On January 13, 2009  we had another great New England Cloud User Group meeting! We had two great speakers - both providing a view into what a startup is doing to change the way we do email archiving (www.soniannetworks.com) and what Sun is thinking about – how to integrate what they build and sell today with the cloud

                Sun’s speaker was Angelo Rajadurai who is one of their evangelists for all things cloud working with startups. According to Angelo they are trying to reproduce their success in the dot.com era by being successful with cloud companies. Angelo spent the first part of his discussion on how great the cloud was for the future in terms of the business model - removing the capital investment requirements, reducing a startups focus on infrastructure and allowing startups to invest in their Intellectual Property and business readiness instead.

                Part II of the discussion was truly brilliant in that Angelo started talking about AWS specifically and got the heads nodding in the room. He talked about how great S3, EC2, and EBS were in terms of providing a great solution to the startups - easy to use, low cost of entry, and very rich in terms of services. Now he had the room hooked.

                Part III - Angelo started talking about ZFS … and at first I was confused - what does ZFS have to do with the cloud? ZFS is a file system that has some very interesting features in it. For one - it can automatically tier storage based on the storage type. Angelo explained that ZFS is designed with the equivalent of an MMU in it that takes blocks and places them in the appropriate memory location - only instead of an MMU it is a DMU (disk management unit) that can put high read data on an SSD and data that is more general on slower storage such as Tier1 or Tier2 and manages file system storage pools. Angelo’s tie in to the cloud was that he created a ZFS to S3 bucket file system - or a FUSE (file system in user space). Now you may be scratching your head - but consider this … what if you could marry localized file system and storage tiering to the cloud as the slowest (cheapest) tier?

                Part IV - Angelo then told us that what they have found is that the $/TB cost are one dimension but the IOPs cost for the cloud are also important. IOPs in the cloud need to be looked at differently because if the IOP results in the data being transferred into and out of the cloud the cost model changes dramatically based on the direction the data is headed. Sending data to the cloud is pretty cheap whereas getting data out of the cloud is very expensive. What SUN has been proposing is a TCO model that positions a SUN Unified Storage Array at a local datacenter (non-cloud) that is used to present cached data to the customers for high volume read and the cloud is used for the archive. For example Angelo referred to a popular photo storage/sharing site that used their own storage, then realized the cloud was cheaper, then found out that the downloads were killing them in AWS costs - so they now used a local cache by storing heavily read photo’s on a local array and use the cloud to persist the archive.

                Personally I think this is a great strategy to use existing storage technology and marry it to the cloud - allowing ILM to now include the cloud as one of the data stores.  Any thoughts or reactions from all you cloud folks?

Here is the presentation sun-cloudstorage and this is the sun-cloud-storage PodCast.

-wayne