS/ids, etc) or you could have a hosting company deal with all that, you just rent servers (shared or dedicated)
You could also probably design and build very high performing e-site on AWS (or GCP or Azure)
 
  • Like
Reactions: vladimir.mijatovic

amalurk

Active Member
Dec 16, 2016
311
116
43
102
How many pageviews do you expect a month? Where is the primary audience? Seems overkill for under 5mil. I would stick on geographically nearest good VPS provider.
 

vladimir.mijatovic

New Member
Jan 12, 2019
27
15
3
This will be helpful, probably to many looking at self-hosting WP, Magento etc.

Just finished and published our analysis about PHP performance based on CPU SKU
2 new PHP performance numbers for a better estimate of real-life performance by a specific CPU / Development - Klik-MALL.com
and think it's a very decent visualisation of what is going on serving PHP.
I also smiled when I saw STH take on Xeon Gold https://www.servethehome.com/intel-xeon-gold-6258r-benchmarks-and-review/ as they wrote the exact location of that CPU on our map.

@Jaguaras
Check this
AMD EPYC 7262 vs Intel Xeon Gold 5222 @ 3.80GHz vs Intel Xeon E-2278G @ 3.40GHz [cpubenchmark.net] by PassMark Software
If Dell has anything on that Xeon E-22xx lineup, that would be my choice. This CPU is not in our chart as I did not find those PHPBench scores I saw, but they were right next to 9900K. It could save you some money as it is a "cheaper" platform.
Second choice for WooCommerce - Gold 5222 rather than that Epyc. CPU will not be a bottleneck serving cached web pages, but a slower CPU can be "noticeably slower" on uncached pages (cart, admin, refreshing cached page...).

Going E-2278g might save you enough to get that BOSS 2xSSD and a separate Intel Optane 480GB+ drive to put into. It's not that expensive. I like it as it tells me that if there is not enough performance, it's probably something with the code and not hardware. We do not have nearly the traffic to push this hardware to its limits, but that info to me (as I am also a developer of that SaaS) is totally worth it. BOSS with 2xSSD will do just fine, too.

I saw Litespeed before and is on my to-do list. We were able to solve our customers' speed issues with NGINX server side cache + NGINX WP plugin and it works great. Reduced TTFB from 1.75s on i9-9900K (best case possible) to < 50ms . Its a quite heavy WP/elementor site Slotrips travel agency | Authentic Hiking, Walking & Biking Trips in Slovenia. Serving Full Page Cache from Intel optane 900p, Samsung EVO 970+ or even ZFS shared storage does not make much matter - it's fast enough.
This NGINX setup we kept as the customer told us they tried a few caching plugins and they did not work correctly or made errors. NGINX setup runs before the request even touches WP/PHP and that cache delete/purge plugin works fine.

@BoredSysadmin points on DC setup and equipment are something to keep in mind. Well written. We do not cover them well enough and have "paid" with a short downtime just yesterday due to outage of our ISP. A plan to mitigate this is multi-layered. I just try to keep it real - we are not that big and important to be frustrated about when an outage of one of bigger national ISP's happens (which was a first that involved us). I saw IBM cloud in AU went offline yesterday, too. :- ) There are limits to what we can aford and even want to do. Figuring out a combination of a nearby failover server, a failover WAN with a separate internet provider, aws DNS route53, smaller temporary hosting on another continent... is on the list. Our uptime is 99.9%+ and this will add to that. But our task involves much more than just serving cached pages otherwise we'd already done it.

