Home Forums Server hardware

About a year ago I built a server with SuperMicro board and E3 Xeon Sandy Bridge processor + about 8GB memory

This has worked out very well, but it slows down some when the front office is doing a lot of scanning or especially during a backup (the staff likes to back up MedWare before they go home; of course AC has auto-backup that can be set for later that evening.)

The question -- for Sandeep and Bert, mainly, since they argued in favor of the server setup instead of peer-to-peer:

What if I upgrade to a new Ivy Bridge E3 1280 ($660) and an Intel Ramsdale 910 PCI-E SSD ($2000) and double memory to 16GB.

Would any of that really make any difference?

Who uses those high priced items?

I currently have a Crucial 256GB SSD which has been very reliable to date.

Before the experts weigh in, I can't count the $ I've spent on hardware that didn't help or wasn't really needed.

Consider a step by step process, in increasing order of $

Double the memory, never a waste

A program or batch file to backup MedWare later in the evening

An old or cheap computer to receive the scanning and then transfer to server after hours.

I doubt you need the new processor or SSD if what you have is working fine most of the time.

Tom,

If I may posit a suggestion, I would add a SATA PCI-e RAID card, and two SSD disks in RAID 1.

It is reasonably straightforward, non-destructive to your current config (always backup), and will allow you to test the relative performance of running the AC directory on the RAIDed disk with the rest the same.

Bert and I differ on this, but based on years of experience in the enterprise space with the increasing use of SSDs, it is less likely that you will find spindle drives than SSDs to serve data from databases.

For years Oracle and IBM has sold servers packed with PCI-e solid-state memory for database servers and transaction servers. They have made a boat-load of cash doing it.

Well, we don't differ as much anymore. SSDs are becoming more and more stable and mainstream. I have a slight bias since I had three out of eight not work, not so much crash. Of course, they were OCZs, which I will never buy again. The CPU is a very good one. Not sure if you need it. The SSD is going to be pretty fast, but as Indy says and he beat me to it, I would definitely go with RAID and definitely back up. I am not sure if we are comparing apples to oranges when it comes to corporations running SSDs or you and me.

Bert - I stand corrected then. Glad to hear we are in harmony.

As we rack new NAS units, our latest is about half SSD, I suspect our next will be almost all SSD and we will reserve the spindle NASs for near-line backup arrays.

Hook me up with a couple of those.

It would be good to know your current drive and network layout.

Are you just using an SSD or do you have a hybrid setup? It's likely not the processor. Those Xeon E3's are quite powerful. The 1280 is only marginally better than the 1230. As in barely noticeable improvement.

Do you have a gigabit network? It could be that your network bandwith is saturated. 100 mbit network can be saturated by scans quite easily.

The other question I forgot to ask as far as setup (in fact I would be very detailed here) was what other storage you have on your server. Are you using regular HDDs for data, etc.?

What is the speed of the Crucial 256 SSD? 1.5, 3, or 6? Is it connected to a 6 Gb/s SATA port? How much space is left?

Thanks for all the input.
It takes me a while to digest this and decide what next steps to take.

I haven't decided to go with RAID yet -- mostly because of the increased expense and complexity, because my personal experience has been that other things fail before the hard drives, and we have multiple daily backups and another machine that can be quickly reconfigured to be the "server" Most we have to re-input is half a day's work. Could happen.

The LAN switches are all 10/100/1000 -- but we still get sluggish performance when the staff is doing a lot of scanning. Maybe the computer the scanner is attached to is only 10/100? So many things to think about. Sometimes I have to take a break and go see patients!

I have been using one of those Crucial M4 (I think) 256G drives that are quite inexpensive on Newegg -- and it certainly has been reliable. Use a couple of other "spindle" drives for backup -- but we also back up to other computers in the office and to a thumb drive.

No one answered my question -- who uses those far more expensive versions of processors and SSD's -- especially those $2000 PCI-E drives? They don't seem to add much in terms of performance. Do you get increased reliability for the money?

What is the real difference between the E3-1230 and E3-1280 processor, besides $400?

***********************************************

Moving to the Cloud avoids all that -- when the Cloud is up and running, that is.

[ It's worth remembering that the idea of a perfect "cloud" environment is pretty old: from Wikipedia, Cloud Cuckoo Land refers to an unrealistically idealistic state where everything is perfect. ("You're living in cloud cuckoo land, mate.") It hints that the person referred to is na?ve, unaware of reality or deranged in holding such an optimistic belief.

The reference comes from The Birds,[1] a play by Aristophanes in which Tereus helps Pisthetairos (which can be translated as "Mr. Trusting") and Euelpides ("Mr. Hopeful") erect a perfect city in the clouds, to be named Cloud Cuckoo Land or Nephelokokkygia....]

That's the E3-1280V2 linked above. You'll need a BIOS update to use the V2. We're talking less than 10% performance gains for double the price. The E3-1280 just has a higher clock speed. Nothing really beneficial about it. I doubt it's related to CPU power. The gains are so small that I recommend the E3-1230V2 (1155)/E3-1230V3 (1150) when building servers. Anything more and I tell them to move to a newer socket.

I do tend to use high priced hardware. My RAID card is $1500 alone. (I need it for my 24 hard drives.) But it only makes sense if there's an actual benefit. Those PCI-E SSDs are good for enterprise database applications not the light read/write load of a medical practice. The main thing is the write endurance with those SSDs. They tend to be reliable and use SLC memory instead of MLC. It's good for companies with very large, intensive databases.


