programming and human factors
by Jeff Atwood
In my recent Windows Vista performance investigation, I discovered the new ReadyBoost feature. ReadyBoost allows you to augment your PC's performance using a USB flash memory drive. It's very easy to use; just plug in a USB flash drive that's 256 megabytes or larger, then navigate to the ReadyBoost tab on the properties dialog for the drive:
The drive has to meet certain minimum performance characteristics (defined in the ReadyBoost FAQ) to be usable for ReadyBoost. Vista performs a one-time performance benchmark on the drive after it's inserted to determine if the drive is suitable.
But what is ReadyBoost actually doing to improve performance? It's leveraging the unique advantages of flash memory..
.. by caching the system pagefile on that USB flash drive.* Subsequent accesses hit the cached, compressed pagefile on the flash drive and bypass the hard drive entirely.
If we've gone this far, you might wonder why we just don't go all the way and use a giant 32-gigabyte flash drive as our primary hard drive. I can think of three reasons why you wouldn't want to do that:
Also, the performance benefits of a solid state hard drive-- even one based on ultra-fast battery-backed DDR memory-- aren't as amazing as you might think.
That's why the best solution might be a combination of traditional mechanical hard drives and flash memory-- so-called "hybrid" hard drives with embedded flash cache. For example, the Seagate Momentus 5400 PSD includes 256 megabytes of flash RAM. This feature is called ReadyDrive, and it's even better than ReadyBoost. Unlike a USB flash drive, the flash RAM on a hard drive can be read before the system is booted, and thus can be used to speed up boot and resume times, too.
It's looking more and more like flash memory is the future. But be careful, because not all flash memory is created equal. I researched USB flash drive performance recently and I found benchmark roundups at hardware secret, AnandTech, and Ars Technica. In my research, I found that there are at least three distinct tiers of flash drive performance today: mediocre, good, and best. The price difference between the best performers and the worst performers isn't much, so you might as well buy the fast ones. The flash drives that performed the best in the above three benchmarks were the Kingston Data Traveller Elite and the Lexar JumpDrive Lightning.
Cheap flash drives are cheap for a reason-- they skimp on performance. Here's performance comparison of three USB thumb drives I had on hand: a 1 gigabyte Iomega Micro Mini, a 1 gigabyte Kingston Data Traveler Elite, and a generic no-name 128 megabyte model I got at a trade show.
I ran SiSoft Sandra's flash memory test on these three drives. The results are summarized below. Note that the bars are stacked, so the total transfer rate is only as high as the largest sub-color in the bar.
There's a big disparity between read and write performance on flash drives. And small files are disproportionately painful to transfer through these devices. The cheaper the flash drive, the worse these characteristics will be. When you go for an inexpensive USB flash drive, that's the tradeoff you're making.
I also ran the command line chddspeed utility on these three drives. Here are the results for the random access read test.
Flash memory is exceptionally strong at random access; my fast WD Raptor drive can't touch these scores.
Here are the chddspeed results for sequential access.
Up to 12 Mb/sec is nothing to sneeze at, but it's nearly 6 times slower than the 68 Mb/sec the Raptor achieves. If you need fast sequential read (or write) speeds, you want a hard drive.
After all this analysis, it's clear to me that traditional hard drives and flash memory are quite complimentary; they're strong in different areas. But flash drives are the future. They will definitely replace hard drives in almost all low end and low power devices-- and future high performance hard drives will need to have a substantial chunk of flash memory on board to stay competitive.
* Yes, it's encrypted, and yes, it is optimized for the limited duty cycle of flash drives. It's even compressed, so that 1 GB flash drive is effectively 2 GB of cache. This is all covered in the excellent ReadyBoost FAQ.
Here are some actual perf numbers from ReadyBoost (USB flash cache) and ReadyDrive (HDD flash cache) from TechEd 2006:
http://www.tgdaily.com/2006/06/15/teched2006_fresh_demos_of_superfetch/
--
Ayers also touted ReadyDrive for being "your notebook's best friend." Powering down the hard drive rotor, he said, will noticeably improve battery life by as much as 15 to 30 minutes on an ordinary notebook computer (like his). He showed a video of a test made at Microsoft where a hybrid hard drive, on a system running Word, was fully accessible to the operating system 100% of the time, while the rotor was powered on for less than a third of that time.
