Slashdot Mirror


What Hurricane Sandy Taught IT About Disaster Preparedness

StewBeans writes: The National Oceanic and Atmospheric Administration Climate Prediction Center is calling for calmer than normal storm activity this hurricane season, which runs through Nov. 30. But it's likely that data centers and IT companies in NYC are still taking disaster preparedness seriously. Three years ago, Hurricane Sandy devastated homes, businesses, transportation, and communication in New York, and taught many companies (the hard way) how to keep the lights on when the lights were literally off for weeks on end. Alphonzo Albright, former CIO of the Office of Information Technology in New York City, gives a behind-the-scenes account of what life and business were like in the dark, cold days following Hurricane Sandy in NYC. He also shares tips for other tech leaders to create their own Business Continuity Plan in case this year's storms take a turn for the worse.

27 of 68 comments (clear)

  1. Geographic diversity by Todd+Knarr · · Score: 5, Informative

    First rule: have facilities capable of running your business in more than one location. Everywhere is susceptible to disaster of one sort or another, but if you pick areas far apart that aren't geographically similar they probably won't both suffer disasters at the same time.

    Second rule: the probability of disaster taking out your main facilities is 100%. It will happen. The only question is exactly when it'll happen, and the only constant in the answer is that it won't be at a good time. If anyone in your organization doesn't like this, remind them that reality doesn't really care what they like.

    1. Re:Geographic diversity by turbidostato · · Score: 4, Insightful

      I should add a rule zero then: Take your time to properly understand your costs and revenues so you can make a sensible investment. Maybe it ends up being cheaper just to close door for a week every 30 years than your A-Bomb-proof continuity plan.

      And then a zero-plus: Make sure you get business-aligment in written. Maybe the board member that agreed to your investment-sensible less-than-A-Bomb-proof continuity plan wants you as scapegoat once the shit hits the fan.

    2. Re:Geographic diversity by Lumpy · · Score: 4, Interesting

      A week? most data disaster you are down for at least 30 days. Hell you cant get an order for servers in from DELL even on rush faster than 2 weeks.

      If your company can survive zero revenue and 100% loss for 30 days, you either are sitting on a mountain of money, or your business is more of a hobby than anything else.

      Oh and if you lose your accounting data due to lack of a bomb proof plan, expect fines in the high 6 figure range.

      --
      Do not look at laser with remaining good eye.
    3. Re:Geographic diversity by nine-times · · Score: 3, Interesting

      Do companies actually bulk-order from Dell any more? This is actually the most I've heard about Dell for months.

      I know Google manufactures their own computers, for the most part.

      So you think just because Google builds their own servers, it must be that everyone else does the same? There are a few companies out there that aren't Google, and yes, many of them still buy from Dell or HP.

    4. Re:Geographic diversity by Khyber · · Score: 2, Informative

      "I know Google manufactures their own computers, for the most part."

      As a former Google employee, I must say you are full of shit.

      Show me Google's manufacturing plants, please.

      --
      Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
    5. Re:Geographic diversity by nine-times · · Score: 4, Insightful

      Take your time to properly understand your costs and revenues so you can make a sensible investment. Maybe it ends up being cheaper just to close door for a week every 30 years than your A-Bomb-proof continuity plan.

      This is an amazingly difficult concept to get people to understand. I've had way too many conversations with people who are sure they need an instantaneous failure-proof disaster recovery plan. They believe their servers should be constantly in sync with multiple copies in various places, such that in the even of a short internet outage, their servers will fail over to an outside copy, and then fail back when the outage ends, automatically and without skipping a beat. Unfortunately, they're willing to spend approximately $0 to achieve this, but that should be fine, because "the cloud" is pretty much free, right?

      It's a similar problem with security. Everyone wants all of their data to be completely secure without any possibility of being compromised under any circumstances, but they also want it to be as convenient as if the data is unsecured, and they don't expect to pay extra for any of it.

      I always try to explain that it's about trade-offs. I can make your data much more secure than it is now, but it'll cost you money, and you'll have to jump through extra hoops to get access to your own data. I can replicate what you need to a remote server, yes, but then you have to pay for the remote server. Depending on exactly what we're talking about, it might not be a real-time sync, or it might not result in anything like an automatic failover. Those things might require special software or services or licenses. Pay enough, and yes, I can probably get you a real-time sync with automatic failover and fail-back, but even then, you could still have an outage. The system that keeps everything in sync and triggers the failover could be the component that fails. Or if there's a total blackout on the east cost, it might not matter that there's a complete replica automatically started on the west coast, if all your employees are on the east coast and without power.

