Day 2 of EMC World 2013 started off with a general session from the CEO of EMC, Joe Tucci. He talked about the software defined datacenter, the shift to even more users which are also mobile, and the ingress of large amounts of data which is influencing storage designs. 
Later that day we also attended a keynote from the BRS (Backup & Recovery Systems) division. To be honest I found todays general sessions lacking compared to 2012. Last year these sessions had a clear strategy on where EMC and the industry wanted to go. Plus they introduced a boatload of new products, new features, etc. The general keynote in 2012 allowed me to zoom out of technology to a high-level view and then adjusting the session schedule around the general keynote subjects.
This year the session was not that overwhelming. Not as many new products were announced, and a lot of time was spent talking about Pivotal. Overall a flashy presentation but it didn’t really inspire me like last years.
Luckily I planned myself a breakout session regarding Isilon protection on an Enterprise scale. It discussed why you should choose a scale-out NAS system instead of traditional NAS in the first place. It also touched a large number of mechanisms that are in place to make an Isilon system as fault tolerant as possible. For example, to name a few:
- File system journal which is protected by a dual-battery NVRAM RAID controller.
- Proactive copies to hotspares for all drives; remove a drive from the setup before it actually fails.
- All (meta)data is protected with 32bit CRC checksums.
- MediaScan which scans the installed drives for errors.
- Integrity scan which does the same as above but on a filesystem level.
- Etc…
The session reaffirmed that you cannot define a replication policy without a proper business continuity plan: how can you design adequate and cost effective storage replication if you don’t know what the business desires?
First of all, traditional backup is no longer going to work. To illustrate: a PetaByte (PB) of data will take a little over 80 days to push to LTO-5 tape. Now imagine you have a fullscale Isilon at 20PB…. good luck!
Isilon uses snapshots to protect data. This snap is then replicated to the other side onto another Isilon. This enables both protection against site failures AND against virus attacks: if your primary data is corrupted, you can fall back to an older copy on the secondary system.
All in all a very exciting product; I should be playing with an Isilon in two weeks during the course!
Questions, comments, etc? Shoot!
