[Info-vax] Don't want that? Well, don't use that. You've got options. (Was: Re: WHY IS VSI REQUIRING A HYPERVISOR FOR X86 OPENVMS?)

[Stephen Hoffman]

On 2021-01-18 15:34:03 +0000, Scott Dorsey said:

> In general, the IT people are familiar and comfortable with 
> virtualization. So I don't think you'll have to sell the idea of 
> virtualization at all.
> ...
> To some extent you can use divide-and-conquer strategies, putting your 
> web server on a completely different machine than your backend database 
> and having an account on the server with only limited database access.

Here is the general hardware configuration that the folks at Parler are 
purportedly looking for:

40 instances akin to AWS i3.metal, each 64 vCPUs, half-terabyte RAM, 14 
TB NVMe storage, or higher-spec. Scylla Cluster. 
https://www.scylladb.com
70 to 100 instances, each 96 vCPUs, 768 GB RAM, 4 TB NVMe storage, or 
higher spec. PostgreSQL Cluster. https://www.postgresql.org
300 to 400 lower-end instances.
Internal network traffic 300 to 400 Gbps, external traffic 100 to 120 Gbps.

For server scale: AWS i3.metal is an Amazon dual-socket x86-64 server 
using Intel Xeon E5-2686 v4 processors, in a 72T/36C/2S configuration.

I'd expect VSI would offer a discount for this configuration, but 
that's still a prodigious purchase for OpenVMS and clustering (SaaS) 
licenses. Cash for Oracle Rdb licenses too, if that's to be the porting 
target for the PostgreSQL database. And cash for the necessary 
Integrity hardware as production OpenVMS x86-64 isn't here yet. And the 
costs of porting over ScyllaDB and PostgreSQL and the rest of the app 
code, and maintaining it. Migrating and/or rewriting and/or refactoring 
the existing environment—whether for porting from AWS and (presumably) 
Linux to OpenVMS, or for improving hardware efficiency—will delay 
getting the Parler site back online, too.

Then there's the added costs for whatever disaster and availability 
work that might be desired or needed—whether the above hardware specs 
are a single site, or are already a disaster-tolerant configuration?  
And any DDoS, CDN, 2FA, or other costs incurred, too.

Pragmatically, these configurations are going to start with the app 
requirements and app options and alternatives, then the software and 
performance requirements for those apps, and only then whatever 
operating system software and other hardware and software dependencies 
that might exist. And porting from AWS running (presumably) Linux to 
OpenVMS is going to be an involved project, which'll further add to the 
costs.

This Parler re-hosting as AWS and other vendors have seemingly had 
concerns about hosting the site: 
https://arstechnica.com/tech-policy/2021/01/filing-amazon-warned-parler-for-months-about-more-than-100-violent-threats/ 




-- 
Pure Personal Opinion | HoffmanLabs LLC