--
In Wednesday's ReadyBoost/SuperFetch demo (pictures of which can be seen in our accompanying image gallery), a battery of four applications were loaded in sequence: Outlook first, followed by OneNote, PowerPoint, and Adobe Acrobat. On an ordinary HP notebook computer, with a 2 GHz processor and 512 MB of RAM, the sequence of apps took 32.6 seconds of load time. With ReadyBoost turned on, load time immediately improved to 26.5 seconds - a boost of 18.7%.
Frankly, that's not too impressive. But as Matt Ayers explained, SuperFetch is learning from its user about the patterns of hard disk pages it tends to load. Having seen the same sequence of data loaded into memory the same way twice before, SuperFetch then has reason to believe this may be a common occurrence for this system - for instance, a startup sequence. [And thus pre-emptively loads these memory pages into the USB flash ReadyBoost cache.] So for the third test, with exactly the same four apps, the load time was enhanced by about 206% over cold loading.
--
Jeff, I've got a page file set up as well with my 2 GB RAM but... it's never being used, except for the handful of pages that the OS swaps out by default. Are you really experiencing disk thrashing with 2 GB?
Chris Nahr on September 14, 2006 3:36 AMSo what happens if the drive is pulled during operation?
Alex on September 14, 2006 3:54 AMAre you really experiencing disk thrashing with 2 GB
When I'm demoing Team System, I use..
- a 1 gigabyte Windows Server 2003 image, running SQL Server 2005, Sharepoint Services, and Team Foundation Server.
- a 512 megabyte Windows XP image, running Visual Studio Team Suite
.. all on a laptop with 2GB RAM.
But even in more typical use, the page file on disk DOES get used. Disabling the page file has some bizarre consequences that aren't immediately obvious. But try it yourself; some people like it:
http://www.codinghorror.com/blog/archives/000422.html
I expect with 4gb of RAM, you probably wouldn't need to worry. Even if you had 32gb of RAM (I wish) your system will still resume and suspend faster with embedded flash cache, though.
Jeff Atwood on September 14, 2006 3:59 AMAlex, your question is answered in the FAQ:
http://blogs.msdn.com/tomarcher/archive/2006/06/02/615199.aspx
I encourage anyone who is even vaguely interested in this feature to read the FAQ; it's really quite good.
Jeff Atwood on September 14, 2006 4:02 AMOfftopic (sorry)...
Looking at the title bar in your screenshot, it looks like you are using graphite colour scheme (and maybe transparency off). I'm using the same, is it me or is the title bar text really difficult to read in Aero? The "Outerglow" around the text looks like a smudge. I've even tried to make the text white and remove the smudge to no joy.
Jeff, would like your opinions on Vista UI in a post...
Damian on September 14, 2006 6:18 AMThis seems silly. If I'm accessing my pagefile often enough that a flash cache of it will help system performance then I should either: 1) add more real memory so my working set doesn't get paged, or 2) reduce the number of running apps (or their collective memory size) such that I'm not paging so frequently.
ReadyDrive sounds interesting, though (if nothing else because I don't have to have a flash drive hanging off the side of my laptop). But the technology isn't specific to Vista.
That's true, adding more memory is better than using ReadyBoost. But that's not always the easiest or most cost-effective option.
Besides the obvious 2GB max on laptops, there's also users who aren't comfortable opening up their desktops. Non-power users can add a ReadyBoost drive in a couple minutes, as opposed to the hassle of trying to figure out exactly what memory slots are available, how many pins the slots have, which are paired, and then finally what to buy that will work well with the existing memory. For most people, buying memory and installing it is not easy. It's far easier to grab a spare 2GB USB drive out of the drawer...
Mark on September 14, 2006 7:14 AMI was at Lexar and M-Systems for a number of years and now consult to the flash memory industry. Conversations I've had with my contacts at Intel lead me to believe that ReadyDrive is a quick fix until "Robson" flash appears on motherboards next year.
The performance and life issue that Jeff points out are generally set by 1) the type of NAND flash, and 2) the flash controller chip that the flash drive manufacturer uses.
SLC - Single Level Cell, which is manufactured by Samsung, Hynix, Micron and Intel has a life span of 1 million program erase cycles (you should live so long). MLC - Multi Level Cell, is manufactured by the above companies, but the leading provider of MLC components is Toshiba. MLC is usuually rated only 10,000 program erase cycles, which makes it much better suited for your iPod Nano than your HDD replacement.
Silicon Motion (www.siliconmotion.com.tw) supplies the fastest flash controller chip in the market. They can offer sustained write speeds of 30 MB/sec.
