Gmail’s last major outage was in May, and it had a number of outages in 2008. But Google says its service is still far more reliable than corporate networks, many of which run Microsoft Exchange servers.
Today, Fortune commented, saying:
In fact, most e-mail goes down. According to Osterman Research, based in Black Diamond, Wash., e-mail systems in mid-size and large organizations have a mean of 53 minutes of unplanned down time in a typical month. Add that up and it becomes 10.6 hours annually. By contrast, observers point out that Gmail has crashed only a few times this year.
In some of the Domino shops with which I am familiar, which happen to be mid-sized and large organizations in Northeastern Ohio, their mail systems are never down. Ever. How is this possible? For starters, most of these companies are not running their mail systems on Windows. iSeries, Solaris, AIX, and Linux servers are the standard. Add to this, they have clustered the servers; some of these companies have the clusters on-site, others use remote sites. Carefully managing their maintenance windows, which do not impact the uptime statistics, contributes to the 100% uptime number. Even with the maintenance window, the clustered servers take care of availability.
There are those that will say that no e-mail system can have 100% uptime, and yet, I can point to several companies that can not only say that, they can prove it. I am certain that I am not alone in saying that in some Domino shops, e-mail service is always available. Management has made the decision that e-mail is crucial to their business and have made the necessary decisions to guarantee that their employees will not lose productivity from a lack of service.
"Never down" or "100% uptime." It may seem unusual, but it is certainly possible.
Link: Wired: Google Explains Why You Didn’t Have Gmail
Link: Fortune: Is Gmail Ready for Business?
Comment posted by kenio carvalho09/04/2009 07:43:34 AM
Homepage: http://www.plansis.com.br/blogkenio
I agree .
Even in a migration process (i have done it) 7.0.2 to 8.5, using clustered servers, the email service never goes down.
Comment posted by David Schaffer09/04/2009 08:02:20 AM
Homepage: http://bloginprogress.us
Agreed. Even in a very small environment I support the last two unscheduled mail outages were caused by switch failures, and since we clustered the mail server there are no more scheduled outages.
Comment posted by tom oneil09/04/2009 08:50:13 AM
You knock on wood when you write this things... right?
Comment posted by Andy Pedisich09/04/2009 09:29:29 AM
Homepage: http://www.andypedisich.com
Clustering, unfortunately, still seems to be one of the secret pleasures of Notes. I'm encountering several older domains that are adding clustering for the first time.
That's amazing to me since the technology is certainly mature and shrink wrapped. It's easy to do and easy to maintain.
If HTTP clustering was a bit more straightforward we'd be 100% set.
Comment posted by Richard Moy09/04/2009 09:58:19 AM
Homepage: http://www.dominointerface.com
The replication and clustering allow 100% up time especially if the servers are located in two different locations with different bandwidth and power providers.
Comment posted by Dan Lynch09/04/2009 10:04:16 AM
We are one of those Northeastern Ohio shops and we see very, very little unplanned downtime with Domino mail services, and we are not clustered. We have enjoyed increasing stability with each release of domino on both Solaris (North America) and Wintel (non-North American). At least 1/2 of the unplanned downtime we may see, is due to things non-domino (platform, o/s, network, facilities, etc). The keys seem to be better code, outstanding admins, and dedication to working all downtime issues with IBM and getting hot fixes and other help that prevents those issues going forward.
Apart from the ridiculous patching frequency that requiresd planned (ansd annoying) downtime, and as much as we may like to bash Microsoft, the w2k3 servers we have around the world running domino 7 and 8 are as rock solid, or in some instances better, than the Solaris boxes. Less obscure code issues, less obscure kernel bugs, etc, no doubt related to the higher prevalence of those platforms.
Many who suggest the cloud is better more reliable etc, particularly the Google cloud, are simply not well informed. If Osterman's research methods are as "rigorous" as some of the others, that kind of information is probably based on a poor sampling methods, anecdotal information, faulty interpretations or many other sins of poor research methodology.
Comment posted by Charles Robinson09/04/2009 11:26:39 AM
Homepage: http://www.cubert.net
In this case it was an issue with rolling your own mail infrastructure, down to the transport layer. Google botched their transport algorithms so they effectively shot themselves in the foot.
I could bring Domino down with some fat-fingered DNS changes or a misconfigured piece of network equipment, too. Barring anything like that, "100% uptime" sounds impossible to people not familiar with Domino because no other messaging system can achieve it in normal operational environments. No matter how you configure any other mail system I can come up with a plausible scenario that brings it down completely or causes information loss, at least temporarily. That simply can't happen with Domino.
Comment posted by Keith Brooks09/04/2009 12:31:16 PM
Homepage: http://www.vanessabrooks.com
Most of my clients run Windows servers, because the products, ahem, require it, and a few RHEL.
If we build it it stays up.
Charles is correct, DNS and other issues may make it appear like Domino is down, but the server never lies.
Most of mine are only brought down to install MS patches every 2 months but to run 90+ days is not unusual for some of them before we purposefully shut them down.
Good admins are hard to find and worth their weight in gold.
Comment posted by David Schaffer09/04/2009 12:59:01 PM
Homepage: http://bloginprogress.us
Forgot to mention -- 100% up time won't stop users from asking "Why is the mail server down?" every time they have an issue with their computer...
Comment posted by Pierre Passin09/05/2009 09:49:21 PM
I personally find the "cluster-awareness" of the notes client too weak. In most cases, users have to do something (get out/get in, etc...) and some users just do not know about it. So even if the mail is up for some less informed people it is definitely down.
Comment posted by Charles Robinson09/07/2009 10:45:32 AM
Homepage: http://www.cubert.net
Pierre, what version of Notes are you using? In 8.5 the Notes client pops up a prompt telling the user the server is no long available and asking if they want to switch to a different one. There has been some discussion of allowing administrators to do away with this prompt via policy and simply have it silently fail over. I don't know if that has made it into 8.5.1 or not.
I do agree, though, that this was a very long time coming.
BlogSphere V1.3.1
Join The WebLog Revolution at BlogSphere.net