SSDs - State of the Product?

[Santilli]

I know, but, it's a good article...

ATTO:

 

[LunarMist]

Here's my boot drive, not an SSD, but not bad...

Why is the transfer graph relatively flat? Is that some sort of array, old SCSI drives on saturated controller? It does not look to be State of the Product.

 

[Santilli]

Very funny.
Anyone have max transfer rates on that 'old controller'?

I've found two drives is good, more is point of diminishing returns in Raid O, with Scsi 15k drives.

I could be wrong. Could be my old 320 Megaraid 1 controller can't get over 120 mb/sec....

 

[ddrueding]

Someone did a review of 4 of the X-25E in RAID-0 on a high-end controller, and it actually showed performance loss in low-load scenarios. Otherwise I would be all over it. I may need to do it anyway, I only have 10GB free of the 32GB.

 

[LunarMist]

OK, OK, I give. I may just get a 64GB X25-E and use it independently. I don't know what to connect it to though. There are no free ports left.

 

[LunarMist]

Very funny.
Anyone have max transfer rates on that 'old controller'?

I've found two drives is good, more is point of diminishing returns in Raid O, with Scsi 15k drives.

I could be wrong. Could be my old 320 Megaraid 1 controller can't get over 120 mb/sec....

Which type of PCI slot do you have? Is that the limitation?

 

[ddrueding]

OK, OK, I give. I may just get a 64GB X25-E and use it independently. I don't know what to connect it to though. There are no free ports left.

You have that many independent drives? I have 4 (X25-E, X25-M, Velociraptor, 2TB) standalone and the array, and it is already messy. When I read the review of 4x X25-E drives, the part that struck me was the complex controller slowing it down; I would stick the quick drives on the local bus and build an array out of all those storage drives with a RAID card.

 

[Santilli]

PCI Expansion • 1x 64-bit 133/100/66MHz PCI-X (3.3V) slot
• 2x 64-bit 100/66MHz PCI-X (3.3V) slots
• 2x 32-bit 33MHz PCI-X (5V) slots

I could have the Megaraid card in the wrong slot.... Maybe I should check...
Tomorrow....

 

[ddrueding]

So if I were to order tomorrow, and wanted something as snappy as an X25-M that was at least 30GB, the best deal is still the X25-M 80GB? I'm building a low-end (for me) workstation and that is a little steep.

 

[LunarMist]

Most workstations are fine with a recent 7200 RPM/32M cache drive. What is it used for that requires an SSD - heavy content creation or server use? What would you have done a year ago?

 

[ddrueding]

A year ago it would have had a 10k SATA drive in it, possibly even a stock 7.2k drive. But the SSDs are getting amazing reviews. I'm not even telling them what is in the box, and they can tell the difference. I recently replaced a P4 3Ghz/2GB/7.2k machine with a 2.66Ghz C2D/4GB/X25-M machine and the difference is really amazing. The SSD upgrade is great because you don't have to be a power user to appreciate it. Just loading word and firefox is noticeably better.

 

[LunarMist]

Someone did a review of 4 of the X-25E in RAID-0 on a high-end controller, and it actually showed performance loss in low-load scenarios. Otherwise I would be all over it. I may need to do it anyway, I only have 10GB free of the 32GB.

Where is that reference? Is it worse in I/O or in transfer rate? I was hoping that maybe a couple of the X25-M drives would be close to a single X25-E.

 

[ddrueding]

Here is a review, though I'm not sure if it is the same one.

 

[Pradeep]

Where is that reference? Is it worse in I/O or in transfer rate? I was hoping that maybe a couple of the X25-M drives would be close to a single X25-E.

The -M has slower write speeds. But the reads are as fast and sometimes slighltly faster than the -E. A couple of Ms should be faster and more GB for the dollar.

 

[LunarMist]

Here is a review, though I'm not sure if it is the same one.

They are probably expecting 4x X25-E drives to scale better. Don't you think that two drives would be better than one and scale (relatively) better than four?

 

[ddrueding]

That review didn't show it, but the first ones I saw (can't find it now) actually showed the single X25-E outperforming the RAID-0 in low load scenarios.

 

[Santilli]

For what it's worth:
In my experience playing with Raid 0's, two drives seem to be optimal for performance. I've setup as many as 6, but, even though they look great on the tests, they just don't seem to be much of an improvement over a two drive array.

I wonder if that might be true with these SSD's, or, it might be a controller issue?

Keep in mind that is for use as a workstation, in both the old macs and current setups.

 

[LunarMist]

We will just have to wait for David to try it.

 

[ddrueding]

Won't be for a couple weeks at least, 32GB seems to be fine for my OS/Apps, minus games (2TB) and docs(160GB) for the moment.

 

[Pradeep]

For what it's worth:
In my experience playing with Raid 0's, two drives seem to be optimal for performance. I've setup as many as 6, but, even though they look great on the tests, they just don't seem to be much of an improvement over a two drive array.