For us atm the focus is more on understanding the workload, available technology in our price range. All optane DB server and new Threadripper 3995wx would solve a lot of questions just with a wallet :) But it would also be a bit more boring :)
For now we are here
https://www.klik-mall.com/web/blog/...rformance-webhosting-on-a-php-lemp-stack/5081
 

Attachments

Jaguaras

New Member
Aug 27, 2020
12
0
1
@Jaguaras
Check this
AMD EPYC 7262 vs Intel Xeon Gold 5222 @ 3.80GHz vs Intel Xeon E-2278G @ 3.40GHz [cpubenchmark.net] by PassMark Software
If Dell has anything on that Xeon E-22xx lineup, that would be my choice. This CPU is not in our chart as I did not find those PHPBench scores I saw, but they were right next to 9900K. It could save you some money as it is a "cheaper" platform.
Second choice for WooCommerce - Gold 5222 rather than that Epyc. CPU will not be a bottleneck serving cached web pages, but a slower CPU can be "noticeably slower" on uncached pages (cart, admin, refreshing cached page...).
Thank you Vladimir for valid points.

Ok so AMD we are putting aside. What if we only have intel socket 3647 and options are: Gold 5222, Gold 5217, Silver 4215R (3.2Ghz / core )
I guess 8 core is a middle point having 3 - 4 Ghz / core?
 

Jaguaras

New Member
Aug 27, 2020
12
0
1
How many pageviews do you expect a month? Where is the primary audience? Seems overkill for under 5mil. I would stick on geographically nearest good VPS provider.
We have specific region and we don't have a good provider for good price. It will be definitely under 5mil pageview but we have pretty great TTFB delay... so this could be a problem of CPU.
 

PaintChips

New Member
Sep 10, 2020
7
10
3
I mostly use a server in a home environment for a test/pre-live duty, I would note on the previous gen PowerEdge the S130 software RAID the performance hit varied upon how active your usage of databases/disk I/O--the H330 provided hardware RAID but lack of cache made it worthless if the server was equipped with more than 4 drives or you really pushed the disk usage too much. I will note I've used a 6 drive setup with a H730 and its extremely zippy even with a mix of SSD+10k spinning rust. On the current PowerEdge models some have one or two NVME slots if raw speed is a must.

I've moved away from being a web developer but every so often I have to deal with a Wordpress matter--back in my student days, using "extra storage/bandwidth capacity" opened the door for some podcasting or fun audio streaming of midi/soundfont created music :D (modern times, some work environments allowed running multiplayer game servers or teamspeak/mumble to avoid using Skype)
 

amalurk

Active Member
Dec 16, 2016
311
116
43
102
You took the average score from Phronnix openbenchmarking website then multiplied by a ratio of base freq/ to turbo and call that the estimated real case score? Seriously? And we should base buying decisions on that? Well duh faster cores are faster but you didn't show there is any real-world difference in TTFB between say a Sliver 4108 and 9900k.
 

Jaguaras

New Member
Aug 27, 2020
12
0
1
You took the average score from Phronnix openbenchmarking website then multiplied by a ratio of base freq/ to turbo and call that the estimated real case score? Seriously? And we should base buying decisions on that? Well duh faster cores are faster but you didn't show there is any real-world difference in TTFB between say a Sliver 4108 and 9900k.
No we don't look much at scores of benchmarks, our goal to make fast server for given budget, because our plans to have more websites. But you could share how you are calculating what type and resources server needs for website.
 

amalurk

Active Member
Dec 16, 2016
311
116
43
102
No we don't look much at scores of benchmarks, our goal to make fast server for given budget, because our plans to have more websites. But you could share how you are calculating what type and resources server needs for website.
In my experience, the return on trying to optimize from a good TTFB to an awesome TTFB is not there, especially if your plan is to do it with a large capital purchase.

There is however huge return on optimizing WordPress themes, the scripts used in the theme and caching and CDN for images. Like if you compare a GeneratePress based theme to one based on one of the slow code builders or a poorly coded jquery heavy theme, it is night and day on the time the page is ready for the user to interact. Same for caching vs no caching.