      It's trade-offs. Spend enough money and put up with enough limitations, and you'll get something that does what you want, although imperfectly. Most of the time, for most businesses, it doesn't make sense. "Good enough" is good enough. But people don't like to be told, "A pretty secure network with a pretty good disaster recovery plan is appropriate for you." It makes them feel unimportant, which most executives and business owners can't live with. They want to know that they should have the best thing possible.

    6. Re:Geographic diversity by dgatwood · · Score: 3, Interesting

      A week? most data disaster you are down for at least 30 days. Hell you cant get an order for servers in from DELL even on rush faster than 2 weeks.

      Maybe true, but you can get a cloud server deployed in a matter of minutes, and you can use that as a temporary (expensive) alternative to servers under your complete control.

      If your company can survive zero revenue and 100% loss for 30 days, you either are sitting on a mountain of money, or your business is more of a hobby than anything else.

      You're making a lot of assumptions that aren't necessarily valid. The amount of downtime and the impact depends heavily on the nature of the company, and in particular, whether sales/income depends on maintaining continuous operation of the business. Take, for example, a company that makes software:

      • On the development side, even if a company's entire repository went away tomorrow, and even if half the development team died, a typical software company could still get back all but the last few days' work (and perhaps a few old branches) by configuring a github instance on Amazon's Elastic Cloud and having the remaining developers push all of the branches from their local checkouts. Downtime would be minimal.
      • On the distribution site, most software companies would be completely unaffected, because distribution is usually handled by a large third-party merchant (Apple, Google, etc.).

      So unless a software company requires critical server infrastructure beyond what they get for free via iCloud, etc., it probably needs very little in the way of disaster preparedness, because the very nature of the work and the tools involved lends itself to being prepared for a disaster automatically.

      On the opposite end of the spectrum, cloud service providers and Internet service companies must have disaster preparedness plans in place, or else everybody who uses their services is screwed. And if they're down for even a couple of days, they're probably going out of business. If Facebook went down for a week, Google+ would become the #1 social network.

      --

      Check out my sci-fi/humor trilogy at PatriotsBooks.

    7. Re:Geographic diversity by sjames · · Score: 2

      It does bring up a good point though. There is a lot of space between the A bomb proof data center with the geographically diverse duplicate data center with hot cutover and DR, what's that?

      As you point out, data backup is essential, but that doesn't imply a full duplicate data center. It may be that a very minimal setup is enough to limp along for a few weeks while things get back to normal. Limping doesn't necessarily mean no revenue.

      It's also useful to note that downtime due to storms and such doesn't necessarily mean lost equipment.

    8. Re:Geographic diversity by fustakrakich · · Score: 4, Funny

      Show me Google's manufacturing plants, please.

      Aren't they on the North Pole? I hope they can float...

      --
      “He’s not deformed, he’s just drunk!”
    9. Re:Geographic diversity by tlambert · · Score: 4, Insightful

      "I know Google manufactures their own computers, for the most part."

      As a former Google employee, I must say you are full of shit.

      Show me Google's manufacturing plants, please.

      As a former Google employee myself, I'm bound by my NDA from naming the East Asia contractors who build the actual equipment. Google generally only provides the reference implementation.

      Do you think Dell builds their own boards? They don't. The majority of their server class motherboards are manufactured by ASUS, based on Intel reference designs (Intel also no longer manufactures desktop motherboards, as of Haswell -- yields were too low).

      If you are curious about who made your motherboard, and run Windows, use the following command:
      wmic baseboard det product,Manufacturer,version,serialnumber

      (If you want a GUI version, download "Speccy", run it, and either look for the "Motherboard" section in the "Summary" view, or click on the "Motherboard" list item to get only that information by itself).

      Other OS's have their own commands, as an exercise for the student.

      P.S.: If the information has been obfuscated, you can usually back-track by looking at the BIOS vendor and version information, and then using searches for updated/same versions of the BIOS based on that, to see which platforms the BIOS vendor says it's for. You are welcome.

