While visiting DLD Tel Aviv last week I had the pleasure of grabbing a coffee with our good friends Roy and Avi, founders of the Israeli company MoovingON.
Scalability comes at a price
We got into an interesting conversation about startups and how beginning with an elastic cloud platform makes sense for them because it allows easy scalability. So, as the business grows and traffic increases the platform expands elastically to accommodate.
Trouble is, the costs are also elastic and tend to start ballooning. But while the business is booming it’s difficult to change course or even find time to plan ahead. The root of the problem is the billing arrangements for every cloud service asset: not only CPU time but memory and disk capacity usage, cashing traffic usage and upstream/downstream traffic usage being charged by the gigabyte.
How, then, is it possible to scale a cloud solution while controlling unnecessary elastic usage – and so control costs too?
Separating big data from real time processes
The answer, according to MoovingON, lies in the architecture of most SaaS platforms today. Generally the services combine on-line/real time services and back-end/offline services.
- Incoming: Gather data for processing from internet resources
- Filter: Filter data into write backend processing
- Real time processing: Process the information into value output
- Output value: Create product usage/value (output) to the consumer/business
- Big data processing: Collect various data from different sources, process them into valuable data to be used by (1) the real time processing engine (2) for business decisions
Of course, every startup wants to be able to scale fast to handle traffic peaks and grow in line with increasing traffic. For those who establish their SaaS in the cloud it makes sense to keep the real-time processes (items 1 to 4) there because they can easily add servers as they grow. The million dollar question, though, is where to locate the big data processing?
The big data processing is basically the “brain” of the solution and it requires huge amounts of information to be processed into statistics data and requires a lot of assets. But whereas the first four items depend on flexibility and dynamism, big data’s requirements are for more bandwidth, consistent power processing, redundancy and data integrity. It can, of course, all be done in the cloud as well – but at what cost?
Best of both worlds: hybrid cloud infrastructure
Using their experience in SaaS uptime management, MoovingON has developed a keen understanding of hybrid infrastructure. In particular, they’ve realized that back-end and big data applications are far better placed in leased server environments using bare metal architecture that can hold a huge number of computing assets at a much lower cost than cloud services. Not only that, they give better processing power and lower cost bandwidth which is just what big data needs. But what about scalability? The architecture can still grow with the business – although it may take a day or two rather than being scaled instantly – but overall this route gives startups a far more robust system right from Day One.
MoovingON and Leaseweb work closely together with various customers in many industries (MoovingON also run their own infrastructure out of Leaseweb’s datacenters in the US). One of their NOC (Network Operations Center) functionalities is to document any activity that occurs in the customer’s environment 24/7. Then, they use this data to analyze and create reports, allowing them to continuously improve their service along with their monitoring infrastructure and reliability levels.
They see the bare metal environment as key: “As this is a core asset of our service and requires 100% uptime, we had to look for infrastructure that can be fast on the one hand and strong enough for processing the information on the other. We checked out many options but eventually we saw that an SSD Bare Metal environment mixed with Leaseweb Cloud formed the perfect combination of reliability and cost efficient performance.”
MoovingON provides various service for UPTIME and on-going maintenance starting with (1) a design and building monitoring solution (Nagious, Icinga, Munim, graphite, keynote, OpsView etc) based on best practices and tailored to customer need, (2) ‘NOC-as-a-Service’ which enables companies to outsource 24/7 monitoring, support tiers and a B2B helpdesk so allowing them to focus on their core product functionalities, and (3) on-going maintenance of the infrastructure. For more information please visit them: www.moovingon.com
P.S. Would you like to learn how to best scale using hybrid infrastructures? Leaseweb and MoovingOn will organize a panel discussion on this topic in Tel Aviv early December. Stay tuned for more details!