PROPOSAL TO SPEND MONEY: Server restructing

Our server infrastructure at the hive is a little mismatched.

We currently cannot use all of the servers we have because we do not have enough power and we blow breakers. (The UPS issues are addressed in a different thread)

We currently have:
2x HP DL320S 12xSATA storage servers (idle, cluster)
3x Dell PE2950 (idle, cluster)
1x DL360G5 (VMWare / schizo)
1x DL360g4 (Linux General / hubuntu)
1x PE2850 with 12 bay SCSI disk shelf (Fastor)

I propose:

  1. Buy 12x2TB drives http://www.newegg.com/Product/Product.aspx?Item=N82E16822148834 @$89 ea = $1068 We could buy less expensive drives, but the drives I chose are not THAT expensive, offer excellent storage/$ and speed/$ and they have a large number of solid reviews.
  2. Install new hard drives in DL320S, 12 x 2TB, RAID + Filesystem losses should still yield 10-12 TB of storage
  3. Run FreeNAS on DL320S. Export iSCSI to cluster machines. Export iSCSI to VMWare. Export CIFS/etc. to machines on network for storage.
  4. Retire Fastor. Retire Fastor’s disk shelf. Mothball 2nd DL320s (cold spare)
  5. Convert 1/3 PE2950 to VMWare server, 2/3 PE2950 for Cluster
  6. Virtualize DL360G4 (hubuntu). Retire hubuntu.
  7. Install single hardened (grsecurity patches, etc.) Linux distribution on DL360G5. Install mediawiki, wordpress, etc. Install in colocation facility to replace old server.

The only FINANCIAL expense that need to be approved by membership for this plan is the purchase of drives.

The net result of this would be to provide budget-baller enterprise class storage to enable other computing resources at the hive. It would additionally allow servers to be re-allocated in a manner to better serve the needs of the space by greatly increasing the speed and resources available for our website along with allowing the computing cluster to be used without creating power issues.

I imagine there will be some discussion on this, but again I’d like to vote on this ASAP.

Thanks,
-Dave B
CTO

Dave,

I’m in favor of the disk farm!

As to the other restructuring - I’m not familiar with some of the nuances so I’ll reserve comment on that.

However, if you are saying - centralize the disk farms and move all available DISK/CPU resources into the “cluster” (i.e. Open-stack cloud???) then I’m in!

With the mini-cloud we can “right-size” and “right-function” static server instances to maximum availability of those resources for other uses.

Regards,
James

Two comments for right now:
We already have quite a few TB of disk space and very, very little of it is used so I don’t really think we need to spend a thousand dollars to upgrade that. Second, is there a reason to stick with VMware? Having used both vmware and openstack a little bit now (just spinning up and down some small instances and playing around) I don’t see a reason not to use openstack for everything. Especially since it is already set up on the servers Ian brought down.

Just my thoughts!

Maintaining an enterprise level cloud computing cluster requires manpower and resources that I don’t think we have at the moment. Not saying that I wouldn’t support adding more hardware and setting up a cloud computing cluster at the Hive, but I think that we should explore other solutions besides Openstack. Ian was looking into Eucalyptus, which from what I’ve gathered is less complicated and would better suite our needs.

As far as spending that much on storage, I don’t think we use that much storage at the space as it is. I could see spending $500-$600 on disks, but $1000? If we are going to spend that kind of cash then I think it’s time we take a serious look at the current and projected needs of the space first, then go from there.