What is inside that flash drive you just bought - no telling without cracking it open. Memorex, Verbatim, Kingston, Lexar, SanDisk, etc., all have the identification strings changed at the chip fabs so that any pinging of the drive returns only the brand info.
I was at the Denali Memcon in San Jose yesterday. Its a gathering of memory guys with nothing better to do. Samsung demonstrated side by side two Dell notebooks. One running a 32 GB SSD, and the other containing the standard Dell 40 GB HDD. The SSD version booted 40% quicker. The point is that notebook HDD's are much slower than the 200 GB that you have in your desktop.
bob witkow on September 14, 2006 7:29 AM
what about using some sort of software Ram drive? With 4GB+ or System ram, I would not miss a gig gone to the pagefile...
http://home.scarlet.be/~tsd59392/RAMDisk/ramdriv.htm
Good stuff, thanks for sharing.
Not that it's relevant, but I'm not so sure about flash drives and hard drives saying nice things about each other, or maybe about the user. I'm guessing you meant "complementary". (Sorry, my parents were teachers -- it's an OCB).
Pops on September 14, 2006 8:58 AMWith 4GB+ of system RAM you probably don't need a pagefile, as the page file is used to augment system RAM. -.-
Jim on September 14, 2006 9:29 AMThat's true, adding more memory is better than using ReadyBoost.
Yes, adding more system memory is always a good thing. But unless you're running a 64-bit OS, your apps can't use 4gb all that effectively.
And as Mark pointed out, adding a flash drive is often the path of least resistance.
Why not do both? Why does it have to be one or the other?
With 4GB+ of system RAM you probably don't need a pagefile
Possibly, and maybe only under x64. I tried running without a pagefile using *2gb* of system memory and I had enough weirdnesses that I eventually was forced to turn the pagefile back on:
http://www.codinghorror.com/blog/archives/000422.html
Jeff Atwood on September 14, 2006 10:30 AMConversations I've had with my contacts at Intel lead me to believe that ReadyDrive is a quick fix until "Robson" flash appears on motherboards next year.
Hi Bob, thanks for posting the great info. I tend to agree that we need this high-speed flash memory somewhere other than a USB port. Intel's approach, putting flash memory on the motherboard, is definitely the most flexible.. but it's also a whole-system upgrade. Buying a hybridized flash drive, or a fast USB 2.0 flash drive, are much easier upgrade paths for better performance.
The SSD version booted 40% quicker
I'm sure, but I bet a lot of that perf was due to the fast random access. The flash hard drive tradeoffs do make sense for portable devices, but I think you could realize similar boot time benefits (with higher performance and a bit higher power draw) using a hybridized flash/physical drive.
Jeff Atwood on September 14, 2006 11:42 AMSo ReadyBoost only caches the pagefile? That doesn't sound tremendously useful, especially as we move to 64-bit machines that can use more than 4 GB RAM. If you're accessing your pagefile AT ALL during normal operation you should install more main memory!
ReadyDrive might be good for notebooks but let's not forget that Vista is reported to be extremely power-hungry. Didn't Beta 2 drain a fully charged notebook in like half an hour? I suppose they need ReadyDrive to get any respectable mileage out of the notebook battery...
Chris Nahr on September 14, 2006 1:34 PMSpeed: ReadyBoost only caches small random reads to the pagefile. Under these conditions, flash memory outperforms hard disk drives. For large sequential reads, hard drives will always win. ReadyBoost is sensitive to random read performance of your flash drive. But there is no need to benchmark, just plug it in, Vista does the benchmarking. YOu will find a detailed entry in the logs. In fact, Vista's test may be better than the tools you used. I.e., it tells you that your 2GB flash drive has 128MB fast flash, and the rest is slow.
Wear: flash memory lasts far longer than 100000 write cycles. Going by vendor specifications, a flash drive used by ReadyBoost would last between 8-80 years of 24x7 ReadyBoost use. So I guess it's not an issue.
Unfortunately you can have just one ReadyBoost file in the system right now, and it cannot exceed 4GB, no matter how big your flash drive is.
Alexander Holy on September 14, 2006 1:53 PMTwo questions:
1) Has their been any talk of using the ReadyBoost usb drive as a Sleep/Hibernate storage device? If you have it attached already, why not just flash the memory to the drive, and power the whole system down?
