In The IBM i Trenches With: LaserVault
March 21, 2022 Timothy Prickett Morgan
When somebody says they have been doing the same job for three or four decades, that is a big deal in the 21st century because that kind of long-term employment is just not something anyone counts on. In the OS/400 and IBM i market, such constancy and longevity is, well, normal. Unremarkable. Expected. Good.
So it is with Brad Jensen, the founder and chief executive officer of Electronic Storage Corporation, a company that is perhaps best known for its LaserVault virtual tape library software, but since the company was founded in 1989 in Tulsa, Oklahoma, it has created document creation, distribution, and management software, report generation software, as well as backup, recovery, and virtual tape library software. It had been a while since we shot the breeze with Jensen about the industry, so we picked up the phone and asked what was going on in Jensen’s neck of the IBM i woods.
Timothy Prickett Morgan: I know that Alex and Dan always did a good job covering LaserVault and related products over the years, but you and I have not talked in quite a while and I suspect that there have been some changes to what you are doing. Refresh my memory on what you do, because I am a dataflow engine, not a storage device. . . .
Brad Jensen: Electronic Storage Corporation still makes and sells virtual tape library software to use with IBM BRMS, HelpSystems Robot Save, and so forth. It uses a standalone unit that uses disks and it can emulate up to 64 tape drives, depending on how many cards you put into the system. The concept is simple: IBM i shops shouldn’t need an operator to be finding and loading and unloading tapes, or making sure that data is actually erased on the tapes, and all that. So what we do cuts down on operational complexity and the possibility for error.
The latest thing that we have added to it is our own de-duplication system. A lot of our larger IBM i customers are using de-dupe products from Cohesity and Rubrik and ExaGrid, and all of these do very well in the marketplace and do a lot more than just IBM i. But all three of them need a gateway, they need a way to get stuff from the IBM i, and on tape emulation and tape library emulation is the way to do it.
For example, with BRMS, you can just totally automate everything and all you have to do is check it once in a while make sure it did what it was supposed to do or BRMS will throw an error. That doesn’t happen very often because we perfectly emulate an IBM compatible LTO-4 through LTO-8 tape drive – whatever.
One of the advantages of using a virtual tape library versus actual tape drives and tape media is this: When you use tapes, if you make an LTO-4 tape, you can read it with your LTO-5 generation tape drive, you can’t read it when you move to the LTO-6 generation. So you are either throwing tapes away or constantly having to re-record from one generation of tape drives to another. Because we are tape generation agnostic, we can read tape data formatted for an LTO-3 drive and write it out in LTO-8 format or whatever. I don’t know why LTO drives only go back one generation. . . .
TPM: Well, it is designed to offer some backwards compatibility, but not too much, and to get you to buy new media and go through this hassle as a secondary cost.
Brad Jensen: We do find some companies, because of auditor requirements or other statutory requirements in their industry, still want to be able to write out a physical tape once a month or once a year or what have you for long term storage. And when they have to write out data to a tape, we can use a PC in the middle and write IBM i tapes from it to a physical drive, and you can take that tape when it is done and use it to load data onto an actual IBM i system. At this point, only about one in twenty customers needs to do this.
But the thing that really, really makes the product LaserVault ViTL sell is noy just the virtual tape library capability, but the replication capability built into the software. Because if you use this with a de-dupe system – whether it is ours or from somebody else – then it reduces the network traffic to copy a day or a week worth of data offsite by a factor of 20 to 1. And that makes it just feasible for some of the larger sites to do this, especially if they are trying to replicate over the Internet, in a matter of hours instead of days.
That is basically 99 percent of what we do with VTL right there.
TPM: What do you see going on in the IBM i market now when it comes to storage and backup? If there are 120,000 or so IBM i customers in the world, how many of them have some sort of VTL capability, and how many of them should have it?
Brad Jensen: Most of them don’t have VTL, and pretty much all of them should. I mean, it’s the modern answer to backup storage for the IBM i platform. The nice thing about using the tape channel is that it is reasonably fast. Most of our customers get anywhere from a low of 250 MB/sec to 700 MB/sec and even 800 MB/sec of throughput from the IBM i. One of the things that I would to talk to IBM about, though, is speeding up the tape channel even more. Out VTL can handle throughput of maybe 2 GB/sec to 4 GB/sec – on the order of 5X to 8X more than what we can get out of the IBM i system.
But the storage is slow because of updating objects and this and that. And I have to feel that with some architecture changes we could possibly get past the tape drive speeds just through software changes. It would also make the real LTO tape drives faster, which is a good thing.
TPM: I agree with you. There is no reason not to do this because there is plenty of I/O in a Power8 and Power9 machine and a crazy amount in a Power10 machine.
Brad Jensen: These systems are plenty fast. Some of the object updates and stuff – for example, if they put the object updates into a queue, and then did those after the save was complete, well you would have a shorter backup times, faster throughput, and your data would be safe sooner. The rule is you want to have at least two copies of your backup – and preferably three. And once you get it onto the PC as a set of files, it’s very easy to get just X copy over to another drive or something.
There’s also another little feature my chief programmer figured out, and there is an IBM virtual tape file, and it is not the virtual tape as we are doing it. You can write out stuff to a virtual tape file, but it has to into the Integrated File System. If they could make that re-routable so that it would go out over TCP/IP and Ethernet. We have 10 Gb/sec Ethernet all over the place now and you can even drive it faster than a Fibre Channel drive. We would be a great partner for such an effort because we didn’t grab someone else’s tape emulator code. We write everything, and we run the device driver and everything above it on an X86 server.
TPM: Can you push it to 100 Gb/sec Ethernet?
Brad Jensen: You would have to have 100 Gb/sec Ethernet, of course, but it is easy to do and, besides, everything is going to NVM-Express flash, and if you put four or eight modules on a PCI-Express card, you can drive something like 12 GB/sec, which is equivalent to running at 100 Gb/sec speeds on Ethernet.
TPM: That’s pretty fast.
Brad Jensen: That is pretty fast. We have a few customers with the new system we are working on, and with a 40 TB RAID array and a 20 to 1 compression ratio, that is effectively 800 TB of backup space. That is a lot for most people. But, then again, we used to see customers backing up 100 GB per night, and now we are seeing customers doing 6 TB or 7 TB per night.
** Editor’s Note: To focus on the VTL and backup products, Electronic Storage Corporation has stopped development and sales on its document management products. So if you are looking to buy dome document management tools, Jensen would be happy to talk to you.