Archive for March, 2009

Solid State Drives

Friday, March 20th, 2009

The solid state drive (SSD) is all the rage. It is an outgrowth of the USB thumb drive that is bigger, has a direct access storage device access (rather than serial as with USB) but is still composed of electronic memory that can hold its state when the power is turned off.

The expectation is that the SSD would be very fast because there are no moving parts and it is all electronic. Anand Lal Shimpi finds that the technology isn’t all together yet in his analysis of the current state of SSD in his report The SSD Anthology: Understanding SSDs and New Drives from OCZ.

The problem has to do with how the memory is addressed and what needs to be done to re-use it. An SSD, like a hard drive, deals with storage in chunks. Unlike a hard drive, the SSD can’t just write over existing data with new data. It has to ‘erase’ the area first. The SSD also needs to abstract the actual location from its address in order to even out memory cell exercise for best life. What all this boils down to is that an attempt to write to the SSD can hiccup if the drive is running out of untrod space to put things. That can create a latency problem and that is what Anand was able to measure and compare in his tests of SSD’s.

This problem is aggravated because it is the OS that knows when a file has been deleted and not the media. That means the media cannot anticipate re-use of vacated addresses and can only prepare areas for new data when it gets a request from the OS. If it can’t temporarily cache the OS write request, it will hold up the show for a bit while it gets the prep done. That may result in delays that can add up and make the user wonder what is going on.

What Anand shows in his measurements is that the SSD can indeed enhance the user experience. A boot process can often be shortened by an order of magnitude as the system reads all of the code from storage needed to run all of its processes and startup programs. He also shows how manufacturers are struggling to understand the technology and its use properly as well. His encounter with OCZ is an example of an SSD manufacturer that took heed of Anand’s findings and took steps to improve their product.

It may yet take a while for the SSD technology to get off the leading edge: to get the capabilities tailored to needs and the costs down to mass market levels. The promise is exciting.

Ubuntu repositories and public keys

Sunday, March 15th, 2009

Extra Repositories for Ubuntu 8.10 You Might Want at I’ Been to Ubuntu describes some of the repositories that can be used in Ubuntu to facilitate installing Acrobat, Google Earth, or the latest versions of Open Office or Wine.

One issue that comes up is security. You will get warnings about these ‘unregistered’ repositories until you provide the key to Ubuntu.

keycode
where keycode is that given in the warning

What happened to CUPS (server.crt symlink)

Friday, March 13th, 2009

When reinstalling Ubuntu 8.10 from a remastersys iso (put on a USB drive with netbootn), CUPS would not start due to complaints about a bad symlink. omns offered fix at crunchbanglinuxforums by editing the startup script. EYo offered a more elegant fix to create the missing ssl links.

# cd /etc/ssl/certs/
# make-ssl-cert generate-default-snakeoil

PiotrAF also suggests “sudo apt-get install –reinstall ssl-cert” to tackle this from the other end.

that seems to work. it will probably be fixed in the install packages soon (if not already). It does raise the question about what these packages are doing …

Remote login, Q&D substitute for VNC on a LAN

Thursday, March 12th, 2009

In Ubuntu, you enable remote logins on a machine with System->Administration->Login Window in the remote tab to enable a server. Then all you need to do is hit the button at the lower left of the client login screen to select the login to a remote computer option.

Juliet Kemp describes this in Remote Graphical Desktop With GDM and KDM, Graphical Remote Control. The process involves setting an xdmcp enabled line in the GDM or KDMrc configuration files and then using “Xnest -query remotemachine.com -geometry 1280×1024 :1″ to launch the remote login. Note that this command sets up a second X windows session on your computer as the :0 is your primary desktop. Then you can use the CTRL ALT F7 or F8 to switch between windows.

The big unknown is the host name or address of the server machine (remotemachine.com in the Xnest code example). When using the GDM client login, you can select from a displayed list of available servers. From the command line, you have to know it (and to spell it correctly!). That’s the dif between the GUI and command line approaches in general. There must be a utility to list available Xdmcp hosts – now to find it …

CFL explained

Wednesday, March 11th, 2009

Design Line has a good rundown on How compact fluorescent lamps work–and how to dim them. Circuit diagrams, wave forms, and other stuff – good stuff.

Throughput bottlenecks

Tuesday, March 10th, 2009

In looking at storage, capacity is not the only issue. How fast you can get data from storage to a destination is also an issue. This path from storage to destination often travels several roads and each has its own limitations.