2) Are any manufacturers planning on making built in flash caches for new laptops, or purpose-built PCMCIA or mini-PCI cards? This seems like the perfect way to bring a borderline laptop into full Vista compliance.
Jeff Jacobs on September 15, 2006 1:45 PMJeff, Since I'm rapidly acquiring a drawer-full of flash drives that Vista tells me are too slow to work with ReadyBoost, I'm curious as to whether you actually got either of your 1gb drives to work with it? (I probably should have asked before I went ahead and ordered the Kingston Data Traveler Elite.)
Richard on September 17, 2006 10:27 AMRichard, most of the flash drives I have on hand actually worked as ReadyBoost in Vista, even the slow-ish ones. Maybe I just got lucky.
But the Kingston Data Traveler Elite *definitely* works, I can personally vouch for that.
Jeff Atwood on September 18, 2006 5:32 AMThere is an windows service in vista called Software Licensing. Description: Enables the download, installation and enforcement of digital licenses for Windows and Windows Applications.
I didn't like the sound of that so I tried to stop disable it. Then got a dialog saying that ReadyBoost service depends on it!
Wtf?
Damian Hickey on September 19, 2006 3:23 AMFunny you post this article today. Just last night I bought my first memory stick where I assumed the differences in speed couldn't be that huge, well they are! I think the stick I purchased has a transfer rate of less than 1Mb/sec if that's possible...
Anyways, great article. I'm glad you pointed out the differences. Being in the industry so long, it's easy to forget some of these things.
Stephane Grenier on September 19, 2006 9:45 AMIn the same order of idea, I've just found out about a pretty interesting product called a href="http://www.mojopac.com/"MojoPaca
Jerome on September 25, 2006 11:37 AMIf it's useful to anyone, I've been building a list of ReadyBoost compatible devices on my site at:
http://www.grantgibson.co.uk/misc/readyboost/
Cheers,
Grant.
XP just wasn't designed to be used without a page file, with 4GB of memory and several GB free, programs sometimes run out of memory and the OS acts weird. When enabled and over 2GB of free memory the OS still dumps stuff to the pagefile. I hope in vista you can really disable the stupid thing.
James on November 7, 2006 8:34 AMTwo things: MojoPac was mentioned above...has anyone actually gotten it to work? I bought it, and it would not start even once. Just says it failed. I contacted support, with no reply at all.
Anyway, let's say I have a Windows Vista 64-bit installation and 4 GB RAM system. Will there be any benefit for me using the ReadyBoost? Even if it is a little or only under certain conditions, I am interested to understand the specifics.
Thanks much.
Chris Hoffman on December 7, 2006 12:28 PMGood article on ReadyBoost performance at Tom's Hardware:
http://www.tomshardware.com/2007/01/31/windows-vista-superfetch-and-readyboostanalyzed/page3.html
Summary: big performance boost for 512mb RAM machines, minor bump on 1gb RAM and 2gb RAM configurations.
Jeff Atwood on February 5, 2007 10:21 AMJeff,
USB Flash Drives and Flash memory cards may not be as fast as hard drives, but the new generation of SSDs are currently roasting hard drives alive.
A-data has a 128GB SSD that has a sustained read of 65MB/sec. Due to the incredibly low latency, they can handle many, many times the I/O of a hard drive.
Slow, 2-3MB/sec Flash Media used for ReadyBoost is calculated to be 10x faster at I/O for smaller files. You can imagine how much faster an A-Data SSD would be than even the fastest SCSI hard drive overall.
A top of the line SCSI drive may see 80MB/sec on the outside of the platter, but the latency is 10x greater or more than an SSD.
Jace on February 6, 2007 7:56 AMHeya Jeff!
I've just noticed that over at Tech Crater, they've hosted a list of the actual speed rating numbers for many flash devices. This way everyone can get the best performance boost out of ReadyBoost. You can see it at http://www.techcrater.com/2007/04/05/readyboost-flash-memory-speed-list/.
Cheers,
rio
We supply 100% brand new replacement batteries that are manufactured to meet the specifications of the original brand manufacturer. If you can not find the notebook battery please contact with us immediately. We will give you a reply within 24 hours.
batteries on April 23, 2008 1:21 PMWe supply a large range of replacement Laptop Battery. Our products are reasonable in price and fine quality such as Acer, Sony.
battery on April 23, 2008 1:22 PMWe warrant that the products sold in our website, other than explicitly stated, are 100% brand new and free from defects in material and workmanship.
laptop battery on April 23, 2008 1:23 PMWe have huge inventory of laptop batteries for Sony, IBM, HP, Compaq, Apple, Dell, Gateway, Panasonic and more.
