discussion new server upgrade (funding)

[Redyscom]

genuine question: what do you need all this performance (and especially RAM and storage) for?

To ensure that the site runs smoothly and is easy to use for everyone, while avoiding the incident the other day when the large influx of people caused a problem (at least, that is one of the theories being considered as to what caused the crash), but also to guarantee space for future growth as the site is improved.

You are not totally wrong, i go to Fthing for the last update game ( like tamara ch3 v9.01) but i found something here and if you try and invest yourself here i think you can find something too like my lastest thread The peoples here are insane and i've experiment it but it's a good thing and it's something you won't find anywhere else. The community is awesome, i'm wrong but it's my right and you can argue with me. I will have support for my point of vue and you will have support for yours we both win. There is no looser on Lewdcorner.
Community is the best thing and they have the most respect here because it's the most lurker. They take, they enjoy and they don't open a thread for nothing, you can verify it. I was a lurker for a long time i know they are the majority.

I agree. I've only been here for a year, I don't like interacting much in forums, but I'm quite active here because I feel comfortable, I don't have that anxiety that if I comment on something, some idiot might come and mess with me like it's happened to me in other communities. Here, it's very rare to see annoying people doing that, and they tend to be quickly shut down.

Of all the communities I've seen, this is probably one of the healthiest. Yes, the site has problems. Yes, not many games are being uploaded, but that's more due to a lack of uploaders. But even with all that, there are people who keep coming back faithfully, some writing and others just reading from the shadows.

Hell, a few days ago when the server went down for a whole day, it took less than an hour for thousands of users to connect when it came back online.

 

[jkj34]

This is so strange to me as both of these are horrible out of date at least in terms of the cpu. Intel Xeon E5-2650L v4 came out in 2016 and Intel Xeon Gold 6152 came out in 2017 and both are several years beyond their last End of Servicing Updates Date. Both are rocking DDR4 which I suppose in the RAM crisis isn't the worst though DDR 6 is about to hit. I mean both CPUs are sub $300 for old new stock and sub $40 for renewed. Which brings the question of if the host is really charging an arm and leg for this or in what situation and for what price point would you accept this? Is selfhosting really out the relm of possiblity? The only thing of real cost in that setup would be the 2× 3.84TB NVMe (Toshiba) which are going for around $1000 at this point, but other than that you could do a lot better for cheaper.

The more newer you get the price skyrockets a thread-ripper cpu by itself can be up to 17grand usd so server rental cost goes way up.

 

[Thcar]

I am making this thread to ask the community to help fund a new LewdCorner server upgrade. ----> Donate Link (goal is automatic meaning it is transparent)

The goal is simple: pay about $30/month less than what we pay now, while moving to hardware that’s at least 2–3× stronger overall in real-world usage. That means more stability, faster page loads, better performance during peak traffic, and way more headroom so we’re not always riding the edge when something spikes.

Below is a clean comparison of what we have now vs what we’re wanting to upgrading to:




edit: goal reached because of @Cyrix
new server purchased, (waiting for apollo to wake up before the merge, expect downtime.)

You must be registered to see attachments

Option A (Current) vs Option B (Upgrade)

Category	Option A (Current)	Option B (Upgrade)	What it means (plain English)
Server / Platform	Current platform	Newer platform	Option B is a newer, stronger overall server build.
CPU	Intel Xeon E5-2650L v4 (14c/28t) 1.7–2.5GHz	Intel Xeon Gold 6152 (22c/44t) 2.1–3.7GHz	More cores + higher clocks = better performance under load and more users online at once without lag.
CPU strength	Baseline	~1.6× to 2.2× more compute (typical server workloads)	More headroom before slowdowns when traffic spikes or background tasks run.
RAM	64GB (4×16GB DDR4-2400 ECC)	128GB (4×32GB DDR4-2666 ECC)	Exactly 2× the RAM = more caching, fewer “memory pressure” slowdowns, and smoother overall performance.
Storage	1× 3.84TB SATA SSD (Intel/S4510)	2× 3.84TB NVMe (Toshiba) (7.68TB)	NVMe is massively faster for databases + lots of files, and 2 drives also ees redundancy options.
Storage “feel”	Fast SSD	Server-grade fast (NVMe)	For DB + file-heavy workloads, NVMe can feel 3×–10×+ snappier than SATA (depending on what’s happening).
Drive failure risk	Higher (single drive)	Lower (dual drive → RAID1 option)	Option B can be configured to survive a drive failure without taking LC down. (like it just did)

Quick why we want Option B

• More CPU power: 14c/28t → 22c/44t plus higher clocks (more users, less lag).
• 2× RAM: 64GB → 128GB (bigger cache, smoother performance, fewer slowdowns).
• Dual NVMe drives: way faster DB/files + redundancy option (RAID1).
• Guaranteed 300M unmetered: consistent throughput for the community at peak times.
• Cheaper monthly: about ~$30 less than the current server while being significantly stronger.

I watched a recent Linus tech tips video with Linus and Linus Torvaldsen building a Linux computer together and Torvaldsen (who created Linux) said something about ECC compatible RAM and that he never goes in for anything other than ECC compatible because otherwise you risk losing your RAM to failures much earlier than not. I don't know anything about computer hardware nowadays but this seems like something you might want to look into if running a web site server is RAM heavy at times.

 

[VSCDB]

I watched a recent Linus tech tips video with Linus and Linus Torvaldsen building a Linux computer together and Torvaldsen (who created Linux) said something about ECC compatible RAM and that he never goes in for anything other than ECC compatible because otherwise you risk losing your RAM to failures much earlier than not. I don't know anything about computer hardware nowadays but this seems like something you might want to look into if running a web site server is RAM heavy at times.

Just a bit of my knowledge as someone working in the HPC scene (huge servers and clusters basically) most servers have been using ECC, which basically is just error correction (ECC = Error Correction Code) where basically you can verify, that even if a bit is flipped, you still get the originally written data from the memory location. So this will not solve this problem. There are possibilities of having less RAM usage by as example running less actual software by deploying an Alma, RHEL or similar server image, and being really careful how the whole infrastructure is set up. But even then an influx in users could again lead to problems without actually having more RAM or scaling the set up with CDN (which might already be used) and a cluster with multiple compute nodes and shared resources (which would have more actual RAM as well).

 

[Slay]

Set the kernel to use "Huge Pages" for a potential 5-10% boost in search speed for free with that new cpu.

 

[oberold]

I see the website getting upgraded. Nice

 

[BLAHBLAH676767]

xeon tends to be a good option when were talking about servers.

You must be registered to see attachments