About ENKI's technology

Are ENKI's cloud instances (VMs) persistent or emphemeral?

ENKI uses a fundamentally different design for its cloud than other cloud providers, based more closely on enterprise compute clusters.   Instead of giving you a slice of a colocated server with its own disk in it that can fail and lose all of its state, ENKI keeps all storage on a central SAN system connected with high-speed Infiniband interconnect.   Should a server fail, VMWare will restart the affected virtual machines on another server, and all the storage-dependent context will be reloaded from the SAN.   ENKI's SAN, being a single point of failure, is designed to tolerate subsystems failures with redundant controllers, network connections, and disk arrays.  This means your instances are never lost with a hardware failure, and if you turn your instance off, it will be there waiting for you unchanged when you start it again.

Why doesn't ENKI allow my application to use as much memory and CPU as it needs, and bill me on whatever it uses?

For our PrimaCloud service, ENKI's Virtual Private Data Centers are currently based on VMWare, though we plan to offer Xen in Q1'12.  Our VMWare instances have a fixed resource (memory and CPU) allocation, which prevents multiple VPSs/VMs on a single machine from interfering with each other.  While VMWare supports memory over-allocation, essentially a form of virtual memory management that eliminates the requirement that the server keep the entire VM memory in physical RAM, ENKI turns this feature off to maintain predictable performance for your VMs.  This allows us to guarantee you the performance you expect based on your requested resource allocations.  You can always request a change to your VPSs resource allocations (self-service available Q4 2011), and your bill will be based on the weighted average of your requested allocation over the month.

What are the LB (load balancer) and f/w (firewall) appliances vendors ENKI uses?

For PrimaCloud customers, ENKI deploys your choice of LB appliance as a virtual machine.  Our standard choice is HAProxy due to its speed, but we will pick the one that meets your application needs.   We can use open source or commercial appliances such as those from Citrix, Riverbed, or others.  Our muscular servers are generally faster than the fastest hardware appliance, allowing our customers fully self-managed and custom load balancing with no compromise in performance.   ENKI offers a Juniper hardware firewall which can be set to block ports or optionally implement stateful packet inspection including IDS and IPS.   However some of our customers choose to also use software firewall appliances within their virtual private datacenters as well for an extra layer of protection and configurability, or to meet specific requirements for regulatory compliance such as HIPAA or PCI.

You often mention Infiniband on your website.  Why did you choose it and what does it do for me?

ENKI has been in the cloud business since 2006.  We realized early on that as processor power was nearly doubling every year, the bandwidth that a server needed to talk to storage and other servers was going up at the same rate.   Our initial cloud architecture used multiple 1GbE links to connect our servers, and we noticed that our clients were experiencing network delays that affected their application performance.   Similarly, clients of Amazon EC2 and other clouds also came to us complaining about these I/O delays.   We decided to solve the problem with such an overwhelming amount of bandwidth to each server that even the server's PCI bus couldn't keep up.  The technology that could do this was Infiniband, which had its genesis in High Performance Computing.  The result is that each of our servers has 2 40Gbit links connecting it to storage systems and to other servers, for a total of 80Gbit/sec of I/O throughput.