Don't get me wrong, I am interested in a fast TTFB but that supposed white paper above on processors showed nothing. It just took someone else's scores and multiplied that by a processor frequency ratio without any actual testing of TTFB. If you want a globally fast TTFB on a budget, caching with a Cloudflare KV worker is pretty awesome. Also, Cloudflare's free DNS is very fast compared to other offerings. I am trying to find but haven't, a prior paper I saw that did actual testing on ttfb with different TLS configs and OS and webserver stacks and that is also very likely a bigger difference than a mid vs high freq processor but I cannot seem to find it right now.

So I would say do more research before you buy an expensive server thinking that will make your site fast especially for one that isn' t that high traffic. Software stack and WordPress theme optimization will make a much bigger difference in a fast user experience.
 

vladimir.mijatovic

New Member
Jan 12, 2019
27
15
3
You took the average score from Phronnix openbenchmarking website then multiplied by a ratio of base freq/ to turbo and call that the estimated real case score? Seriously? And we should base buying decisions on that? Well duh faster cores are faster but you didn't show there is any real-world difference in TTFB between say a Sliver 4108 and 9900k.

First - you can tackle TTFB on 2 fronts: hardware and software.
This discussion is mostly for hardware as software brings up a lot more to the table (but will reply to that in a separate post). The assumption is that @Jaguaras will keep their shop on Wordpress/WooCommerce and changing the platform is not an option.

Second - serving fully cached blog pages (FPC) as static responses compared to Webshop tasks are two totally different beasts. Apache bench results for one Wordpress site we host:
- With FPC: 1.200 reqs/sec (TTFB < 20ms without moving the CPU usage much)
- No-FPC: 11-12 reqs/sec (TTFB > 1.15sec on a i9-9900K pretty much maxing the available cores when PHP workers tuned right).
FPC storage was not even on Redis, but on a Samsung EVO 970+ NVMe. For FPC you do not need a lot of CPU muscles. When you see truly low TTFB (instant loads) it's 99% fully cached or even an amp page.

I compare this task (FPC vs. non-FPC) to 2 math questions:
1. What is "5 x 7"?
2. What is "23.5 x 7845 - 50.58"?
One is a pre-prepared answer that can be answered by a child and is returned almost as a "picture" from memory (the fastest, but also most limited storage like RAM) while the other needs processing. How much time is needed for getting the result? That depends on the "CPU".

Disqussing tradeoffs of FPC and what WooCommerce requires is even more off topic. What is without doubt is that any webshop needs solid non-FPC performance.

My analysis was focused on "What CPU will process the same PHP code for Non-Cached pages faster" and "all other things equal, what should I get". As you have to run your site on some CPU, it's a good practice to try to pick the best tool for the job.


Chart and table can serve as a framework for PHP processing power as they (imho) show a lot of usefull info:

- Chart puts PHP[CPU] landscape of hundreds of SKUs into some perspective. Going up improves TTFB, going right will add user/requests concurrency. (I decided to size up the bubbles based on cores rather than $$$ as prices of CPUs can vary a lot over time).

- It also puts SKUs into some relative position to each other.

- Nobody should have an issue to say up until sept. 2020 that i9-9900k and its close relatives are the fastest CPUs for PHP. I find it usefull saying to customer "This is the fastest this site can run. Is it good enough or you gonna remove some plugins or open other options?".

- But saying i9-9900K is the single best solution is not entirely correct as there are cases with lots of paralel requests where a high-core count could provide a better AVERAGE TTFB than a single 8/16 CPU. This is not visible from PHPBench scoreboard, but it is from this chart. Finding those break even points on TTFB vs. Paralel reqs. charts will be the next step for some.

- @amalurk said "You took the average score from Phronnix openbenchmarking website then multiplied by a ratio of base freq/ to turbo and call that the estimated real case score? Seriously?"
Real score will depend on a lot more than just a CPU. But it's an estimate of "processing time required in relative terms among CPUs".
We have been using 3x dual E5-2640, dual E5-2690, E5-1650v4 and i9-9900k for couple of years. PHPBench and Passmark single thread scores were reflected on multiple PHP sites, frameworks, internal benchmarks... PHPBench has its real-life value. But I think this score is only part of the real case story.

