Storage company Zadara offers a product for IT managers who know they have to embrace the cloud but want to keep a firm hold on their priceless data assets.
It offers what it calls 'on-premise as a service' storage, with no up-front capital costs, available on SSD, SAS and SATA drives. The service provides consumption-based charging for storage and service, maintenance, and support are included in the per GB price: additional capacity can be supplied and installed as the customer needs it. The company's COO Noam Shendar talked to ZDNet about the thinking behind the service.
ZDNet: The cloud is everywhere and flexible, but keeping your data at home offers security. Can you get the best of both for your storage?
Shendar: The founding idea for the company was that there is a better way of delivering IT to the customer than we have today.
When you look at doing IT in a cloud way you should not think of it in the limited sense of the public cloud. The public cloud is wonderful and will account for at least half of IT as the years go by.
But the other half will still be on-premised and still won't have the opex, pay-as-you-go, cloud-linked model for the financial arrangements with the customers.
This is the part we latched onto. This is the part that illustrates the entire way that IT is changing. We have the opportunity to do all of this.
Now we are a storage company and that's our background. So we asked, are any of our storage products capable of serving this new nation? The answer was, absolutely not.
So we saw a cloud opportunity in storage but we knew that the only way that a cloud storage service would work was to be built it from the ground up.
To do that we began by asking our customers questions like 'is your EMC server a good product?' and 'is your NetApp array a good product?' and the customers said yes.
So our next question was: do you like your experience obtaining the product, living with it, and refreshing it? The answer was an almost universal no.
They said: 'I hate it that I have to pay a lot of money up front for something that I may have to live with for a long time. I hate that I have to plan multiple years in advance which is an impossible task. I cannot know what two years into the future holds, let alone three or four years.
'I hate these disruptive refresh cycles which mean I have to have long evaluation cycles. And, most of all, I hate the fact that after three to five years, I have to start this task all over again.'
You commit to one system and that means you are limited by that system. If someone invents something better, you can't take advantage of it because you are stuck with your system.
We thought that cloud opened up new opportunities. Now with cloud you can get storage easily and affordably, but it is not very capable.
So we thought, what if you could offer fully featured storage a cloud service? What if you could offer the same performance, reliability and compatibility?
In other words, the features of an EMC's but with the elasticity and offered by a fully-featured storage service that you pay for by the hour. We call that Enterprise Storage as a Service.
Our system is not a toy, it is not basic, it is not something that you have to relegate to a different class of applications. It is the main event.
You can offer it immediately and on-demand. And we can offer you what you need right now, not what you need next year. We ship that to you, at no cost, and we remotely operate that for you.
So this is introducing software-as-a-service (SaaS) to your portfolio?
No, what we are talking about is the entire experience. What the customer gets from us is the entire appliance.
We talk to the customers to find out what sort of drives we should be offering and we get visibility into their next expansion because we always ship more capacity than the customer needs. We don't charge the customer for that but we want the customer to be able to expand at the push of a button.
After shipping the drives to the customer, the customer opens up the boxes and the most common form factor is a supermicro server with 36 drive slots in four rack units. They are 24 drives facing forward towards the main isle -- or cold isle -- and then a dozen more drives facing backwards towards the hot isle.
They get a supermicro-based chassis and they are standard servers. We don't customise them in any sense. They have a standard motherboard, a standard CPU, standard memory, standard 10 gigabit or 40 gigabit NIC, and one gigabit management ports, and so forth.
Now let's say that there are five boxes within these servers. Well then, there is a sixth box that has the networking. What the networking kit has is a bunch of colour-coded cable and two, high-speed, 40 gigabit switches and then one, one gigabit switch with a bunch of ports and this is the management and our redundant, heartbeat network [a private network which is shared on by a cluster and is not accessible from outside].
We carry the heartbeat over the redundant 40 gigabit switches and then there is a third heartbeat network which is also over the one gigabit switch and all of the management traffic goes over this.
And last there is a firewall.
Why a firewall?
Because if we need to talk to this system remotely we don't want to be -- and in most cases won't be allowed to be -- inside the customers' firewall.
In this case, the customer connects the firewall to their external internet connection, either directly or through the cluster. We send a simple instruction guide using a colour-coding system -- these cables go here, these cables go there and so on. Then the customer connects the internet to the public internet and then lets us know when everything is on.
We connect to it remotely from one of our two locations: one is in Irvine CA and the other is in northern Israel [in Haifa] which gives us 24/7 coverage.
At that point we bring the system up and check that the system is responding as expected and then we tell the customer that everything is ready.
The customer has a provisioning portal -- a web-based portal that is internal to their network and accessible through a private IP address. That provisioning portal is exactly the same as the provisioning portal in the public cloud and at the same prices.
The point is that the customer has the same cloud experience even on a system that is completely private to them.
Also we take exactly the same system -- the same hardware, software, provisioning portal etc -- and we drop it into public cloud sites around the world. We use Amazon Web Services (AWS) sites around the world -- 20 to 30 of them -- for this, so that the customer has access to all of those as well. So, the customer can use AWS around the world at the same price as using the system on-premises.
The point is that the cost is the same. The storage cost that is, the AWS cost, is down to AWS and the customer.
The principal aim here is to make it non-punitive for customers to change their minds about anything at all.
So this is the cost of the whole system -- hardware and software?
Yes. The hardware, the software, the maintenance and support, the monitoring, the SLAs, and so on.
But if this is all standard, what is your unique point of difference? What makes you different from any other storage vendor?
If you look at NetApp as an example, you can't teach any NetApp offering to scale up or down non-disruptively and on-demand. Secondly, we have a unique, patent-protected, multi-tenancy that does not exist in any other vendor's system.
Now, when we say we are shipping five servers to the customer you are probably picturing a five-shelf array. Those five servers are actually five piles of resources and by resources they are actually combined into one that is, in itself, elastic. By resources, I mean drives, CPU cores, and memory.
So, say one of our customers needs a three-drive array, you can do that. When you create the array you can choose the drive mix and quantity that is right for that array.
You can create as many VPSAs as you need and each behaves as a standalone array in every respect. They cannot access each other. They behave like separate boxes that you put in separate corners of the room.
So if you set up the system and run with the equipment you need and then find that you need to get more, quickly, you and do that and it all happens transparently?
Yes. The philosophy of the product is that you cannot make a mistake. The founders are all veterans of the storage industry and we have, collectively, been at every storage company you can name.
And there were things that drove them nuts about the industry and they finally had an opportunity to address them.
One thing was that scalability only works in one direction -- the vendors made it very easy to grow and impossible to shrink. The rationale is obvious but it doesn't serve the customer.
The second was that you could make mistakes. There were decisions that you could make that would paint you into a corner and were irreversible. It could be blade strategy, or how you partition, and if you made the wrong choice you couldn't undo it. You would have to move the system somewhere else, re-configure it and then move it back again.
Thirdly, there was the requirement to spend many months planning. People were sitting in rooms doing long-term planning and know that they would get it wrong or IT would get it wrong.
So everything we did, we did thinking of what the customer would wish for if they could wave a magic wand.