    10. Re:Geographic diversity by turbidostato · · Score: 2

      "This is an amazingly difficult concept to get people to understand. I've had way too many conversations with people who are sure they need an instantaneous failure-proof disaster recovery plan."

      Given your nickname it seems no wonder that you grasp the concept. Exactly yes to all you say.

      "It makes them feel unimportant, which most executives and business owners can't live with."

      That's true, but I'd say it's only half of the story. Specially business owners are quite sensible to the money part and quickly understand that there's no point in expending 1M upfront and then 100.000 a year in protecting from an outage with a 10 years recurrence that risks 500.000 in loses. From my experience, the other half of the story, specially dealing with non-director's board level management is the CYA part. You make the assess, the numbers are sensible and everybody agrees with that but, in the end, you are still gambling; most managers will gladly pay a lot of (company's) money just to avoid the chance of being the one having to go upstairs to tell the big boys that shit happened above and beyond the disaster recovery plan coverage -just as expected, but still...

      Some anecdotes, just for fun:
      -We need wide geographical coverage. We can't stand a local disaster since it would make us lose about 500.000 a day.
      -Well, your CRM says 80% of your billing comes from companies within this industry complex... do you think a local disaster won't impact your sellings anyway?

      -We need new offices within 72 hours in case this building is lost because we can't afford paying our employees for nothing.
      -Well, do you remember that, even if the building is gone off-hours without live loses, as per local labour law the company can forcibly set the dates of 11 out of the standard 22 days of yearly holidays so in case of disaster you can send your people home two weeks without it costing you an extra dime in wages nor lost productivity along a year? (this one was not so simple, but the nut of the case was this)

      -We need live migration between those two datacenters so in case of disaster there's sub-minute service disruption.
      -You see... this is quite costly and all you are basically doing is serving almost 100% static content with only minor changes. Wouldn't it be better is you spread the static load and/or let the content converge somewhere between 1 and 6 hours and drive down the costs ten-fold?

    11. Re:Geographic diversity by cyberchondriac · · Score: 2

      That made my day, thanks. I'm familiar with the "We think we should have all the greatest and best services that Fortune500 companies enjoy, even though we're funded like a corner convenience store.. and can we implement this with fewer employees and a smaller budget than we had 5 to 10 years ago?"

      --

      Look back up at my post, now look back down, you're on the Internet. Now look back up. I'm a signature.
    12. Re:Geographic diversity by Software · · Score: 2
      That should be

      wmic baseboard get product,Manufacturer,version,serialnumber

  2. What it should have taught by CanadianMacFan · · Score: 3, Insightful

    Nothing. Disaster recovery plans are like backups... if you don't test them every so often then you assume that they don't work.

    Companies should have already tested their plans and known that they worked so that when any interruption from the storm kicks in their backups would take over as planned.

  3. Re:New York by turbidostato · · Score: 3, Funny

    "And what the fuck is up with "The National Oceanic and Atmospheric Administration Climate Prediction Center is calling for calmer than normal storm activity this hurricane season"? You don't call for that, you predict it. Calls are to be answered (or not). Predictions are to be met (or not)."

    Nononono... This is the United States of Almighty America. When NOAA calls, hurricanes abide! (or else, we send Chuck Norris).

  4. Katrina & A Datacenter/ISP by i.r.id10t · · Score: 2

    Wasn't there a datacenter guy who posted here on /. when Katrina hit about all the stuff they went through keepign things up and running at some sort of minimal level?

    Been drinking and google-fu is off but perhaps someone can post it. IIRC it included a blog of what was goign on, etc.

    --
    Don't blame me, I voted for Kodos
  5. No Serious Precautions Taken by JimSadler · · Score: 2

    First we must be willing to build very large facilities capable of storing thousands of tents and food enough to last people for many weeks. And we need to do this in regional centers such that deliveries can take place the day after a storm leaves an area. Areas such as Miami simply can not be evacuated as the population is way too large. A strong storm can knock out roads and rails and put a large area into severe isolation. A few days worth of groceries will simply not help much. In my area our grocery stores were destroyed when three storms hit us back to back. Getting a car on the road was next to impossible and quite dangerous. Gasoline was not available at all as the gas wells were all flooded. I had no power for three solid weeks. A situation like that can get to the point at which people raid each other trying to keep from starving. I see no effective measures at all. If another Katrina hit New Orleans the results would be very much like the original Katrina. The repairs made are not designed to stand up to a class 5 hurricane. And it is only a matter of time before New Orleans gets hit by class 5 storm. Miami is in the same boat. We roll the dice here constantly and count on good luck not to bring on a class 5 storm.

