IT Employment
by mkbyers
I respect Mike's opinion, but he was way out in left field suggesting mainframes are only just now waking up to the need for disaster recovery. Mainframers invented it and honed the techniques that pc-based architecture DR now claims as its own.
I've been involved in DR for 10 years now. Each year our mainframe systems & application staff travels to a DR site with backup tapes and totally recreates our systems (2 mainframes). We are able to recover to the day before the "disaster", as we take incremental backups every night and send offsite. We have 72 hours to totally recover our systems to support our 14,000 users who are located all over the country.
I think that Mike's point was that recovery time objectives have gotten shorter and that recovery time objectives have gotten much closer to real-time. Further, that the new RTO and RPO are not achievable using only tape backup.
If you really have 72 hours to get things back to normal for your 14,000 users, consider yourselves extremely lucky. And try not to act surprised when you find out that nobody has considered the implications of that 72 hours in the last ten years!
by wrlang
I think both your posts are true. Mike states very clearly in his first two paragraphs that mainframe DR is being ignored by some companies because mainframes are very stable. He then goes on to make another point about the volatility of RTO and RPO.
His points are well taken. That people tend to concentrate DR first on the systems that fail most often. Mainframe DR could take a backseat to other platforms since they fail more often leaving little time to develop DR plans for mainframes.
I personally don't know of any mainframe shop that has no DR for their mainframe, but know many shops that have no DR for other platforms.
An interesting viewpoint nonetheless.
by mkbyers
I should have made it clear that we have 72 hours of testing time. We actually have our systems operational in about 24 hours. Our mandate, and this is a government mandate, since we are a contractor, is actually 72 hours. Our timeframe to get things back to normal is re-evaluated every year, so there shouldn't be any surprises there.
While our being unavailable is a major pain to our users, unavailibility of our applications are not life threatening, nor would it lead to loss of revenue,although there are penalty implications (or ultimately, loss of the contract) if we fail to become operational within that timeframe. So, yes, we are lucky that that type of pressure wouldn't be on us.
As far as recovery time objectives are concerned, I wasn't considering anybody else's because the questioned asked was: How does your mainframe shop handle DR? The type of DR planning that we have in place would be totally innapropriate for an airline, for example.
by mkbyers
This isn't about whether or not the hardware platform falls apart. It's about having your machinery blown to smithereens and having to leave the premises and recreate your environment at another site! Most companies cannot afford to have a mirrored mainframe in a different physical site. But some companies, like the NY Stock Exchange, probably can't afford not to have this.
That the mainfame runs like a Maytag washer should have no impact whatsoever on whether or not people have a DR planin place. That's akin to saying because my Honda car is so reliable, I don't think I need insurance for a possible wreck.
by miketalonnyc
In reality, DR is a horribly overused term =)
High Availability, wich people often use the term DR to describe, is the idea of immediate use for a single data-system or group of data-systems, with the assumption that the data-center survives. The columns I write attempt to cover this idea as well as DR.
DR, by contrast, is the assumption that the data-center did not survive, and the data must survive somewhere else to be utilized in the rebuilding efforts.
Both are viable issues in the mainframe world right along side the open systems world. Neither should be ignored, though many corporate budgets don't permit HA to be implemented, as those tend to be much more expensive solutions.
Mike Talon
miketalonnyc@yahoo.com
by littledragon
In my 24+ years in the mainframe world, I have never encountered a mainframe shop that did not have a DR plan in place that was tested at least once a year, usually 2 or more times. However, even in mainframe shops with full DR in place, these sameorganizations often have little or no DR in place for their distributed systems. Not only do mainframe shops have DR plans in place, but these plans are frequently audited both internally and externally (for banks, this is a banking regulatory requirement) and the recovery times and objectives are often reviewed as part of this process. I am not sure what Mr. Talon's exposure to the mainframe world is, but he certainly does not need to preach about DR to us!
And as a previous respondent mentioned, the stability of the platform has absolutely nothing to do with the need for DR. Disasters can strike any data center at any time.
by miketalonnyc
While it is true that most of my hands-on experience has been in the open systems world (and oh boy does it show!), I have dealt with many shops that utilize mainframe systems.
The vast majority of these shops utilize tape backup as the sole DR system for the mainframes. If you have a 72 hour RTO, then tape is great, provided you can find another mainframe system to play back the tapes onto, and you can suffer the loss of up to 24 hours of committed data which never got to the tape systems.
For the majority of mid-tier mainframe businesses, the overwhelming expense of the mainframe systems makes it fiscally impossible for them to own a second system, so DR has often been overlooked, as long as the tapes are being made to meet regulatory requirements.
Now we are faced with pressure to address the fact that entire datacenters can be destroyed, and services that offer shared space on DR mainframes and other cost-effective measures suddenly look very appealing.
You are correct, DR is a cross-platform issue. However I can say that for many companies DR is an issue that has been ignored for far too long, on all the platforms they support.
Mike Talon
miketalonnyc@yahoo.com
by littledragon
When I stated that Mainframe data centers already have DR plans, I was not implying that they just have a document and some backup tapes. I meant that they have full, tested DR capability, including a specific back-up site with appropriate back-up machines (how else could you do a live DR test?). Some very large companies have their own back-up data centers complete with network switching functions, and others have DR contracts with Disaster Recovery service providers who are paid to maintaincompatible mainframe configurations.
So when I say a data center has a DR plan, I mean that they have made full provisions for a backup machine, necessary peripheral equipment (disk, tape, printers) and network access - not that they just back up some data to tape once in a while.
Yes, most mainframe data centers do still use tapes as the backup media. But these tapes go offsite at least once a day, and in many cases, 2 or 3 times a day. And the back-up tapes include system backups,application backups and data backups, including incremental backups on the recent transaction processing files. Service Level recovery times are almost always in the 6 to 12 hour range with minimal or no data loss.
Some of the largest companies are trying to eliminate the tape media problem by using Disk mirroring, but this is still a very expensive option compared to using tape.
My personal opinion: Any busienss that can afford a computer system to support a critical business function cannot afford NOT to have a complete DR plan in place, which includes a back-up site, a back-up machine (or ability to get one quickly), and full systems, applications and data recovery capabilities. Either that, or a really good insurance policy....
Here's a comment from one of our readers:
Most organizations that I work with are planning for DR for their mainframes. They are more concerned about the mainframes than other systems, although UNIX is also considered as having mission critical applications.
I would like to see some more discussion on DR related to organizations and how they are planning on protecting their systems, multi-hop SRDF, XRC, PPRC, etc...
More importantly, I'd like to discuss where are they placing the data centers to facilitate this. All of these technologies require proximal systems to be able to accomplish the very low RPO / RTO that we discuss in e-commerce.
Will we see such discussion in the near future?
by miketalonnyc
I'd say those are great topics for discussion, and I think you'll be seeing some discussion of them right here.
What do you all have to say in response to this posting?
Mike Talon
miketalonnyc@yahoo.com
This discussion is closed to new replies.
Sort by Trending