Purpose:
Consumer grade computers proved to be too unreliable for mission critical server operation. The objective is to upscale the computer infrastructure to enterprise class hardware, allowing us to implement reliable RAID 5 and system failover. This is for a small user base that requires high quality service. Even simple SATA RAID level 2 systems proved to be unreliable delivered with motherboard provided controllers. These are proven reliable controls that should work without issue for many years more than this equipment should remain viable. Hopefully SSD technology will render this platform obsolete and cost ineffective within a few years.
Approach:
Purchase late model Dell PowerEdge 2950 III servers that are coming off lease. We integrated 2 ESXi servers on Dell 2950s one has 16GB and the second 32GB. Upgrade the hard drives to 1T SATA enterprise class drives.
The SAS hard drives were replaced with 1T Enterprise Western Digital drives (WD1002FBYS). Both computers had the batteries replaced, after some hardship with RAID errors.
Lessons Learned:
1. Buy only servers that have PERC 6/i controllers maybe the PERC 5 is ok who wants to find out.
2. Replace the battery, the RAID array will fault if the battery is dead. Intermittent problems will occur if the battery is marginal.
3. Use Enterprise drives that are less than 2 Terabytes.
4. ESXi needs to be licensed to have backup strategy
5. Veeam is an excellent backup/migration product
Outcome:
This project was successful. I would keep the same approach if I were starting again today in 2013. The RAID arrays will write at ~80MBS under ideal conditions. The servers are running AD, VOIP, Exchange, Ubuntu PHP Web applications as well as IIS ASP applications. The server room is much cool and more efficient now that the old Pentium class servers are gone. The backup and migration strategy has realty simplified. It was prudent to invest in software over hardware. Future migrations to new server will be effortless with the virtualized servers. One of the advantages to this approach is that you can restore server between the ESXi boxes to test the restoration. Also, the restoration for all OS disruptions is the same. The quality of the 2008 era enterprise equipment is impressive. The price is extremely low. Server centers liquidate 10, 100s if not 1000s of theses server at one time overloading the market. We purchased these for well under ~$500.00.
