Home Forums Solid State Drives

What is everyone's experience with SSD. Are they worth the investment?

Also in a tower desktop can I put 2.5" drives?

Thanks!

Ben

Well worth it if implemented properly. RAID1 or better, regular backups off the array, used where I/O performance will be best realized ...

There are brackets you can find on Amazon that handle 2.5 to 3.5; the ones we get hold 2 2.5 drives in one slot. You should check to see if you have sufficient power connectors in the box.

Depending on your current config, you may need a PCI[-e] RAID card as well.

I run AC with an SSD drive in a quad core with 12 gb or ram, I should have gotten the 200+gb size SSD....runs great, boots up very fast....am very happy with it....drive prices are reportedly coming down again soon per my tech friends.....I use Newegg to buy things

I was thinking of putting this in our terminal server that I am trying to set up. We have providers that need to access amazing charts from outside the domain name.

We will have up to 5-7 people that will be accessing amazing charts from the terminal server so i wanted optimal performance.

Our "server" too may need upgrading. Currently it has 8GB with and i5.

What would be a suggested setup for us?

Thanks!

Ben

Ben,

AC is going to see the most significant performance boost from using the SSD [in RAID] to run the Amazing Charts directory on them, which would include the database files and Imported Items sub-directory.

I would install the directory on a separate drive [mount point], but keep in mind that AC will assume the C: drive, and some Microsoft installs INSIST on the C: drive.

Unless you are running on a full MS SQL install, memory has limited utility on the server beyond general head-room and running other services.

Is there a way to copy everything on our current hard drive (Windows Server 2008 R2) to an SSD drive and just swap it out with out an hitches?

You are an optimistic soul!

There are easier ways and harder ways, but for the highest success factor/lowest frustration factor, I recommend Acronis Backup & Recovery. It does the best of any of the tools we have used recently for moving between dis-similar hardware. We use it primarily for bare-metal restores to dis-similar hardware, or P2V (Physical to Virtual) transitions.

It is also a great tool for running multiple scheduled backups for the professionally paranoid.

It is available in the channel, let me know if you decide you need additional help.

Thanks Indy!

Yikes! The server license is quite expensive. Hmm. This will be something i have to look into.

Thanks so much!

I am putting together a server would it be better to get SSD than even a 15K RPM Hard drive?

Thanks!

Ben

Ben,

I have had excellent luck just using the built-in Windows Server backup to move a server install from one hardware to the next. Even disimilar hardware.

JamesNT

You can check out PING for an open source alternative:
http://ping.windowsdream.com/

It works pretty well. I've also used Macrium Reflect Free Edition just last week too to migrate my install over to a larger SSD.

In my opinion SSD is not yet ready for critical data. I did put it on my main last year, using Intel SSD. Initially it was good but then after couple mos , mysterious crashes, so I went back to reg hard drives.

I am no longer in the IT enterprise field and not keeping up what going on so I am not so sure about large corporate SSD implementation.

Looking in all Newegg, Amazon feedbacks, there same problems, mysterious crashes, BOD....even a small percentage, less than 10 percent. That might be why only 3 year warranty on most of them.

Go with old, trusty hard drives, even they are slower, but hard drive technology is mature, even the warranty is longer 5 years...

Wait some more time on SSD for AC main computer, but put it on your game PCs now.

I love SSDs on my clients. Have a few problems. To be honest, I wouldn't use SSDs on a production server yet.

But reboots in 12 seconds. Wow!

Thanks, a helpful response! Are you implying I have game PC's in the office?! Never! And I would NEVER play Minecraft at work.

Are SSDs any more reliable today? (For a desktop server)

Why are they reliable/suggested only in clients?

SSDs are still relatively new compared to traditional hard drives. Longevity testing is not as extensive as hard drives. The way lifespans are being determined is by simulating intensive workloads for short amounts of time.

In my opinion, certain models are ready for server usage. Just as with HDDs, there are enterprise/commercial models (Seagate Cheetah 15K, WD Raptor, etc.) and home models (Seagate Barracuda, WD Blue, etc.) Some are more reliable and better than others.

The models I would recommend in low to medium-end servers are the Samsung 840 Pros, Intel 520, Plextor M5 Pro. All with 5 year warranties. For the high end servers, the Intel 710 with eMLC memory would be very reliable. Anything above that and you'll want to look into PCI-E SSDs like the Fusion IO.




