[sims] Sims Server Performance Benchmarks

[Chuckster]

150+ student school (SQL 2008 R2, Solus 2, Discover)

 

Sims server

HP ML370 G5 (6+ year old H/W), VMware ESXi 4.1, Win 2003 STD SP2, 3GB RAM, 2 Cores Xeon @ 1.8GHz, 2 x 72GB 10K SAS drive (RAID 1+0), 2 x Gbit NIC

 

Client machine

Intel i3-3220 3.3GHz, 4GB RAM, 250GB HDD @ 7200 RPM, Win 7 Pro SP1 x86

 

Sims Benchmarks

2.31 Seconds Client Load time

5.52 seconds Reports >> Students >> addresses >> Year 7 >> 1 Reg Group

0.76 Seconds Search "Smith" student name in main screen search box

 

 

Sims Database

Sims.mdf - 1.85 GB

Sims.ldf - 184 MB

Simsdiscover_Primary.mdf - 73 MB

Simsdiscover_Primary.ldf - 69 MB

 

Thankfully, for us, ours is relatively 'fast'.

 

Edited bit: I should add that at first run time, the load time is 18 seconds - less on the server. Any subsequent client runs after the first run time loads in < 5 seconds.

Edited November 11, 2013 by Chuckster

[matt40k]

@CAM agree about Discover, whoever got that pushed though Capita needs a medal.

 

Performance Management was also meant to be free but had to be charged due to the same issue.

 

Not sure I believe that. Don't think any MIS provider would complain if Capita started developing for free! If they would, let me know and I'll go issue some "customer feedback"

[localzuk]

@CAM agree about Discover, whoever got that pushed though Capita needs a medal.

 

Not sure I believe that. Don't think any MIS provider would complain if Capita started developing for free! If they would, let me know and I'll go issue some "customer feedback"

 

They do complain to be fair. As it means adding new features, outside the tender, which due to the dominant market position could be claimed to be anti-competitive, as they would be 'forcing' their way into a new market. Kinda like Microsoft bundling Media Player and Internet Explorer into Windows etc...

[vikpaw]

They do complain to be fair. As it means adding new features, outside the tender, which due to the dominant market position could be claimed to be anti-competitive, as they would be 'forcing' their way into a new market. Kinda like Microsoft bundling Media Player and Internet Explorer into Windows etc...

 

Absolutely, and they are always at the edge of being accused of abusing their dominant position. And, usually not allowed to discuss a lot of these issues in public.

 

I don't think this would apply to fixing existing issues, or improving older products. Modernising a product e.g. Nova, which will still essentially do the same task, isn't abuse of position, just making your customers happy. Giving away improvements should be mandatory!

@matt40k - if Capita gave away intouch for free every messaging / texting company out there would complain, and rightly so, it's something that was never sold as part of core SIMS. I'm not sure what they'd gain by giving it away, apart from lock-in, and maybe some good press for a change

 

The smaller companies have it easy in a way, as they can copy what SIMS does, or wants to do, and give it away with their core offering and get away with it, plus it's probably easier to program, make it look good and work well with existing systems.

 

I doubt very much if being King of the Castle is an easy position to be in.

[matt40k]

You're being silly now @vikpaw. InTouch is new business, upgrading Nova isn't. It's common sense. They don't do it because if they jumped from v6 to v7 for everything everyone would need to re-tender as it's a new product. It's why a lot of companies that sell to government release a new product but still support the old version because they know if they didn't support the old version they wouldn't win the tender again.

 

If it was really was the fear of other MIS suppliers complaining if they upgraded, say Nova, or the reporting engine, isn't Capita large enough to take it on the hip? Let it go to court, then it'll be public knowledge that a certain MIS supplier(s) is basically say hey, we're not happy that Capita is trying to save Schools money and improve the overall education of our children. Technically speaking, Microsoft sue them because schools aren't buying Lync because Capita included a messaging and alerts service within SIMS. Even if this was 100% true, Microsoft would never do it just because of the bad PR. For me the law stops them from setting the rules rather then improving the market. They can't for example define a CTF for staff which would not work for the community of MIS providers at large, or enforce licensing costs.

[vikpaw]

I'm serious. I don't get it. You shouldn't have to re-tender just because the version increased, even if it was the entire underlying architecture or do you .....?

I don't think anyone would complain if they upgraded Nova or reports, but they defo do if they release new products / 'features'.

 

I'm not sure these cases reach the public domain, i don't think the complainers dare hold their hands up, sometimes it's probably anonymous complaints.

 

I think they have it wrong and they should be fixing the old issues rather than investing in new areas, but the impact of this re-tendering could be the reason if what you say is true.

 

We should probably split this chat off into a different thread, we've really veered off the benchmarking process now.

[matt40k]

I don't get it either mate, but we had it here when the HR system went ASP.net from ASP. That's why we're still using the ASP version (well till it gets replace next year). I get why you have the tender process but it's like other processes, it works 95%+ of the time, but the other 5% it shots you in the foot.

 

Not sure it is anonymous, if it is, it shouldn't be.

