Mission critical computing for the masses
Everyone benefits from networking advances
The provisioning highly-available IT service, once the sole province of large enterprises, is today available to everyone.
Consider the ftServer line from Stratus. These servers contain numerous redundant components, are equipped with dozens of sensors and are shipped with “uptime assurance” software monitored by Stratus as a managed service.
Stratus claims 99.999+ per cent uptime and has entry-level servers within financial reach of most small businesses.
A whole new board game
For the do-it-yourself whitebox crowd, advances in hardware resiliency and commoditisation over the past decade have been game changing.
Chenbro, long a whitebox favourite, offers chassis with various redundant and hot swappable components.
Networking advances have trickled down to common availability. The need for outdated protocols such as RIP has almost disappeared, replaced instead by much more resilient successors such as OSPF, EIGRP and BGP.
Life’s a switch
Switching features such as port trunking, spanning tree and VLANs are now tick-box options available on commodity switches from companies such as D-Link.
Support for multipath storage (MPIO) is everywhere. Although MPIO disks are still the realm of truly enterprise fibre-channel drives, network-based storage technologies such as iSCSI have brought storage redundancy and reliability to everyone.
In a great example of commoditisation, Microsoft has joined in, offering a very mature and proven software iSCSI target free to anyone with Server 2008 R2.
One download launched iSCSI from the storage appliance and deep Linux voodoo market into mainstream small-business availability overnight.
Getting a cluster running was once a black art; today, it is simple. Combine Server 2008 R2’s Failover Clustering with the iSCSI Target and Hyper-V. Your collection of Wintel machines have become a point-and-click cluster of highly available servers.
Disaster-aware software is another popular trend. It is increasingly common to find applications coded to deal with common hardware or network faults.
No omelette please
Modern hybrid cloud applications are environmentally aware , operating in different modes depending on network conditions and synchronising data between instances when connectivity is restored.
Worried about putting all your mission-critical eggs in one cloudy basket? Companies like Rightscale have you covered, offering portability across multiple clouds.
Cloud brokerage – creating a “cloud of clouds” – is becoming a viable way for organisations of all sizes to skip in-house IT all together.
The past decade brought high availability and fault tolerance to Wintel and Lintel computers near you. The result is an expanded skills base with real-world experience in mission-critical computing.
The next decade belongs to the cloud. From companies the size of Intel to your humble local sysadmin, the skills, tools, hardware and experience to manage mission-critical cloud services are everywhere.
More importantly, they are becoming highly available to all budgets. ®
Critical thinking needed
I read this and couldn't help but think of far too many freshly qualified noobs I've met over the years - usually the type who have just done some industry cert and now know all there is to know.
This is new! Therefore it's better! And all our problems will be solved!
There's never any reference to whether a) there was a real-world problem in the first place b) if there was then this this actually solves it and c) even if the first two conditions are met, whether the benefits are worth the costs of adoption. This whole article suffers from that same lack of critical analysis.
So yes, thank you for reminding us of some of the more significant develpments of the last 15 years, but without any additional context this is little more than a list of management-speak buzzwords.
Stratus approach is so Pre 1990s.
First we had VAX Cluster.
Then in 1999 I think Cluster with 2 x NT servers sharing at least two external storage shelves on separate 2 x SCSI bus (3 was better performance)
kill power randomly on one Server or one Storage shelf and no interruption of service. A cheap commodity server with single PSU is fine.
Obviously each shelf and server needs it's own PSU and separate UPS. But no server or shelf level expensive on board redundancy.
I have never - honestly, not once in my life - believed something was "better" because it was new. I viciously and vociferously mock people who equate "new" with automatically "better." So I will take your comment in stride and try to do better next time.
As you pointed out, the article really was largely aimed at “reminding folks of some of the more significant developments of the last 15 years.” This article was never really meant to be a standalone: it is part of a series of interrelated articles. It is possible you have not read my previous articles on the topic that do go into far more depth on many of the issues you raised. Here are links:
As to “was there a problem in the first place,” well...yes! The problem is the same as it has always been: how to most efficiently deliver IT services. There isn’t “one true answer” to that problem; cloudy services are just one more tool in the toolbox. Neither innately good nor bad for their recent appearance; cloud computing must be viewed with the same sceptical eye we would use when analysing any technology.