dmj

Take a look at Vagrant, it's really useful for creating reproducible VirtualBox machines: https://developer.hashicorp.com/vagrant/intro

There are other free options if you don't want to go the full boatware hyper-v method: we used to use KVM as a hypervisor, it meets all the requirements for being lightweight and stable (it's the same hypervisor GCP and AWS use) and there are some good gui tools available (virt-manager, cockpit, proxmox). There are supported options if you are desperate to pay someone (RedHat Enterprise Virualisation, Proxmox etc)

This is quite good: https://www.hibob.com/

I came across this open source tool today, not had much time for a deep dive into it, but it might fit you needs: https://www.danswer.ai/

It's a strange one because, either a) the school is the third party and under the contract isn't even allowed access to SIMS at all b) the school is the customer and is allowed to have access to their database file. Sounds like ESS have really stuffed LA's with this contract as well.

So the LA have a contract with ESS and the school is the third party. So under the ESS contract, the LA cannot allow the school to access SIMS? Given the contract is between ESS and the LA, i'd just follow the advice given earlier and reset the SA password on the database and create a backup for arbour

it's 2024. This is a job for a container.

Fixed that for you

If you want to do this with AI, in a google environment, start reading up on Vertex (not an AI model itself, but a collection of tooling - you can use whichever model you like, including opensource ones). You'll probably need the Document AI The idea is then you use Retrieval Augmented Generation to create your own model that's been trained on your data. You can interact with that model via an API and create a frontend app. That's the basic idea anyway. To be perfectly honest, I would approach this from a different angle. Rather than document what you have, document what you want and then build it from the documentation. This is really an Infrastructure as Code technique where you plan out what you need and built - this serves as your documentation (with appropriate comments) and your configuration.

Like outsourcing IT support to MSP's?

Can't you just use Office 356 as a service provider and use another identity provider or host your own. I'm sure the Google IDP is still free for schools. Then you can setup the MFA how you see fit. It doesn't seem very long ago we were all hosting our own IDP's and integrating third parties with them, has edugeek just lost that collective skillset now or are microsoft forcing schools to use their auth?

Anyone would think the application and database would scale with load

It's an archive backup. You know the saying if your data isn't in three locations it may as not exist. Snapshot every fifteen minutes, this is what's used for daily recovery tasks, as it's local. Only the archive storage that needs to go offsite does, so that would be weekly diffs and monthlies.

Ingress is free, egress is very expensive, like £50 per TB. If we need to recover then the company would pay without complaint.

er yes. multi regional archive shouldn't be more than £2 per TB per region

Looks like it, reddit is convinced at least. good old fashioned regression test failure by microsoft.

That's three times what we pay. GCP Archive storage is £2.04 per terabyte for single region archive https://cloud.google.com/storage/pricing#regions

also: try on another operating system - if you don't get memory issues on Linux (and you ran memtest on linux) then 99/100 it's a mobo driver issue.

Did you try reseating it? it might help. Perhaps try increasing the voltage on the RAM (ever so slightly!) may improve stability. Difficult one if memtest looks good.

Run memtest and make sure it's not bad: https://www.memtest86.com/download.htm edit: reseating the memory might help, but my money is on bad ram.

They're all generic linux switches AFAICT, that's what keeps the cost down.

At last. an abhorrent piece of software.

It was a long time ago, and I wasn't directly involved, but If I recall correctly Capita complained that the spec wasn't sufficient but also refused to help work on it and withdrew (I personally believed they never wanted interoperability, and that was a big selling point for its current owners ). I'm not sure if Bromcom were ever involved and Arbour weren't around back then. The government cancelled BECTA and the support in the UK kind of dried up as none of the major providers would play and the government wasn't directing them to do so. It's still big in North America, on version 4.3 by the looks of it. Perhaps some of the larger MAT's would pick it up and integrate the spec into their tenders, I think thats the only way it would get traction.

With SQL queries presumably. Merged into a4l. The spec is going strong, but a few notable absences (in terms of members) https://data.a4l.org/sif-specifications/

Utter nonsense. Anyone with basic knowledgedge of how unencrypted databases work can see this. No it doesn't? Sorry I just want to pick up on these two points. point 1) is actually fair - database extract is a pretty terrible way to get data and they are correct that it doesn't guarantee data integrity, a seemingly insignificant database migration could cause a lot of mishap. An API is usually a far better way to do this, but in ESS/SIMS case their API is just insufficient so there's no real alternative. Point b) an SQL extract contains all the information required to build a copy of the schema - it *is* their code, their SQL instructions etc, etc. They have the right to claim copyright over it to prevent others from making derivative work, but nobody is doing that, so they are scaremongering plain and simple.