Before the advent of Peak Hosting’s AlwaysUp Architecture,® downtime was seen as an unavoidable drawback of running a technical operations department. Conventional wisdom held that downtime could be minimized, but it was next to impossible to eradicate. We took this as a challenge to overcome, and Peak Hosting has raised the bar for system availability: downtime is unacceptable for any reason. No Excuses. No Exceptions. No Downtime.
Peak Hosting’s unrivaled consulting experience architecting, building and running the largest and most profitable web properties has enabled us to create Best Practices that ensure that you have the right hardware, networking, architecture, and procedures to ensure that your service is always up! Plus, we have an experienced staff – the best engineers and architects on the Internet – ready to fill any “knowledge gaps” encountered on the road to implementing your complete hosted solution.
We custom design your solution and provide implementation and support teams capable of delivering what your company needs, at the lowest possible total cost of ownership. Even before they walk through the door, Peak Hosting experts know the right architecture for your vertical, so your site will never fail in the first place. Peak Hosting’s AlwaysUp Architecture,® with built-in redundancy designed for every moving part, means that component failure will never escalate to service failure. Peak Hosting’s technical operations service takes care of everything: architecture and design, server hardware, storage, network, load balancing, operating system, applications, security, out-of-band, data centers, BGP, virtualization, database, monitoring, alerting, performance trend analysis, back-up, and archiving – all while significantly reducing your total operating costs and ensuring that downtime never happens.
100% AlwaysUp SLA® No Excuse. No Exception.
Peak Hosting stands alone as the only cloud or managed hosting provider to deliver a 100% SLA, with zero exceptions or carve outs for scheduled or emergency maintenance windows. Only Peak Hosting is able to maintain, repair, and upgrade our environment with zero interruption to our clients’ service. Our expertise, architecture and design provide the certainty that your service will never be down. Most managed hosting providers claim to provide 100% uptime. However, they all carve out scheduled maintenance windows and emergency maintenance windows, which are effectively defined as, “Whenever they’re needed, for as long as it takes.” This is unacceptable! To say you will always be up except for when you’ll be down is saying nothing.
An outsourced managed service provider, whether the architecture is in the cloud, a dedicated server, or some hybrid, is either committed to being up 100% of the time, or it is not.
No company should be down because it’s convenient, cheaper, or easier for their provider to take them offline while performing maintenance. Unfortunately, commodity managed hosting vendors have learned they can increase profitability by skimping on redundancy. They simply don’t consider the complete ramifications of system or component failure to their customers’ business. They believe if they swap “like for like” hardware within an hour, all will be forgiven. However, your staff will still have to spend hours or even days re-deploying your code, data, and configuration to fully restore your service. At Peak Hosting we ask the question, “Instead of replacing it in an hour, what if it never failed in the first place?”
Peak Hosting’s fully 2N architecture means that component failure will never escalate into service failure. Even though we deploy only enterprise-class hardware, it’s not as if our power supplies, hard drives, or PDU (power distribution units) fail less often than anyone else’s. What makes a big difference in our client’s outsourced experience is Peak Hosting’s commitment to preventing failures before they happen. We constantly ask: “What if it didn’t fail in the first place?” We find it unacceptable, the widespread rule of thumb that if failed hardware is repaired within an hour, everything is fine. Everything is NOT fine if your system is down for an hour. This lame service philosophy inevitably results in lost customers. If you are always up, you’ll make more money because the cost of regaining a customer that’s been lost due to downtime is enormous.
A full one-third of the cost of Peak Hosting servers goes into redundancy for extra power supplies, hard drives, RAID controllers, PDU, UPS, Generators − everything needed to ensure continuous uptime. Our industry veterans designed this level of redundancy into Peak Hosting’s architecture, knowing that the impact of downtime to your business is massive. Every time your staff is pulled away from their job to redeploy code, configure applications, add data to systems that failed and then replace hardware that is now essentially useless, it costs you money. Peak Hosting’s 2N AlwaysUp Architecture® is designed so that every moving part on a server can be maintained without taking the service offline.
How does Peak Hosting protect against motherboard or CPU failures, problems that can be catastrophic? We’re not miracle workers, but we can plan ahead by using a bullet-proof server design. Every service on Peak Hosting servers is striped across different physical systems in different racks with no single points of failure (SPOF) between them: the applications and the service remain online, even when an individual component fails. Additionally, Peak Hosting only deploys Dell hardware, so our clients can rest assured they have the highest quality and most reliable equipment in the industry. And if something goes wrong, our clients incur no extra costs. As part of our full Operations-as-a-Service offering, we never charge service fees for maintaining, repairing, or supporting your environment.
