's servers take weekend holiday's servers take weekend holiday

Summary: Charlie Wood runs Spanning Partners, a small company that provides a way to track salesforce information using RSS. He recently blogged about receiving a notice from salesforce.


Charlie Wood runs Spanning Partners, a small company that provides a way to track salesforce information using RSS. He recently blogged about receiving a notice from about scheduled maintenance that shut down his access to for six hours, from 8 PM PDT to 2 AM PDT tonight.

I get notices all the time from about "scheduled downtime". This time the email describes a six-hour window this Friday night during which may not be available.I want to use to capture requests for customer support but those come in at all hours of the day, especially since my business is global and my customers are all over the world. There's no way my support page can be down regularly for hours at a time.

If wants to make itself part of the infrastructure of not only 9-to-5 businesses with American customers but also 24-by-7 businesses with customers all over the world, they're going to have to change something fundamental about their operations.

sfdowntime.jpg is very transparent about the status of its datacenters, and makes its performance data information public. It turns out that two of salesforce's eight servers, serving North American customers, will be down for the six-hour window, for fixing bugs, new features and performance tuning. The maintenance window may not last as long as six hours, according to John Taschek, vice president of market strategy.


The downtime may be acceptable to North American customers who want to take a rest Friday night, but for global customers with operations active during the time period that hit those two servers, it will be the kind of inconvenience for which you want your service provider to be held accountable. Multitenancy has some cost advantages, but individual customers can't schedule their your own downtime.

It's early in the game for software-as-a-service. Google, Yahoo and other major on demand site suffer from outages and scheduled maintenance downtimes for some of their services.

Consumers are more tolerant, or forgiving because the services are generally free, than business users, who should expect at least three nines (99.9 percent uptime), which is 8.76 hours of downtime per year, which salesforce is already bumping up against this year. Uptime of 99.5 is a work week's worth of downtime (43.76 hours).

Last year, suffered outages and has been addressing the problems by adding a datacenter and investing heavily in infrastructure.

In an email Charlie asked:

Does anyone think PayPal could have hours of scheduled downtime and get away with it? Of course not! Why not? Because payment processing is a mission-critical, 24x7 function. Shouldn't sales and support be no different?

Five nines would be downtime of less than 5.26 minutes per year. You have to conclude at this point that, the poster child for the 'Business Web,' is business mission semi-critical.

I asked CEO Marc Benioff about having 25 percent of's mulitenant servers down for six hours. His company is well known for talking about seamless upgrades as new capabilities are added.

Benioff said that the scheduled maintenance is not included in SLAs or uptime numbers. That doesn't make total sense--the servers are either up or down.

He also said that there isn't a company or service that doesn't have to periodically do heavy duty maintenance (I guess that means shut down its service). Does he include the telcos and payment processing services? He added that customers expect regular maintenance, including downtime I presume, to maximize performance. True.

I talked to Wes Benwick, CEO and president Bennett's Business Systems, which has 35 salesforce subscriptions and has been a customers since 2004. He volunteered that he has had performance problems with the salesforce platform lately.

"We've had performance issues. This morning I tried to run a report and got a session error. I clicked the back button and was able to get the report. It's a little aggravating, but I got the work done," Benwick said. . "It stinks that they are having these problems but you expect it with growth they are having."

Regarding the scheduled downtime, Benwick said, "We are an 8-to-5 company. We are used to maintenance windows on weekends. We have more downtime with systems we control in house."

In many instances,'s uptime will be better than what a company can do managing its own infrastructure to support a CRM application. But as the company markets its enterprise platform vision, taking on Oracle, SAP and Microsoft, 99 and a half just won't do.

What do you think?

Topics: Enterprise Software, Hardware, Servers

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • You would think they would use clustering technology

    That way if one goes offline.. users can seamlessly be redirected to another.. or in the case of clustering.. capacity is lost.
    • They are running clustering

      They are running clustering. Dan probably meant 8 instances. I REALLY doubt only has 8 servers in their entire operation.

      The problem is that Servers have to be shut down when you apply patches. Oracle has several dozen patches per quarter that needs to be applied. You can't keep the instance running while patching it.
  • Charlie can capture support cases in during the maintenance

    Charlie should capture his support cases in using the Web2Case or Email2Case feature. It will queue the new cases until the maintenance is complete.
  • Eight servers, you have to be kidding!

    What are they running, converted PS3s :)
    • 8 servers

      Small correction. They are not "servers" rather instances. A collection of hardware
      and software serving dedicated region or user grouping.

      I would be curious to know what was last year uptime for PayPal, Gmail or Microsoft
      services. Does any one know?
      Tomasz O
  • get your house in order

    If I read one more article about Benioff running around touting all the large Enterprise business they've won recently I think I'll scream. has thousands of current customers that they are doing a shoddy job of supporting. Get your house in order and make your current customers happy. It's wrong and unethical that these maintenance times are losing business for their customers - i.e. can't capture prospect query forms 24/7. And the previous posters suggestion about a work around - absurd. Business do not have the time for constant workarounds for basic software functions. At the very least, a CUSTOMER RELATIONSHIP MANAGEMENT system should be able to capture customer online query forms. Forget about google ads, forget about complicated add-ons through apps - just get your basic system running seamlessly. No wonder there's virtually no customer support by the company when it comes from the top - the quote by Benioff is the equivalent of shrugging one's shoulders and saying oh well, I'm on my way to the broker's office to sell yet another $850,000 in my own company's stock as I have been doing daily for the last year or so. Get your house in order before we all walk.
    • Interesting how they say this doesn't count as down time

      Like Dan said; it's either up or down.
  • My "servers" never went down this much

    When I worked in a consulting firm, my "servers" never went down this much. Servers = instance. Instance = many servers.
  • This is tougher then you think

    While I love the idea of kicking SalesForce in the shins a couple of times (we're a competitor -- shameless plug heap CRM And I do think they need to get their house in order; I do want to say that running a large scale web app is extremely difficult. It has a way of revealing your weaknesses in very scary and disastrous ways. Now, thats no excuse, but just so everyone knows just having a cluster doesn't solve all their problems.
  • wht just pick on

    Let's get the on-premise vendors to share their customer up time stats and costs...then we can benchmakr all of them - see