Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
This is pure curiosity: I am aware that the "post title" functionality has been removed. Yet, in some instances, it seems some posts are still copying the title from the thread's title.
Example, this was a post from yesterday, where you can see the subtle line that divides the post title from the post:
I thought today I would take a moment to explain my thinking in dealing with the following problems that I mentioned earlier. Following is a quote from what I wrote a few days ago about the origin of how I developed MwenBank.
I want to write about …
I could not find a way to forcibly re-instate the titles. Is it a glitch or am I missing how to do it?
As I mentioned many pages back as something I wanted to do... I left Google Cloud Compute and reverted back to something far simpler. Instead of a triple node MariaDB Galera replication system, with multiple physical nodes for the application servers, I reverted back to something that served us well for years. A single node setup.
Just like trading, you have to always be learning. The reason I went with GCP in the first place was because of rolling hardware upgrades and reliability, both having been the weak point of our old setup (aging hardware).
However, that aging hardware was mostly because we owned it and we co-lo'd into a datacenter. If I didn't own the hardware, I would have been more willing to ditch it earlier and upgrade in between Intel release cycles.
But GCP added complexity that proved to not solve problems for us, but create them, and it was also noticeably slower due to optimization for mutli-core deployments but at slower clock rates.
So I searched and found a high-clock rate replacement better suited for our workload. Less cores, but higher clock. I went with an i9-10900X for its stability, in a server class motherboard. Sure, we give up ECC ram but my experience tells me it's not important in real-world, where higher clock certainly is.
I also upgraded to an 8TB NVMe, specifically the HGST SN200. It can do 1.2M IOPS and 6GB/sec transfer, a significant improvement over anything we've had before.
When you look at our biggest threads, like the SP500 Spoonalysis thread, it was taking 2.5-3.0 seconds to load per page on the GCP hardware. Now we are down to less than 1 second, a massive improvement.
You should also see similar performance gains when creating a new post, as that was one of the slower aspects of the old system as well. The current bottleneck in that regard is related to the number of Followers in the thread, as notifications are processed in real-time when you submit the post (architecture of vb 3.8).
Since we gave up triple node redundancy with Galera, I am instead doing real-time hot backups of the databases using mydumper, combined with real-time file system sync using lsyncd. We have a much smaller backup node on a private VLAN at the same datacenter, to hold our backups and with the ability to step in at any time to replace the master in an emergency.
Anyway, having moved to Dockerized containers a while back, there is no real issue migrating from one host to another any longer. But please do report any problems, nonetheless.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals, U308 and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,059 since Dec 2013
Thanks Given: 4,410
Thanks Received: 10,226
Isn't the entire business class server industry built around the opposite of that? ECC, more cores and threads and lower clock speeds, ie Intel Xeon chips?
Yup. But for us, it's better with the i9 to get the high single thread spread, and that means non-ECC. In all my years, only time I've seen ECC error bits flip and log is also when there is system instability which you would catch or be aware of regardless. ECC isn't really "saving" you in any of my own personal experience at least.
What do you think the most direct Xeon is for the i9-10900X?
Should also mention we are running DDR4 2933 now in this system, a significant improvement over what we were seeing with GCP. It was very noticeable. I'm sure something was also lost in their virtualization, as we were of course using Compute Engine, whereas now we are back to our own dedicated physical box.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals, U308 and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,059 since Dec 2013
Thanks Given: 4,410
Thanks Received: 10,226
Me? No idea. My last several trading desktops have been dual Xeon server class desktops. Not sure if its due to the ECC or not but desktop or application crashes are virtually non-existant. (Although I am now having some issues with Excel and the RTD links with the new Trading Technolgies platform but thats another story).