I wonder if that might be true with these SSD's, or, it might be a controller issue?

Keep in mind that is for use as a workstation, in both the old macs and current setups.

What people have found is that scaling with RAID is highly dependent on the RAID controller, for example the 500MHz Intel controllers (IOP 341) were limiting performance to around 450-500MB/sec on the Areca controllers. As the cranked the speeds up to 800MHz, max thruput also increased. The IOP 348 is now running at up to 1.2GHz, and the Adaptec cards based on this chip will cook to death unless cooled with an extra fan based heatsink.

 

[sechs]

I only have 10GB free of the 32GB.

The horror!

I have 8.5GB free on my 36GB drive.

 

[LunarMist]

36GB? That is the capacity of a SCSI drive.

 

[Bozo]

This looks interesting: Performance loss over time with SSDs.


http://www.pcper.com/article.php?aid=669

 

[Stereodude]

http://www.pcper.com/article.php?aid=669

I was on the fence about getting one, and that article has convinced me they're still not ready for prime time.

 

[LunarMist]

I saw that link elsewhere. I use the X25-E quite heavily, but I am not seeing any deterioration of that sort. I also don't know how accurate ATTO is for the prupose. My Intel SSD has two partitions. The boot partion is about 8GB and has been restored at least 50-75 times so far, and the data partition becomes practically filled with temp files which have been written and rewritten many times.

 

[timwhit]

Some users of non-Intel MLC-based flash drives have worked around their small write slowness by using Microsoft Windows SteadyState. Officially meant to be supercharged version of System Restore, able to ‘reset’ the entire OS back to a default state on each reboot in a VMware snapshot-like fashion, it provides MLC flash users with an added bonus through its execution. SteadyState reroutes all disk writes, regardless of their randomness, to a contiguous ‘change’ file. This brings small write performance much closer to the ‘sequential write’ speed of a given drive. In the case of the X25-M, it will significantly reduce the internal fragmentation that occurs as a result of random writes. If you are a Windows user and willing to deal with the various pros and cons of running something like SteadyState, it may be worth checking out, regardless of the SSD you may currently be using.

That's a pretty interesting idea. Does anyone have any more information about how this works? Can I run this with 2003 Server?

 

[ddrueding]

None of my Intel SSDs (currently 4) are showing this performance degradation.

That's a pretty interesting idea. Does anyone have any more information about how this works? Can I run this with 2003 Server?

That is an interesting idea. Got link? I might try to put some of the slower SSDs back into service if they can get faster.

 

[timwhit]

That is an interesting idea. Got link? I might try to put some of the slower SSDs back into service if they can get faster.

Try: http://www.cdfreaks.com/news/15287-Free-utility-speeds-up-SSD-writing-on-XP--Vista-32-bit.html

And: http://www.ocztechnologyforum.com/forum/showthread.php?p=318673#post318673

This also looks promising: http://www.easyco.com/

 

[timwhit]

MFT looked like the best option until I noticed that you can't use it on a boot volume.

SteadyState doesn't support Windows 2003 Server.

Looks like I am out of luck right now. Damn.

 

[timwhit]

SteadyState doesn't support Windows 2003 Server.

I can now back this claim up with experience. I removed the SteadyState installer's version compatibility check with a handy tool called Orca. It definitely installed, but after I rebooted the system it would almost get to the login screen then reboot again. I tried to mess with some settings in safe mode, to no avail. (Why does safe mode support nothing higher than 640x480?).

I restored from backup and am back in business. The whole thing took about an hour. I guess I never excepted it to work, but it was worth a try.

Acronis True Image works wonders for this kind of toying around.

 

[Santilli]

HMMM. Wonder if formatting the drive with a larger block size would exacerbate, or reduce this problem? While a larger block size might reduce the drive capacity, it might reduce this problem?

Also, thanks for excellent links. It looks like I should wait a bit, and, this might be a part of a general machine upgrade. Also looks like MSFT is working on fixing some of the problems, and, Windows 7 compatibility is going to need to be a requirement.

Good reading...

 

[timwhit]

HMMM. Wonder if formatting the drive with a larger block size would exacerbate, or reduce this problem? While a larger block size might reduce the drive capacity, it might reduce this problem?

Also, thanks for excellent links. It looks like I should wait a bit, and, this might be a part of a general machine upgrade. Also looks like MSFT is working on fixing some of the problems, and, Windows 7 compatibility is going to need to be a requirement.

Good reading...

There are strategies for formatting these drives with a block size that matches the size of the erase block. When a MLC SSD needs to write some data, it has to clear the space before it can write to it. Most drives have to clear 512KB of space before writing. So, even if you are writing a 1KB file, 512KB still has to be cleared. Someone correct me if my understanding is incorrect.

