San Diego Cloud Hosting Services

Having “HA” storage gives you redundancy at each and every point – from the power and storage hardware all the way through the network to the hypervisor. The M5 Cloud primary block storage in each Availability Zone runs on fully redundant clusters. Each storage cluster is multi-headed, with internally redundant arrays. All systems have redundant power supplies, network connections, HBAs, and network switches. Even though our world-class data center facilities have industry standard UPS and diesel generator systems, we’ve taken things one step further with additional in-line UPSes for the storage arrays.

Why is this set-up better than local storage as is common in VPS providers? M5’s SAN-based storage allows for virtual machines and disks that can be hot-migrated for faster resizing, in advance of maintenance and in response to hardware failures. This translates to absolutely no down-time for you when we’re doing scheduled hardware maintenance, in most software maitnenance, and very little downtime when a catastrophic hardware failure occurs. All of these issues may cause downtime for typical VPS providers.

Our Primary Storage is connected to the storage network by multiple, bonded, 10GigE connections to multiple switches, in a fabric configuration. The fabric configuration means that if there is a failure of one network component in the data path, then the effective bandwidth is reduced, but there is no delay for fail-over (no spanning tree protocol). We have engineered the storage network with enough bandwidth to make sure that even in a degraded state, there is plenty available for the affected host, cluster, pod, or Availability Zone.

The block storage is based on ZFS, taking advantage of all the performance features that can be added to a ZFS system. The host systems have lots of RAM and the storage arrays are configured with many mirrors. We use mirrors of S-TEC ZeusRAM for the ZIL, and many S-TEC SSDs for the L2ARC. The storage hosts run the fastest cores available in Intel’s Xeon E5 series for fast compression. Compression allows for fewer IO operations and actually increases IO performance in trade for CPU power. For performance purposes, we do not de-duplicate.

The M5 Cloud network is fast, low-latency, and redundant. It uses a converged network fabric with redundant switches and dual, bonded, 10 GigE to every storage node, every hypervisor and to our core network. The channel bonding hash algorithm used gets maximum total throughput from the hardware, and hardware flow control prevents dropped packets during TCP “microbursts.” All maintenance can be performed with zero downtime and no perceivable effect on performance. If any port or cable fails, there is ample capacity on the remaining link.

The M5 Cloud sits on M5 Hosting’s world-class network, with multiple upstream providers, and leverages the power of the InterNAP Flow Control Platform to optimize outbound routes. M5 Hosting is Autonomous System Number 21581.

All of our service offerings specify a minimum guaranteed CPU power, but you can use more when it is available and not in use by other customers. This allows for bulk jobs to go faster when it’s available, but provides a guaranteed level of performance for your latency-sensitive applications. The amount of guaranteed CPU power is specified in cores and in MHz or GHz.

Many other cloud providers and VPS providers just say “vCPU” or “vCores” without specifying how many other people are on your cores. Some only specify that you get a “proportional share” of the CPU power, but again you don’t know how much that is. With M5, you’re never left in the dark.

You can access the console of your instance, as if you were sitting at a keyboard and monitor (or using a KVM switch). This can help with the troubleshooting of non-booting or non-accessible instances.

Our cloud is built on Apache CloudStack. The Apache CloudStack web interface is an easy to use portal for you and your team to manage your cloud resources.