Next week I will fly to Boston to attend my first full Tech Field Day conference as a delegate.
Last year I was lucky enough to be invited to the smaller scale Tech Field Day Extra event at VMworld Europe and I really enjoyed the experience, so you can imagine my excitement when I received the invite from Stephen and Tom to join them at TFD14.
Not yet time to pack a bag, but definitely time to start doing some research on the three vendors that will present at TFD14: ClearSky Data, Turbonomic and Datrium.
Let’s start this TFD14 Preview Series from ClearSky Data: from what I understand ClearSky has a very unusual approach to Cloud Storage which is normally intended for secondary/object storage kind of use cases. ClearSky has developed an interesting architecture that allows for storing all your data in the cloud while not sacrificing the performance (and use cases) typical of primary storage – all of this going beyond the obvious caching technologies that have been around for some time.
All revolves around the set up of so-called Points Of Presence (POPs) in key US locations: for you to be able to access ClearSky’s service you should be at metro distance from the nearest POP. This could be seen as the main obstacle to the adoption of this solution, but plans are to expand ClearSky’s Global Data Network worldwide, so this concern should quickly go away. The POPs will host ClearSky’s equipment and will be connected on one end to the customer’s premises (using 2x 1 Gbit leased lines that ClearSky will contract and deploy for you as part of the package) and on the other end to AWS (still with very capable network links).
When customers sign for ClearSky’s service all they have to do is to find 2U’s in one of their racks to deploy ClearSky’s All-Flash Edge Appliance and connect it to whatever needs to consume that storage. Writes are sent to the Edge Appliance in “write through” mode and they will immediately hit the main storage at the POP (this is doable because of the extremely low latency that the dedicated links can provide). So, the Edge hosts “Hot Data”, while the POPs host both “Hot” and “Warm” Data. All of this, plus “Cold Data” is eventually sent to Amazon. ClearSky claims that “worst case scenario” for reads of cold data can take at most 50 msecs.
The main advantage of this model is that customers can have the benefits of on-premises primary storage without having to invest in a SAN infrastructure; the very simple PAYG model allows for customer to expand and shrink the amount of “used” storage (yes, used, not allocated!) and pay only for what they need, without having to buy storage in big chunks upfront.
A couple of interesting use cases are for Application Mobility: data stays in the same POP – or it can be migrated across POPs – while it can be accessed by compute nodes in any location (as long as they are close enough to the POP). Imagine turning off DC1 and restarting your Apps in DC2 without having to move your data (as it stays in the POP + Amazon Cloud). All you have to do is to re-warm the Edge Appliance in DC2!
Another use case gaining momentum is the integration with Data Analytics tools: both Splunk and Elastic Stack are supported. These tools are IOPS intensive (that will be taken care of by the Edge Appliance), but they also tend to consume increasing amounts of storage space (and this is when “infinite” elastic storage available in ClearSky’s cloud comes into place). ClearSky seems to be the right match for this type of workloads.
According to ClearSky, their customers use their solution for VMware (it is VAAI compliant) and to run DBs and Exchange, so Tier 1 apps seem not to be a concern.
This is what I discovered (and hopefully correctly understood) about ClearSky and I am really looking forward to hear from them next week about the product roadmap and advancements.
Stay tuned for part two and three of this TFD14 Preview Series where I will discuss about Turbonomic and Datrium.