Log in

No account? Create an account
entries friends calendar profile Previous Previous Next Next
Software Dev. - Ed's journal
Software Dev.
Anyone feel like having a go at a software project? There may be some cash in it, for a good solution. (I can't guarantee that I'm afraid, but other solutions we've looked at have either been crap, or ludicrously expensive)

The problem:
My department charges various of our customers for storage usage. This allows us to basically keep supplying storage on a price/GB model, and be able to afford to buy more when disks fill up.

We need to be able to break down the storage used on our servers, to allocate charges to the appropriate subgroup.

The complicated part is that some of these servers are windows, some are Unix (IRIX, Linux, Solaris, HP-UX) and a few are 'NAS' boxes (network attached storage).

We need a utility that will collect the information on storage usage.
Ideally it will be easy to deploy (an annoying installation on 500 servers makes me cry) or even agentless (Connecting CIFS shares remotely for example).
it'll collect data all through the 'disk tree' in a lightweight fashion
And then it'll allow me to report on who owns what, where.
The servers it collects from will be on a mix of standalone systems, NT domain members, and on a variety of networks, so should be able to handle different authentication credentials (saved, so they don't have to be re-entered each time).

For the complicate part, it needs a fairly easy to use system that:
Displays anything that _isn't_ 'allocated'.
Allows 'hierarchical' allocation. E.g. I can set /customer-services/hrd to be owned by a 'customer' of Customer Services, Department tag of HRD. And then have /customer-services/hrd/temp owned by ITC.

The reports should be 'historical' so we can compare changes in charging month on month. Ideally also include 'disk usage monitoring'.

Mostly it will be single server, single customer situations, but the exception is things like our 'user storage' spaces, e.g. file servers.

Oh, and just to top things off, it needs to be relatively lightweight, and not totally cane our network if we try and analyse 40Tb across 500 servers. (Although it's acceptable for it to take a long time to run, as long as it completes in under a week, that's fine).

It should be managable over a web interface (Java if you really must). Server it runs on is fairly free choice, be it windows, linux or solaris (This has an impact on authentication to NFS/CIFS) at a worst case we can deploy one of each. The interface should be manageable at the helpdesk level of expertise, for day to day usage. (By day to day usage, I mean generating the report each month, and correcting 'allocations' of directories for the next report cycle)
4 comments or Leave a comment
cthulahoops From: cthulahoops Date: June 1st, 2006 12:12 pm (UTC) (Link)
Sounds like an interesting project, and one that really shouldn't be too difficult once the NFS/CIFS pieces are solved. The efficiency part seems like the toughest requirement.

If there was cash in it, I might be able to have a look at it - but not until after my viva in three weeks time. How quickly does this need to be solved?
sobrique From: sobrique Date: June 1st, 2006 12:54 pm (UTC) (Link)
Each month I'm gathering info with a a perl script on unix, and a vb script on windows.
I was vaguely thinking that a 'two box' approach, mapping the shares on a Windows 2000 server, and nfs mounting unix stuff on a unix box would be the way to go, and essentially doing a recursive 'df'. Parallelise a bit for speed, but split across multiple boxes to avoid killing stuff.

The config file for 'storage allocation' isn't actually all that convoluted, but the amount of work in 'dealing with' the allocation, corrections, data migration is enough that it's starting to annoy me (and of course, it's only me that can deal with it.

None of the problems are particularly complicated, it's just when you put it all together you end up with a morass of evilness that becomes hard to put a coherent UI on top of.
From: (Anonymous) Date: June 1st, 2006 12:36 pm (UTC) (Link)

Software engineering

I may be up for this.... My name is Steve Bradshaw, and I'm based in Cambridge, and can be contacted at 5@pcfive.co.uk. I've just started freelance programming and IT technician work, so I have some time for such a project

Where are you based and what would you require from me ?

Let me know


sobrique From: sobrique Date: June 1st, 2006 12:58 pm (UTC) (Link)

Re: Software engineering

I've not really got as far as a 'hard spec' yet I'm afraid, it's just I know I have this problem, that needs solving. How it's handled depends quite a lot on how much time/effort/cost is involved on the deliverables.
4 comments or Leave a comment