The Wikipedia SATA page has a nice comparison chart for hard drive buses. Curent drives can transfer data at a maximum of about 120 MB/s which is covered by the standard SATA or PATA buses. The Firewire and USB ports often found on current motherboards can handle data at about half the hard drive speed or at about 50 to 60 MB/s. The device bandwidth page notes that the old 300 baud modems that could handle 30 B/s were at human reading speed. That means current buses operate at about a million times human reading speed.

A storage device might also need to traverse a local area network. Most common is Fast Ethernet which can handle about 10 MB/s which slows down to about a third that in practical situations. If Wireless is involved, the speed drops to below a quarter MB/s with the latest and greatest standards maybe hitting 10 MB/s or so.

For comparison, the standard bus in the computer, the PCI bus, can handle 130 MB/s and the new version, PCI express, can handle 250 MB/s for each of up to 16 channels. The memory buses tend to run from about 1000 MB/s (5 year old PC133) to 6,400 MB/s (PC2-6400) in modern mid level PC’s. Digital audio runs about a third of a MB/s and the latest HDMI video with 1080p runs about 1,275 MB/s.

But, back to loading up a storage device. The two buses usually used are USB and fast Ethernet. USB looks to be about 5 times as fast. For time, let’s consider 100 GB and the slower ‘fast Ethernet’ in practical terms. Netgear gets actual speeds of about 20 MB/s in tests for its NAS devices, which are rated as speedier than most, using gigabit Ethernet. Fast Ethernet runs at a tenth of that speed so consider about 3 MB/s as a good rate for calculation. 100 GB is about 1E5 M bytes divided by 3 MB/s yields 5E4 seconds, which is about a full day.

The bottom line

If you are looking at loading up an external storage device with the typical wires between boxes approach, plan on it taking a while. A 500 GB external drive can take nearly a day to fill over a LAN or USB link going full tilt. Even drive to drive inside the box you are still talking about tens of minutes. Modern storage devices hold a lot of stuff!

Fax trends and implications

Friday, March 6th, 2009

Ten years ago, the office fax was a critical part of business. Back then, the transition was from stand-alone analog fax machines to scanners and modems connected to a computer. Even a small office (like ours) could integrate a database to a fax program and send out fax blasts for clients. Of course, the ease of being able to send out facsimile ‘en masse’ meant that abuses occurred and that stimulated legislation and regulation to inhibit unsolicited faxing.

Nowadays, it seems e-mail attachments are taking over a lot of the fax volume. Fax is still needed, though. Part of this is due to how the law treats facsimile documents and how it defines them. PayPal is a case in point. They got into a banking ruckus about the identity of clients and that meant they needed to verify the status of those clients representing themselves as, for instance, nonprofit associations. One of the ways to satisfy PayPal’s needs was to get a fax cover page from them with case identifying bar codes. Fill in the blanks on the cover page, attach the necessary documentation, and fax it back so they then had proper documentation on file to satisfy regulatory requirements.

Since the very nice PMFax on OS/2 became a bit long in the tooth, I have been using Hylafax on an Ubuntu server with a couple of US Robotics v.everything modems. I have also installed it on the eBox machine. Hylafax, like eBox, is a bit more potent than needed for a SOHO environment. That means that setup and management is a bit of a pain.

eBox does not yet support fax services. It is also not a feature found in network appliances like the ReadyNAS that provides for many other protocols and for printer sharing. That is one indication that fax service is a rather low priority. A fax server such as Hylafax can be installed alongside eBox without encountering conflict problems. It also seems possible to install on a well endowed NAS as those have USB ports (modern modems are even powered via USB), print sharing, media streaming service, and user access controls already. A fax server doesn’t require that much processing power so it shouldn’t overload the computing capabilities of the device.

An alternative I am testing is an all-in-one device with print, fax, and scan. The HP M1522nf was on sale with a good rebate a while back (thanks slickdeals.net). The biggest problem with this it is limited in its Linux software and doesn’t cache incoming faxes in a way you can avoid printing them but just download from the machine’s web server. It does have spam fax blocking so you can reject faxes with selected fax ID’s (which are required by regulation but sometimes ignored).

With an automatic fax detector on your line (see Command Communications for example), you can filter out junk faxes and often eliminate the need to print any fax you receive. The fax can contribute directly to your data store. With the proper setup, it can be a convenient means to communicate. It was ten years ago but is a bit more of a pain now.

These parts and pieces add up. Simplifying the hardware would be nice and it would also reduce the residual power drains. That might only be a few dollars but it does add up.

