M$/T-Mobile SideKick cloud failure....

[Dr Mordrid]

Never did think much about this whole 'cloud computing' thing....

Link....

Mobile Computing

Microsoft/Danger: It's a Sidekick disaster, data gone?

By Julio Franco, TechSpot.com

Published: October 11, 2009, 11:13 PM EST

After numerous service interruptions during the past week, on Sunday afternoon T-Mobile informed owners of the Sidekick smartphone that user data, which is stored in Microsoft/Danger's servers could have been lost for good. By design, the T-Mobile Sidekick doesn't store much of its user data on the device itself, but uses the cloud for storing address book information, photos, calendar and application downloads. And while service outages are still commonplace in today's cloud-based services, this could be one of the most high-profile data loss occurrences in recent history.

Reportedly, the data loss was caused by a failed upgrade to Danger's storage area network, only to discover later that no backup was available. Microsoft's involvement only aggravates the issue (they acquired Danger in 2008). Undoubtedly, this will come as a complete blow to Microsoft's image, both as a cloud service provider (see Azure) and to its mobile platform division, having just recently announced the Windows Phone initiative.

An official follow-up statement is expected Monday regarding the status of the potential backups. In the meantime, T-Mobile is alerting its affected customers (number not disclosed), letting them know they shouldn't remove the Sidekick's battery, reset the devices or let them lose power. Syncing with a desktop application should be a safe measure to be taken at this point if your Sidekick has some info on it.

[Jammrock]

Never did think much about this whole 'cloud computing' thing....

Link....

"Clouds" are just a buzz word to describe load balanced or reverse proxy web hosting. That's all a "cloud" is. They still require the normal redundancy checks on the back-end to make them truly failure proof (as much as is possible), especially on the storage controllers which is your most expensive and most important piece on the hardware side.

From the looks of the articles I read the issue boils down to sloppy work by the tech workers involved in the SAN upgrade. The key issue, in my opinion, was probably the fact that most people don't think that SANs need to be backed up because they are redundant by nature (dual controllers, RAID arrays on the disks, hot spare(s) available, etc).

But the large disks that are in SANs these days make them more susceptible to failure than most people think, because rebuilding a large drive from parity can take a full day or more. If a second drive goes out in that time your data is toast. Unless you use up a great deal of your SAN real estate using a RAID 6 (double parity, 2 drive failure) or RAID 50 (mirrored RAID 5), or RAID 60 (mirrored RAID 6), which most people refuse to do because SANs are so damn expensive and the bean counters feel those type of arrays are wasted money.

I think Google has the right idea. They don't use more then a simple RAID 1 on their servers with no back-end SAN. Then they keep copies of their data on multiple systems in multiple data centers to provide redundancy. Though most people don't have the money or software expertise to pull that off.