 

[LunarMist]

There are strategies for formatting these drives with a block size that matches the size of the erase block. When a MLC SSD needs to write some data, it has to clear the space before it can write to it. Most drives have to clear 512KB of space before writing. So, even if you are writing a 1KB file, 512KB still has to be cleared. Someone correct me if my understanding is incorrect.

I use 32KB clusters. That should help a little, no?

 

[timwhit]

I use 32KB clusters. That should help a little, no?

What SSD do you have? I think it is going to vary manufacturer to manufacturer and model to model.

If you want a lot more information on this check this thread.

 

[LunarMist]

X25-E

 

[timwhit]

I wouldn't worry about these fixes with that drive, I don't think it suffers from the problems.

 

[LunarMist]

Hmmm.

 

[udaman]

X25-E

Hmmm, foget about Winblows 7, it's just a poor copy/attempt to emulate a superior OS....Mac OS X (older version of course)...get a Mac, get Snow Leopard ...server version (only 2 Snow Leopard's, unlike multi-version 7).

Intel replies to solid-state drive 'slowness' critique



http://news.cnet.com/8301-13924_3-10168084-64.html?part=rss&subj=news&tag=2547-1_3-0-20

A review, however, entitled "Long-term performance analysis of Intel Mainstream SSDs" on technology Web site PC Perspectives claimed, among other things, that the Intel X25-M solid-state drive may degrade in performance as a result of "internal fragmentation" and that "a 'used' X25-M will always perform worse than a 'new' one" and, in some cases, drives "would drop to significantly below manufacturer specs."
The reviewers claimed that they made an effort to reproduce real-world scenarios. "Dozens of different scenarios were played out on our drives. XP / Vista installs, repeated application / game installs, batch copying of files...were all liberally applied to the X25-M." The review concluded that "all three of our SSDs suffered a drop in performance regardless of the type of workload applied to them."

Hmmm, me thinks dd needs to reconsider Andriod and go with the rumored updated iPhone (virtually all the new smartphones are eliminating a physical keyboard, lol), go to the new Server ver of MacOSX .

Mac OS X Snow Leopard Server to pioneer ZFS ahead of desktop

http://www.appleinsider.com/article...d_server_to_pioneer_zfs_ahead_of_desktop.html




Snow Leopard Server to ramp up scalability and performance
http://www.appleinsider.com/article...r_to_ramp_up_scalability_and_performance.html

iPhone-savvy Wiki services and remote access

Apple's web-centric approach to serving businesses' information sharing needs extends to Mac OS X's collaboration services, which provides web-based wikis, blogs, mailing lists, and RSS feeds tied in with Open Directory users, comparable in some respects to Microsoft's SharePoint services.

In Snow Leopard, those features will be enhanced with search across multiple wikis, a template optimized for mobile use on the iPhone, and a central My Page site customized to provide access to all of the updates to the intranet wiki sites a user selects to track.

Along with sending push notifications to mobile users outside the company's local network, Snow Leopard Server also enables mobile access for setting up secure incoming connections to remote users, providing them with proxy service access to their corporate email and intranet websites.

Snow Leopard shared performance updates

Snow Leopard Server will also inherit the same kernel updates as the Snow Leopard desktop version, with full 64-bit addressing to handle massive amounts of RAM. That's a particular advantage in the the server realm, where applications can take full advantage of wide resources to accommodate more simultaneous network users. Leopard Server already employs 64-bit versions of many of its non-kernel services, from Apache web hosting to email.

The move to a 64-bit kernel will give Snow Leopard Server security advantages as well, as noted in a previous article. Other new architecture changes due in the Snow Leopard kernel will also benefit the Server side, including Grand Central technology for optimizing performance on multiple-core and multiple-processor hardware.

Exploring Windows 7 on the Mac: the Taskbar
http://www.appleinsider.com/articles/09/02/06/exploring_windows_7_on_the_mac_the_taskbar.html

Once past the fairly painless installation of the Windows 7 beta, Mac users will be struck with deja vu. This new version of Windows looks more like Mac OS X than any previous edition ever offered by Microsoft.
.....


One size fits all

There are still some differences between the Taskbar and the Mac OS X Dock: the Windows 7 version (which may yet still change before its release) must be manually "unlocked” before resizing it, and then can only be resized in half inch-sized increments.

Even so, resizing the Taskbar neither resizes the icons (as it would in Mac OS X) nor provides more vertical room for organizing Taskbar items (as Windows users might expect). You can’t organize icons in vertical rows, making it fairly useless to change the vertical height of the Taskbar. There’s also no Dock-like magnification.

Taskbar app icons sit within a metallic looking panel which becomes glass-like blocks that highlight when the app is active in the foreground or running. The early betas of Mac OS X in 2000 similarly lacked both a transparent background and smooth vector scaling (below, Mac OS X DP3).

 

[Handruin]

More mac propaganda. You're tiring uda.

 

[LunarMist]

Why would the OS be relevant? An SSD should work with all.