Storage media history

Thursday, March 5th, 2009

Computer Data Storage Through the Ages — From Punch Cards to Blu-Ray (print page) has a pictorial list of the most popular data storage devices. Some of them (those I still have evidence for in the closet!) include

IBM Punch Card, 1725 – 1975, 960 bits (80 characters or bytes) – usually used as card decks that sometimes were several feet thick. Originally designed for the US census, they could be manually manipulated for statical evaluations.

IBM Magnetic Tape, 1951 to present – the half inch circa 70′s mainframe tape of 2400 to 4800 feet could store 100kB – the mainstay for mainframe archival purposes.

Audio Cassette Tape, early 70′s to late 80′s, 700kB per side on a 90 minute cassette – essentially like using a modem except to tape rather than phone line.

5.25″ floppy, 1976 – 1982, 1.2 MB in high density, 360k was common circa 1980 for microcomputers

3.5″ floppy, 1982- , 1.44MB although 720k single density were common in the early years.

CD, 1980 – , 700 MB.

Colorado Backup (tape), early 90′s – 2001, 14 GB in later versions, 100 MB early.

CompactFlash, 1994 – , now up to 64 GB. This has branched out to SD memory and other card types used in portable devices and also the USB flash drives. Modern card readers often talk about being able to read 20 or more interface types for these devices.

Zip Drive, 1994 – 2003, 750 MB

DVD, 1995 – , 4.7 GB or 8.5 GB dual layer – this is an upgrade to the CD and is being upgraded by HD-DVD and Blue Ray starting about 2006. All use optical technologies on similar disk media.

So the capacity increases, the space requirements decrease, the cost of media decreases, and the more wide use in the consumer market brings the media drives down in price. For instance, both floppies and CD drives started out about about $400 and ended at a tenth that price. Inflation for the 20 years between the peak use times for these drives would make the comparison even more interesting.

About backups and media requirements

Tuesday, March 3rd, 2009

Back in the 80′s, the start of the micro-computer and home computing era, a backup could be done with floppy diskettes. PC Tools Backup would fit on one floppy and a complete system backup would fit on maybe ten or twenty. This was when the operating system was only a few hundred KB.

In the 90′s, the media moved to tape. First was the floppy interface tape drive that used tape cartridges with a capacity of 100 MB or so. A bit later came SCSI based tape drives that could handle 2 GB. Operating systems such as Windows 95 or OS/2 came on floppy sets. The CD ROM as a distribution and backup media became common in this decade as its 700 MB capacity was plenty for applications and software and sufficient for many backup needs. At the tail end of the 90′s, DVD’s became available. These were like super CD’s as they could hold more than five times as much data as a CD.

Floppies were cheap and didn’t require any extra hardware. Tapes were often an order of magnitude more expensive than floppies and required rather expensive tape drives. CD’s and DVD’s got back to fairly inexpensive media and their broad appeal brought the cost of the drives to handle them down to only moderately expensive.

Compression was a big deal. Most of the files subject to backup were text or simple memory representations that could be easily compressed down to somewhere between a half to a tenth of their original size in a backup. The files also tended to be rather small with sizes measured in kilobytes.

The pictorial presentation both on screen and in data is now the norm. In documents, fonts, illustrations, and layout descriptions mean that the word processing files are sized in megabytes rather than kilobytes. Images have increased in resolution and color depth. Audio and video both require significant data to represent anything other than a trivial segment.

One trend has been to compress data for storage in files. Word processing files used by Open Office, for example, contain the many components of the document compressed and collected as one file. A library of routines for software is similarly composed. Video and audio files often used sophisticated compression schemes that utilize a knowledge of human perception. What this means is that further compression to reduce backup media requirements doesn’t pay like it used to. Instead of being able to expect a backup file store to be less than half the size of the original files, you are lucky to get just a ten percent reduction in size and have to do an awful lot of computing to get even that.

The situation now is that an operating system needs several GB of backup media and many PC users have hundreds of GB of files that need to backed up. Even with DVD’s that gets to be a problem. Fortunately, hard drive storage technology has made it possible to provide the media space for modern backups at a reasonable cost. The interface technologies such as USB have also improved to make it easy to temporarily attach a storage device to a PC and copy over large amounts of data rather quickly.

Technology is also making other backup ideas possible. Broadband I’net is bringing up the idea of commercial backup servers so your data is backed up somewhere in the ‘cloud’ of internet space. Redundant drive array technology is reducing the risk of drive failures.

Now, what do you do when you don’t need that data anymore?