The only hard drives that have 5 year warranties anymore are the high end/commercial WD and Seagate Drives (e.g. WD Black, WD RE4, Seagate Constellation, Seagate Cheetah,etc.). Almost every other has a 2 year warranty. It's basically a monopoly in the hard drive market with Seagate and WD being the top 2 and Toshiba bringing up the end. Nobody is exactly jumping to get in that field anymore...

Most of the time I need a drive to be replaced under warranty. I'll just buy it and get the other one replaced later on. The turn around times can be weeks. Even doing an advanced RMA is a headache and I'd rather just pay the money to get it ASAP. Amazon Prime makes that one day shipping nice for only $4. I just want people to be realistic about the warranty. Sure it's nice but don't expect it to be an immediate solution.

I've recently started using Samsung 840 Pros in servers. Samsung is a good brand and they were awarded several OEM contracts from various manufacturers including Dell. Intel is pretty rock solid as well. Though the two I use most at the moment are Samsung (5-10 in production) and Crucial (Over 20 in production).

Sandeep, thanks.

This past week the server in our office was upgraded completely and it has an 840 (not pro) SSD for now. Its only 120 GB (I know its too small for the server) but I wanted to see if Solid State would increase the speed for the 7 workstations. So right now we are basically test driving it.

Originally I knew the SSD would be a world of difference - at least for the server. And my plan was to upgrade to the 500GB capacity once a deal came around. But now I am hesitant after reading this thread and about data corruption incidences with SSD in servers.

So now, I will probably be researching traditional hard drives. Most likely the Velociraptor 1000RPM. The 128GB SSD will go into the Dr.'s laptop.

SSDs & servers -- is that a good idea? Don't SSDs have a limited number or read-write cycles before failure, often abruptly? What do our hardware gurus think?

Write endurance isn't really an issue anymore. Maybe in the '90s it was. For a 128GB drive, you would have to write around 40 Gigabytes a day to it for it do wear down in 5 years. From my experience, failure rate is strongly correlated with manufacturer and model. (Same in the case of hard drives.)

I thought this was a good article for people considering using SSDs in servers:
http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923.html

Another good article on the best SSDs for the money. I would advise you use this to spec/upgrade your desktops. It appears the author shares the same experience when it comes to Crucial SSDs as I. Had them for I think 2 years now. Still running just as good as when I got them. And I've installed them in over 25 systems. All of which are still running well. But, as with all things, they fail. Never hurts to have a backup. You should always have a good backup.

http://www.tomshardware.com/reviews/ssd-recommendation-benchmark,3269.html

The SSD company I have most faith in right now is Samsung, followed by the Crucial M4 and Kingston stuff. Many of the other companies have had fiascos with firmware (OCZ comes to mind).

Intel would probably be number 1 for me, but they are also the most expensive. Not by much though. I liked the Samsung 830 but I don't like how they switched from MLC NAND to TLC NAND with the 840 (non pro). I grabbed 2 830s right before they went out at a bargain price.

I forgot about Intel. I would considering them the top dog in terms of reliability.

Regardless of brand, you have to be hyper vigilant about firmware updates. Even the Crucial M4 had issues where they would abruptly quit after logging about 5000 hours. It was patched with a firmware update.

Yea. I was one of the first to reach the limit. Got it a couple days after release and ran it 24/7. They did release the update swiftly at least.

Intel hands down. Then the M4. Lastly the OCZ. And, I am using it right now. Scary.

One tip I found. Turn your defrag scheduler (on by default) off when you used SSD. I think it's been crashing mine every night. Could be wrong. Sounds good though.

I have been using two Intel SSD drives on an M-Tech server for over a year with no problems. DO NOT defragment SSD drives. That will ruin them. You can google this.

DoctorMel,

I bow down to your willingness to use SSDs on a server. No, seriously, that may help me take the leap one day.

I have eight computers, six worked great with Intels and OCZs. Two different. Spent hours troubleshooting. Installing a new M4 to one computer fixed it. On the other, I had to turn it off every night something I hate and couldn't do, obviously, on a server. I also think that SSDs and Lenovos don't play well together.

Again, not being critical. I admire you and that it works so well for you.

Because if your server crashes, you lose a lot of important data. Who cares if your client crashes. By using your clients, you get faster performance on AC, and you can trial SSDs.

This is just I, but I would never run a server on an SSD. Yet!

Even I touted the bootup time. But, why is that important unless you need to reboot all the time. On your server, you only need to reboot for updates. I generally go months without rebooting except for updates that need to be rebooted.

Barcafan, nice to hear your comments. Would you consider changing your display name or putting a signature at the bottom? Just makes it easier to address you.