Of course, it would be cheaper to cobble together commodity “white label” servers. But at the end of the day, the client would pay dearly when they are repeatedly offline because of junk hardware failures, inconsistencies between different shipments, and their absence of 2N redundancy.
Our diagrams demonstrate that Peak Hosting’s servers have the most redundancy possible and that component failure (network, PDU, power, hard drive) does not require taking the server offline. Because we stripe our clients’ servers so that any two servers performing the same function (web, database, noSQL, Hadoop, etc.) are not in the same physical rack, there are zero potential ways the service can be interrupted.
AlwaysUp SAN,® part our AlwaysUp Architecture®, extends to every element of the Peak Hosting platform. Our EMC VNX 7500 SAN solution has multiple controllers, power supplies, 10 Gig iSCSI interfaces − everything that it needs to remain continuously up. Additionally, our EMC VNX Monitoring and Reporting system continuously monitors every one of Peak Hosting’s SANs from every one of our data centers and proactively notifies EMC of potential hardware problems BEFORE they fail! What other Managed Hosting provider can claim their SAN vendor will dispatch both technicians and spare parts for your JBOD or direct-attached hard drives? Peak Hosting doesn’t believe in wasting time and money reacting to problems. Instead, we proactively keep them from happening in the first place.
Peak Hosting’s AlwaysUp Architecture® extends to our network as well. Every Peak Hosting network device − routers, switches, load balancers, firewalls, VPN, etc. − has dual power supplies and is deployed in a 2N configuration so there is always a back-up system providing continuous service.
Even so, there is no way Peak Hosting can control the ups and downs of individual parts of the Internet. However, we do everything we can to ensure connectivity. Here again is where the depth of experience and the pedigree of our engineers and architects come into play. Our staff has built some of the largest networks on the planet, so they know how many things can go wrong with Internet connectivity. They know the complexities of BGP (Border Gateway Protocol) and how ISPs speak to each other. This is why Peak Hosting has eight different ISPs to control both the inbound and outbound path for your traffic. Whenever an individual ISP is having a “bad day,” we route traffic around the problem so your customers and revenue are not impacted – and your cash register never stops ringing.
One of the most common reasons for downtime is network unreachability. ISPs are polite enough to tell their customers when they can expect to see a service interruption, euphemistically called maintenance window notifications. Other managed hosting and cloud providers are happy enough to let BGP rebalance their traffic after an ISP goes down (even if this means at a minimum 90 seconds of traffic being black holed).
At Peak Hosting, we prefer our clients never even suspect there might be a problem − we have so many ISPs that we gracefully bring down any ISP performing a maintenance window. We do this two hours before the window starts, then bring it back into service the following morning when we’ve contacted their NOC to receive the “all clear.” This is all invisible to you and you will never be interrupted in the middle of the night by an unpleasant alarm saying your site is down. We prefer that our clients are never disturbed in the first place. It’s all part of our, “What’s the most we can do?” attitude, ensuring that Peak Hosting clients have the best possible experience.
Other providers offer only 1 Gig or perhaps as much as 4 Gig trunks to connect network equipment, only Peak Hosting deploys 10 Gig Ethernet across our entire topology. This ensures that your applications are never slowed because of congestion from other random third-parties, over which you have no control.
Peak Hosting believes that 10 Gig Ethernet should be the standard for reliability. Modern applications such as Hadoop and NoSQL can easily generate over 700 Mb/sec per server. Modern applications drive bandwidth unheard of just a few years ago. If your managed service or cloud provider is not entirely 10 Gig, your server and applications will be fighting for network throughput and subject to packet loss, high latency, and jitter (variability in latency that destroys performance in all distributed applications).
With Peak Hosting’s network architecture, your systems enjoy being just two network hops between devices. This keeps the “network diameter” as small as possible while still being fully 2N redundant. This network topology only works because of our proprietary “pod” design to limit the total device hop count between systems. While this may sound like a minor point, it’s massively impactful for latency, packet loss, late collisions and jitter (variations in latency). Peak Hosting is the only provider who can deliver 20 microseconds of latency between servers where public clouds languish with 1,000 microseconds (1 millisecond) to 1,000 milliseconds.