Jeff, Most if not all of what I read in the article and comments section were correct from what I have found so far in the HDD, HHD and SSD markets. Yes smaller devices like notebooks and tablets will defiantly benefit from the HHD's and SSD's alike. HOWEVER, the limited read/write times of SSD's puts a cramp in me actually using them currently, well that and the price per gig is still way too high. Perhaps when they offer "instant on" like my old HP palmtop does when you power it up AND it will reboot back into an original state when both batteries are removed. Its slow, old and cannot be upgraded but its still useful. The ASUS Eee 900 has alot of those features with a modern Linux OS. I'm not a "power-user" "programmer" or "coder" but I know what I want my machine to do and what sort of environment it will be expected to perform in (high heat and high vibration) so I choose products to fit that need best. Keep up the good work and keep the rest of us informed as new products are released.
Bounty on May 1, 2008 2:06 AMI could be wrong about this, but won't your PC crash horribly if you (accidentally) unplug you flash drive while using it as a readyboost device? If so, no thanks.
Paul Sanders
www.alpinesoft.co.uk
I could be wrong about this, but won't your PC crash horribly if you (accidentally) unplug you flash drive while using it as a readyboost device? If so, no thanks.
Paul Sanders
http://www.alpinesoft.co.uk
I have a similar setup, WD Raptor 74gig (primary), but have recently purchased Hitatchi P7K500 500 gig 7200rpm with 16 meg cache. I got a surprise when I ran the test against it.
Just like your Raptor (mine is WD740GD-00FLC0) I got similar results in performance numbers. For the a Multisector read test line, one sample for here:
(HD) (blocks) (Repititions) (time) (Speed mb/s)
WD740GD-00FLC0: 147 4000 4187 68.571620
P7K500: 147 4000 3187 90.087661
Hmmmm, the 90 and the 68 where times for both drives that were representational of their top speed; hitting it multiple times for different blocks consistently!
Here are the differences:
1) Hitatchi was formated with 8K blocks, where Raptor took the default.
2) Hitachi has an 16meg cache where the raptor only has 8mb.
3) Raptor is running the system OS (XP)...
Motherboard (P4P800-E Deluxe) only to my knowledge supports SATA 1 (150).
Either the cache is coming into play heavily, or the program has a quirk which formatting 8kb sectors comes into play boosting the Hitachi....
OmegaMan on August 30, 2008 2:29 AMRunning a 4GB Desktop at home with an 8GB Paging File. My 2GB Flash Drive does make a difference in normal applications speed and response times.
I recently came across your blog and have been reading along. I thought I would leave my first comment. I don't know what to say except that I have enjoyed reading. Nice blog. I will keep visiting this blog very often.
Ann
Ann on March 23, 2009 11:02 AMFLASH is a half-way house to what we all really want which is something that behaves exactly like SRAM (none of those nasty DRAM refresh cycles) yet retains its state in the absence of power.
It used to be caleed CORE memory and it is on its way back if this (http://www.freescale.com/webapp/sps/site/overview.jsp?nodeId=0ST287482180C6B6253) article is anything to go by. Yes the densitiy is nowhere near high enough but it is a sign of things to come.
If MRAM takes off, FLASH will become an interesting footnote in the textbooks.
Links:
http://www.freescale.com/webapp/sps/site/overview.jsp?nodeId=0ST287482180C6B6253
http://rfdesign.com/next_generation_wireless/freescale_mram_memory/
Are there any real advantages to using a hardware solution for PC Remote Access over a software solution? I tend to think the software is easier to update and upgrade, but I really don't know any advantages of the hardware end.
its great laptop inside shame about the small dell keyboard, i shud no i own one .thank you.
Well, MLC is usuually rated only 10,000 program erase cycles, which makes it much better suited for your iPod Nano than your HDD replacement.
Singles
this is indeed a nice blog you have..i am sure that they will learn lots of things about this blog..thank you for the information...
blackberry spy software
cell phone tapping
cell phone spy software
cell phone spying
free cell phone spy
cell phone monitoring
It is a nice blog.Here in this blog the author tells something about vista and flash drivers.
Dannie Kosanke on July 26, 2011 3:23 AMThe comments to this entry are closed.
Subscribe in a reader
Subscribe via email
|
|
Traffic Stats |