I'd be curious if you think it'd better to just take the official PHPBench score and multiply it with core count? Or maybe with a thread count? Maybe with an all-core turbo speed would be better than base, but that number is something Intel started hiding and I am not sure CPU gets to those speeds with short-lasting bursts that PHP requests require. But it is clear that opening freq. is CPUs base freq. That is why WS CPUs feel snappier for PHP websites.

- The table is potentially detecting "trap CPUs" with calculated "Estimated single core PHPBench perf." as some CPUs can look good in synthetic PHPBench benchmark with idle computer running a single core at turbo freq. But its low base clocks can be devastating for TTFB on heavy PHP sites. Xeon Silver 4108 has +56% PHPbench relative to E5-2640, but an estimate single core in "real case" is that it would have only 12% better performance on a busy server. It could be even less than that @1.8GHz. So if you ask should I take Xeon Silver 4108 or 9900K the answer for PHP is to take i9. Even more - never take a Silver or let alone Bronze Xeon for PHP. There are many better and cheaper options for this task.

I do not have the option to test all CPUs. I just tried to make some conclusions how PHP workload operates and how this might affect sites based on the hosting environment.
- It also reveals a bit more what to expect with shared, cheap packages by Webhosting providers. Their interest is more in line with "the most cores for the least electricity (read low freq.)" than "let's provide great TTFB". Partly understandable as one site with unoptimized code and tons of plugins might jam those 4-6 cores for 30sec (difined limited time by server settings) before returning Err message while all other clients will be waiting in queue for the unfit neighbour to finish his turn.
- You can also expect this - once it gets to your turn, it will be closer to 2.5Ghz than 4Ghz. That is significant.



My rules of thumb for a purchase:
- small 4c/8t workstation CPU will very likely perform better that a mid-range 6-10c/12-20t CPU.
- 1 socket, high-end workstation CPU will likely perform better that a dual socketed mid-range server and in ~80%+ cases better than some dual socketed 2x 64/128 low-freq behemoth.
- if you need more than 1 high-end WS CPU, you are probably already thinking about horizontal load-balancing (but this brings developers along as app must be set for it).
- if your SLA and UX goals are based on TTFB, freq. is your base input. If you have lots of paralel cores AND want to provide great TTFB/UX levels, go with load-balancing multiple small & fast servers.

These alone are some low-hanging fruits and money saved.
 

Jaguaras

New Member
Aug 27, 2020
12
0
1
In my experience, the return on trying to optimize from a good TTFB to an awesome TTFB is not there, especially if your plan is to do it with a large capital purchase.

There is however huge return on optimizing WordPress themes, the scripts used in the theme and caching and CDN for images. Like if you compare a GeneratePress based theme to one based on one of the slow code builders or a poorly coded jquery heavy theme, it is night and day on the time the page is ready for the user to interact. Same for caching vs no caching.

Don't get me wrong, I am interested in a fast TTFB but that supposed white paper above on processors showed nothing. It just took someone else's scores and multiplied that by a processor frequency ratio without any actual testing of TTFB. If you want a globally fast TTFB on a budget, caching with a Cloudflare KV worker is pretty awesome. Also, Cloudflare's free DNS is very fast compared to other offerings. I am trying to find but haven't, a prior paper I saw that did actual testing on ttfb with different TLS configs and OS and webserver stacks and that is also very likely a bigger difference than a mid vs high freq processor but I cannot seem to find it right now.

So I would say do more research before you buy an expensive server thinking that will make your site fast especially for one that isn' t that high traffic. Software stack and WordPress theme optimization will make a much bigger difference in a fast user experience.
Thanks @amalurk I appreciate your own experience. Yes cache is very good, BUT at this moment when new client visits our website he got load from with huge delay of TTFB, when second time he is browsing with way less TTFB because of caching. Yes we tried with cloudflare free dns but it's just for firewall purposes more. We don't have multi country visitors, we have locals. First we tried to take a look at coding and after that we thought about to change the hardware part. Thats why we are looking to own a server. You mentioned about 4108 vs 9900k about real world TTFB, but in my opinion 9900k is not a server related CPU. In which parameters you think need too look for better performance from CPU?
 