Depending on which RAID system you employee in your server gives the 'safety' re you could go RAID 10 with 4 Intel Cherryville 180GB SSD's would run $680 for the SSD's, though guess that changes the I in RAID to an E (expensive.)

Depending on which RAID system you employee in your server gives the 'safety' re you could go RAID 10 with 4 Intel Cherryville 180GB SSD's would run $680 for the SSD's, though guess that changes the I in RAID to an E (expensive.)

I don't see the need for more than a RAID 1 (unless you need the capacity) with SSDs. They are already so damn fast compared to HDDs. Also, things like SSD caches are becoming more popular for people who don't want to take full step to SSDs. Also if you think about it. You really only need the database on SSDs. (And as far as I know, even physicians practicing for 10 years don't have DBs larger than 200 MB.)

These SSDs you see failing usually have nothing to do with write endurance but instead of the quality of the parts that go inside it.

Write endurance isn't a real issue for our databases who barely see any action. Those eMLC and SLC SSDs are targeted for those people who write large amounts of data per day (I'm talking on the order of hundreds of GB/day).

Agree the databases prob aren't very large and the amount of info asked for at any 1 time ie a pt chart isn't really that big
So 2 Intel Cherryville 240 GB SSD's in Raid 1 would be about $400.00 really not a bad price, I probably should get the second SSD and RAID 1 my system now.

You could always detach and move the databases to a RAID 1 of low capacity SSDs. 240GB seems a bit large unless you plan on throwing the II in there as well. This wouldn't be necessary if full SQL was used since it would probably just all sit in RAM.

I said 240GB for those 'biggest' practices, I currently have a 180GB Cherryville and am just waiting for the next price break, thanks for your input glad you keep up on this board.

OK, the voice of reason. I think Sandeep just says put the OS, maybe programs, etc. on the system drive. You could still put the Exchange mailboxes, Sharepoint data and other stuff including AC on HDDs.

I just think I would go through one more cycle (a server of five years) before going with SSDs. My programs smoke including AC on clients with SSDs.

I think the redundancy is meant for an unforseeable crash rather than one that is possibly likely.

But, to each his own. And, as Sandeep will tell you, I am always looking for speed, speed, speed. But, not on a server.

I mean if you're going to use some high quality Intel drives. I think you would be safer than you be with a hard drive. I really do like their testing process. They analyze every failed SSD they get back to improve their tech. Test thousands of drives and really do their best to keep the lowest ARR in the business. MTBF doesn't mean anything. What you should be interested is in the ARR or Annualized Return Rate. Very hard info to find but some stores publish and can give you an idea of how often they *really* fail.

Is there something different with SSDs in RAID?

I am trying to install AC 6.0.9 on a new server - IBM x3200 M3 with BR10il, so I decided to give the Samsung 840 Pro a go as the second RAID1, with a pair of WD as the first RAID1, with intention of putting AC on the Samsungs. Before I decided to go with SSDs, I had the setup working well with 4 WDs in two RAID1s running on the same server.

However, even though the server can recognize each individual drive perfectly fine, I cannot seem to be able to convince the IBM ServerGuide for W2K12 to set up the second RAID1 (it gives a RAID 1 + 1 option, setting up the first two drives as RAID1, and remaining drives ad RAID1). It would set up the first RAID1 fine as Array 1, but giving a mysterious error message "Exceeding the array limit, the maximum is 1%".

I did have trouble finding trays that would support the SSDs physically in the correct spatial relation for the pinouts, so I initially tried ICY DOCKs so they fit perfectly in the bay, but they did not seat the SSDs properly. So I currently plugged the SSDs directly on the backplane of the RAID bracket and supported them on cardboard (gasp). However, each drive seems to work individually fine, just could not RAID them.

I don't know. Sandeep will help you. But, for me, SSDs are just not ready for a production server. Just my opinion. I have them on all eight clients.

Hi Jack,

This is an unrelated question.

Why 6.0.09?

This was the dog of dogs for us, significant speed issues, and some MU reporting problems.

We have been happily using 6.1.2 for the last year or so, doing very well with this version.

Gene

I just need to get my infrastructure done while waiting for 6.5. As usual, not really sure when 6.5 will be general release. I still need to work in the meantime, so I need to migrate the current setup. If/when I upgrade to 6.5 the hardware will be all ready. I was running AC 6.0.9 on the SBS Server 2008, and recent updates of SBS made it more of a resource hog. I am trying to get a hyper V replica going, installing separate AC server and its replica on different virtual and physical servers.