Some of my notes from this session may be useful for others.
Information vs Data – Information is good. More information means more business. The bad is the cost to store data. Data is information captured. The goal is to always store the information in the most efficient way possible. – I agree with this on some levels but there are other goals that go with this.
The hardware data capacity is not keeping up. HDD density is only going to be 36x in 2020. While Data is going to be 124x. Compute is going to be 103x. So we need more compute resources and way more hdd’s to keep up with data growth over the next 9 years. Oh and the budget is only going to be 2x.
The 3 dimensions of storage efficiency is simplicity, efficiency and performance. I’ve experienced the issue of looking only at $ per GB and GB per sqft etc etc. Performance then came around to being the most important. $ per io, io per sqft, etc etc. While maintaining a good position of the 3 dimensions.
The first stage in the journey to the cloud is Simplicity. Number of VM’s growing quickly. Just now starting vm deployments.
The second stage is large volume of tier 3 vm’s. Utility type stuff over in the efficiency area.
The 3rd stage is performance when we put the tier 1 in. Cost is the last thing we should worry about because best performance is what is needed. This is where we were in 2010.
The next generation unified storage is VNX and VNXe. The big boy is the 7500. The smallest is the VNXe3100. – My initial information on VNX makes it sound like it solves all of my NS-960 woes.
The where’s waldo slide fits us perfectly. We play the 3 card monte with storage. Where’s the bits where’s the bits. The bits are here the bits are there. Where’s the bits where’s the bits. Constantly moving stuff around to make the pieces fall into place. Capacity optimization!
By 2012, 80% of all storage capacity sold will be for file-based data. If I keep assigning storage to Celerra like I have been then I agree. Most of our data these days is VMWare Datastores and Celerra NFS/CIFS mounts.
Backup storage is roughly 5x-10x production data. I agree with this. We showed that with Recoverpoint and DataDomain and Tivoli we have many xGB for 1 GB.
VM Storage is only 1/1200 of the enterprise data? I disagree with this. Of course my company is around 99% virtualized. We have about 50% of our data in Datastores and about 50% in NFS/CIFS on Celerra.
AS all of this data growth occurs IT starts to ask the business “do you need to store that” We should ask “Should we store that like this?”.
To reduce production data, compression at file and block level provides 40-50% typical savings. File level dedup typically around 10%. Block level between 20-28%. VNX uses the first 2 types.
CPU to HDD peformance gap – CPU improves 100 times every decade – disk speed hasn’t. The solution is FLASH.