Not really. Importing those scans in the cloud would take a long time. If your network is feeling it now even at a 100 Mbits. Imagine how it would feel at 10 megabits. Most people find workarounds those bandwith limitations. Keep the scans in folders or something of that nature.

Moral of the story. The hardware you have should be more than enough. It's something else related to networking or some configuration issue. Maybe you have more than just SSD storage. 256GB seems really small for all the practice data and AC.

Oh -- Thanks Sandeep.

Lots of other storage. The M4 is just for OS and programs.
Storage elsewhere. "Spindle" drives.

I think the server MB must have only a 100MB network card built in -- that is what I am using, and the Task Manager says we are running 100 even though the switches are 1000.
Maybe need add-on card for the server?
Need to look into that.

Ipso facto, quid pro quo.
So little time, so much to know.

Your speeds are going to be bottlenecked at your lowest connection especially if it's the server.

You may want to just do that first. But, while you are in there add a 1TB HDD.

Hi Bert --
I'm going to have to spend some time increasing LAN speed. Don't really know what the problem is, too many other things to do. Currently running 100mb -- fast enough for most things.

If part of the LAN is 1000mb and part is 100 -- will it all run 100, or will the 1000 legs run at full speed? In other words, can I just upgrade the server and the computer attached to the scanner -- or does everything have to be upgraded at once for it to work at fast speed?

Already have 1TB -- need more! Not sure why -- but I think the server backup just keeps adding backups until the disk is full, then starts writing over the oldest ones.

The connections need to be 1000. More likely the line goes from Server (probably has a 1000 NIC) to either a switch or the router. These, then, would need to be 1000 as well. This is most likely your biggest expense.

If you had a separate NIC card for the router to a room, you might not need to upgrade anything else, but it becomes impractical if you have more than 2-3 computers. The rooms probably have 1000 NICs as well, these have been standard for a while.

The server does keep adding backups. You should go in periodically and clean them up. I just deleted about 300 G of backups (I do full with imports) spanning back about 3 months. More room is better. How many backups do you need? (I generally keep about 1 a month when I get more than a month old.)

His NIC card is 100MB according to one of his posts.

@Tomastoria 1TB is likely enough. It just wasn't clear, at least to me, that the extra space was in the server. I, myself, put all of my data on a different drive than my OS for multiple reasons. I would think all of your shares would be on a different drive. At least the ones where speed isn't an issue.

One thing you haven't answered is the speed of the SSD and do you have it connected to the correct port. I doubt you have 1.5Gb, but you could have 3 or 6. A 6Gb SSD is going to be faster than a 3. On the same token, If you have a 6Gb SSD, but you connect it to a 3Gb port, your performance will significantly drop.

As far as 100 vs 1000, etc., your speed is dependent on the lowest bandwidth. I wouldn't make it complicated. If data has to run through a 100Mb card, then it will be slower than a 1GB card, dependent on where information is going through the card. But if computer one and two are 10MB, computer three doesn't really care if it is 1GB and goes to a 1GB switch and then to the server. While your NIC card settings can be 1000MB, you are best to set it at Auto-negotiation, then the cards will figure out the best setup for the best performance.

As far as the SSD PCI card, I doubt too many people use it for a Small Business Environment. It is a little bit overkill. It is an enterprise card. You don't have to do a RAID for fault tolerance IF you have good backups of the ENTIRE server, but it does always help to have RAID for redundancy and increased performance.

I am a little confused on the Medware. Are they just closing Medware on each machine, then running a backup? That's pretty nice that you have staff like that. My staff wouldn't know how to do the backups. Then, again, it wouldn't be very hard to teach them. But, I usually see patients after anyway, and I am completely OCD about backing them up.

There are scripts all over the board on how to auto delete your backups. Personally, it is simpler to just use AutoDelete:

http://cyber-d.blogspot.com/2005/10/cyber-ds-auto-delete-101.html

Before you change NIC cards, I would make a 1GB file out of PDFs and copy them from the server to other computers with 1GB NICs and 100MB NICS and time them. Then do the same once you install a one gig NIC. But, as everyone has suggested, I would definitely go with ALL 1GB connectivity. And, of course, make sure you are using Cat5e. You only need 1GB on your switch. Your router firewall won't have anything to do with your LAN speeds.

Most of the GB should be Gb. Sorry.

Once again, Bert, Thank You.

Things running pretty well right now -- just looking ahead to see how to improve performance in the future.

Yes, the staff does all the backups.
With MedWare, we back up the entire MW folder -- programs and data. MedWare has never told us (despite asking) which files need to be backed up and which are stable or even redundant. They have never made backup really efficient, thought it would seem a simple enough (and very useful) little add-on.

Since I am the only "IT guy" in the office, and I really have no training -- just bootstrapping -- I am trying to educate the staff as much as possible so they can help and figure out some of the simple things that screw up.l Some of them are very savvy indeed.

There is always something going wrong in these computer things -- they are much too complicated, and non-intuitive, even in 2013-- to really work well in an office that has been paper-based for 35 years.

You can always zip and back up the .TPS files (database) and not back up files which can be recreated with an install. It is too cool that you can copy the folder to a thumb drive and plug it in at home and run it from there. Personally, I do what you do and back the whole thing up. Takes two minutes longer.

I would have them back it up, then you back it up before leaving for home. This is what I refer to as a modular back up. Back up the whole server every night, then back up the individual backups.

Medware has great support I think. Plus, updates are simple when you see the green update icon at the top.

The only frustration is is it Medware, Sage, Vitera, Viagra, what is the name? Very confusing.