Jaguaras

New Member
Aug 27, 2020
12
0
1
So if you ask should I take Xeon Silver 4108 or 9900K the answer for PHP is to take i9. Even more - never take a Silver or let alone Bronze Xeon for PHP. There are many better and cheaper options for this task.
I was asking about these types of CPU's: Gold 5222, Gold 5217, Silver 4215R (3.2Ghz / core ). I added this Silver CPU because it has good speed / core and only has 8 cores. If I understand correctly it is way better to have with less core but with good speed / core. But why is this Silver 4215R CPU is worse than Gold 5222 or even Gold 5217?
 

vladimir.mijatovic

New Member
Jan 12, 2019
27
15
3
In my experience, the return on trying to optimize from a good TTFB to an awesome TTFB is not there, especially if your plan is to do it with a large capital purchase.

There is however huge return on optimizing WordPress themes, the scripts used in the theme and caching and CDN for images. Like if you compare a GeneratePress based theme to one based on one of the slow code builders or a poorly coded jquery heavy theme, it is night and day on the time the page is ready for the user to interact. Same for caching vs no caching.

Don't get me wrong, I am interested in a fast TTFB but that supposed white paper above on processors showed nothing. It just took someone else's scores and multiplied that by a processor frequency ratio without any actual testing of TTFB. If you want a globally fast TTFB on a budget, caching with a Cloudflare KV worker is pretty awesome. Also, Cloudflare's free DNS is very fast compared to other offerings. I am trying to find but haven't, a prior paper I saw that did actual testing on ttfb with different TLS configs and OS and webserver stacks and that is also very likely a bigger difference than a mid vs high freq processor but I cannot seem to find it right now.

So I would say do more research before you buy an expensive server thinking that will make your site fast especially for one that isn' t that high traffic. Software stack and WordPress theme optimization will make a much bigger difference in a fast user experience.

I agree with this @amalurk
"There is however huge return on optimizing WordPress themes, the scripts used in the theme and caching and CDN for images." No server is powerful enough to be fast for "bloated sites".

But not so sure about this
"If you want a globally fast TTFB on a budget, caching with a Cloudflare KV worker is pretty awesome."
I checked this Deploy a static WordPress site · Cloudflare Workers docs as we are talking about WP and am myself curious for all things that can make WP/PHP run faster.
But this imho does not have much to do with WooCommerce/webshop

Cloudflare stated themselves Deploy a static WordPress site · Cloudflare Workers docs
"There are some features available in WordPress sites that will not be supported in a static site environment:
- WordPress Forms
- WordPress Comments
- Any links to /wp-admin or similar “internal” WordPress routes"
and
"...With a static version of your site being served,..."

That is one way to solve TTFB - serve FPC to users. But tradeoffs serving cached webshop content are much bigger than doing this with a blog.

So I am not sure if I am missing something as this will not solve a "real time" need for processing
- cart,
- keeping the stock status updated
- providing personalized experience for registered users
- invoicing...
in a webshop.



If FPC can be heavily used:

- serving fully cached pages will likely not be a bottleneck even on a 4/8 speedy CPU and will provide solid TTFB for the other side of the world and < 50ms for local users.
- Put media files on a CDN and bottlenecks for FPC (server side cache) will go in order:
Internet Connection > Network card > Storage (if served from files) > CPU/RAM.
If your FPC comes from a plugin and touches PHP, it will be a bit slower, but still very likely fast enough.
- i9-9900K, E-2278G & co. are the fastest ones for Redis, too. As Redis often goes along WP/Magento/PHP... it is a welcomed free "good to know". Redis is single-threaded and does not care if you have lots of cores. It will work better with a faster one.