  6. Re:Not an option for High Frequency Traders by Hognoxious · · Score: 3, Insightful

    Terrible. How will we cope without them?

    --
    Confucius say, "Find worm in apple - bad. Find half a worm - worse."
  7. 1920s electricity infrastructure by dbIII · · Score: 2

    It taught us that 1920s electricity infrastructure shouldn't be in use in one of the richest cities on the planet - wet wood in contact with high voltage is a bad idea and the inevitable fires happened.
    Funny thing is I know a transmission guy who said "I told you so" based on what he said in the 1960s. Fifty years later that shit was still in service and it burned.

  8. History repeats itself by Monoman · · Score: 2

    Eventually things pretty much go back to the way they were before. I remember seeing a discussion about the lessons learned from Hurricane Andrew (not just IT specific) and how after 7 years things that were important were forgotten or deemed less important. I'm sure the same happened with Hurricane Katrina, Sandy, and many others. It seems to be our human nature that these things eventually wear off and become less important. I think Neil Degrasse Tyson was on Joe Rogan's podcast a few years ago and touched on the subject as well.

    --
    Keep the Classic Slashdot.
  9. Test your backup generators .. by nickweller · · Score: 2

    The backup generators failed as the fuel pumps couldn't be powered as there wasn't any electricity to power the pumps ref. Don't site your critical infrastructure in the basement. ref.

    1. Re:Test your backup generators .. by Monoman · · Score: 2

      Backup generators stopped running when they ray out of fuel because the tanks couldn't be refueled for various reasons. Power outages, roads blocked, behind on schedule, etc, etc. .... Nobody thought they would need an SLA on refills.

      --
      Keep the Classic Slashdot.
    2. Re:Test your backup generators .. by Cid+Highwind · · Score: 2

      Or they thought having an SLA for refills meant their fuel trucks get through... even when trees/power lines/National Guard soldiers are blocking the road.

      --
      0 1 - just my two bits
    3. Re:Test your backup generators .. by Monoman · · Score: 2

      Exactly. PHBs think an SLA is a guarantee. Even a "guarantee" isn't a guarantee in some circumstances. I live in "hurricane alley" and can tell you that when a storm hits that things don't usually go as planned.

      --
      Keep the Classic Slashdot.
  10. Final Rule: Test the Disaster Recovery Plan by wiredog · · Score: 2

    Story time: A few years ago I was working on a web app for a US intel/LEO agency in northern virginia. The app had started as a demo, then kind of grew. Like a fungus. It was never really designed, much less designed to shut down and restart unexpectedly. There were some other similarly "designed" apps running in the data center.

    The data center, being under the flight path for an airport, had a continuity of operations ("coop") plan and hardware. The "UPS" was a big generator with a switch so that it would take over when mains power went down. There was also a system designed to handle hot mirroring of everything and switch all network traffic to the backup center if the main center went down.

    A great system which was never tested because what if the test takes the system down for 15 minutes and we thus miss the opportunity to prevent the Next 9/11 and Thousands Die and, worse yet, we have to testify in front of Congress?

    So one day the fire marshall came through the building and, as part of his testing, hit the Big Red Switch. The switch designed to detect this and start the generators (and which was reported to cost $15) failed. All the systems went down, hard. The network switch in place to notify the hot backup site and send all the traffic there also failed. And the Vital Systems Protecting Our Nation From the Next 9/11 went down, worldwide.

    Don't just have a plan, test it.

    p.s. We never were able to determine how much, if any, data was lost....

  11. Re:Not an option for High Frequency Traders by sociocapitalist · · Score: 2

    Not an option for High Frequency Traders. Geographic diversity means locating your fiber optic connect further way from the transatlantic fiber head ends which make HFT possible.

    Nope -

    FHTs will typically have a presence in two colos per exchange and, depending on where they trade, with multiple exchanges each exchange being in a different country or region.

    If the disaster is big enough to take out both colos for a given exchange it has probably taken out the exchange as well.

    --
    blindly antisocialist = antisocial