RAM was the new disk - and it will be again

By Sean McGrath  Add a new comment

Way back in time when a top of the range PC had two 10 megabyte hard disks and 640k was a lot of RAM, I used to use RAM as disk. My trusty old MS-DOS 2 machine provided a device driver RAMDRIVE.SYS which, when loaded, took a lump of RAM and made drive F out of it. It was terribly, terribly useful. Back then I wrote a lot of software in Turbo Pascal which used a technique called overlays to break through the 64k barrier. (Yes, you read that correctly, I did say the 64k barrier.)

Somewhere betwixt there and here, RAM disks have fallen by the wayside. Perhaps modern operating systems are so super smart that they handle it all transparently. I don't know to be honest. It sure doesn't feel like it when I'm sitting at my laptop, watching the hard disk light flash. And that is on a machine with 2 GIG of RAM for goodness sake.

Anyway, the great thing about putting Turbo Pascal overlay files onto the RAM disk was that they were code - not data. If the machine had a power failure and you lost the contents of the RAM disk, so what? It was just code. Thinking back to those days got me thinking about an aggressive RAM caching strategy that I do not think is being automatically exploited by modern operating systems. Namely, laptops.

Here is what I am thinking. The problem with using RAM for persistence is that you loose it when the power goes down. You can only afford to aggressively use RAM for persistence caching if you have a very reliable machine and very reliable power. Cue the pictures of rack mounted, air conditioned servers with uninterruptable power supplies.

Ok. But hang on a second. My day-to-day machine is a laptop. It is very reliable and guess what? It has a built in uninterruptable power supply. It is called a battery pack. I spend most of my working day tethered to the mains with the battery pack doing nothing. Even with a degrading battery such as mine, I can be confident of 30 minutes power before the battery dies. More than enough time to flush the cache and do an orderly shutdown.

I know I can probably rig it all up under Linux using tmpfs but I don't know what is involved in automatic shutdown which would involve probing the state of mains supply to the battery. I guess I'm just suprised that the RAM disk trick is not more commonly exploited in laptop-tethered-to-mains environments.

We are at an interesting point in personal computing. Most of us have more storage space and more raw CPU power than NASA had in its entirety a couple of decades ago. Sadly however, we cannot effectively utilize a lot of the storage because it so darned slow to access. Similarly our honkin' great CPUs spend there time running around NOP loops waiting for disk - and also RAM - to catch up.

We need all the help we can get to keep our CPUs doing work rather than idling. RAM disks have a role to play in that - and not just in server farms.

ITworld LIVE

Operating SystemsWhite Papers & Webcasts

White Paper

A Comparison of PowerVM and VMware vSphere (4.1 & 5.0) Virtualization Performance

This technical white paper presents benchmark results showing greater VM consolidation ratios than demonstrated in previous benchmarks and demonstrating the extent of the performance lead that PowerVM virtualization technologies deliver over x86-based add-on virtualization products.

White Paper

Consolidating Lotus Domino x86 Workloads on IBM Power Systems

Read the white paper to learn how moving up to Lotus Domino 8.5 and consolidating with IBM Power Servers can help you boost performance results and ROI.

White Paper

Task, workflow & issue management for teams. Try free!

Need a flexible system for managing team tasks, issue tracking, and automating and managing workflow processes? Comindware® Tracker helps you do it all.

Webcast On Demand

Best Practices in Monitoring VMware

The benefits of virtualization are unassailable: increased agility, scale, and cost savings to name a few. However, so too are the monitoring challenges posed by these environments-including complexities, lack of visibility and control, and inefficiency.

Sponsor: Nimsoft

White Paper

How Nimsoft Service Desk Speeds Deployment and Time to Value

For years, many support teams have been hamstrung by their traditional service desk platforms, which require complex, time-consuming coding for virtually every aspect of customization. This complexity makes it costly and difficult for support organizations to adapt-and places an increasingly substantial burden on the agility and efficiency of the business as a whole.

See more White Papers | Webcasts

Ask a question

Ask a Question