 

PS: Yes, just a bit. Still not seen any terrible benchmarks yet, 32 seconds to load SIMS isn't too bad. Still takes me 10mins to login to Windows in the morning!

[zag]

We have just spent 3k+ on this software and hardware getting it all updated and I am still get regular complaints about it being "slow" from users.

 

- I've purchased brand new hardware

- Paid Capita to come in and do the install (they did a great job)

- Did all the recommended things like putting SQL data on a 2nd drive, using SSD, 10GB of RAM, 4 CPU cores ect ect

- Set it up as Capita advises

 

And its still horribly horribly slow to the point where my users are blaming me (so I need someone else to blame haha)

Edited November 11, 2013 by zag

[matt40k]

Explain how you are limited to slowest part and you have tracked down the bottleneck to the users - 11 secs - you're typing too slow!

 

To be honest, 11 secs startup in the morning to cache pretty much everything, leave it open all day, then shutdown at the end of the day. Not sure how you could really explain any futher costs trying to improve that. Lets say you get it down to 5 secs, that's 6 secs saved per day say 50 staff, that's a saving of 5 mins a day, that's what £4 a day efficiency saving? Pretty sure getting a tea lady to bring all the staff tea and coffee would make more sense. Are you sure it's not that all your users are connecting via 1 wireless access point or something silly like that?

[vikpaw]

It's not the login time it's the report time that is a killer. I know where @zag is coming from. When a teacher or any staff want a list of kids waiting 15 secs is just too long, and perception will be bad.

[CyberNerd]

 

To be honest, 11 secs startup in the morning to cache pretty much everything, leave it open all day, then shutdown at the end of the day. Not sure how you could really explain any futher costs trying to improve that. Lets say you get it down to 5 secs, that's 6 secs saved per day say 50 staff, that's a saving of 5 mins a day, that's what £4 a day efficiency saving? Pretty sure getting a tea lady to bring all the staff tea and coffee would make more sense. Are you sure it's not that all your users are connecting via 1 wireless access point or something silly like that?

 

 

11 seconds adds up.

10million kids in the UK. Capita have an 80% monopoly with 8Million kids.

5 days per week, 38 weeks per year. 8000000*5*38*11 = 193,000 days !!

 

Assuming the teacher doesn't move rooms so doesn't need to log in every period to do a lesson then Capita are responsible for a hell of a lot of lost learning time every year in the UK.

[vikpaw]

That would be right if every kid wanted to login to SIMS once a day and wanted instantaneous response.

 

Plus Capita have way more than 80%!

 

I'd concede 3000 days if your 10M start value is right.

[CyberNerd]

That would be right if every kid wanted to login to SIMS once a day and wanted instantaneous response.

 

Plus Capita have way more than 80%!

 

I'd concede 3000 days if your 10M start value is right.

 

I got it from Wolfram Alpha.

 

It probably makes more sense to work it for the 620,000 teachers then. Say 80% make that: 0.8*620000*5*38 = 26,000 hours.

Anyone care on working out how much this is costing the DfE based on an average £35k wage?

[caffrey]

Do the specs of client machines make any difference to load times ?

[CAM]

I got it from Wolfram Alpha.

 

It probably makes more sense to work it for the 620,000 teachers then. Say 80% make that: 0.8*620000*5*38 = 26,000 hours.

Anyone care on working out how much this is costing the DfE based on an average £35k wage?

 

But 11 seconds is still insignificant. That's a politician/manager statement to justify a snap decision.

[zag]

Do the specs of client machines make any difference to load times ?

 

Absolutely no difference at all

[zag]

It's not the login time it's the report time that is a killer. I know where @zag is coming from. When a teacher or any staff want a list of kids waiting 15 secs is just too long, and perception will be bad.

 

Yes exactly. I'm not really complaining about report speed here, I can understand that takes time. Its the overall usage of the software, in particular searching for students is tediously slow. Loading the app takes a lot of time also which creates a bad impression right away.

 

These should... in theory be easy things to fix.

[jinnantonnixx]

I'm trying to get around performance problems, and on my travels I discovered this; it won't help but I enjoyed the article. (I know education discounts are large)

Are Commercial Databases Worth It? | Coding the Wheel

 

Nowadays I like to think of SQL Server and Oracle as the Death Stars of the relational database universe. Extremely powerful. Monolithic. Brilliant. Complex almost beyond the ability of a single human mind to understand. And a monumental waste of money except in those rare situations when you actually need to destroy a planet.

 

 

Which brings me on to my next question, which I shall phrase rhetorically. Most (or many) of us are running enterprise-grade OSes, enterprise-grade SQL servers on enterprise-grade kit. This stuff is supposed to be able to cope with millions and millions of transactions. So why does pupil data for schools cause problems?

Edited November 12, 2013 by jinnantonnixx

[vikpaw]

I swear the app and the client specs play a huge role. I frequently pull data out of SQL servers direct and never have performance issues.

The problem is between the server app call receiver and the client aka: PIBSACRAC

[localzuk]

The problem is the way the data is stored.

 

Pulling pure data from the server is pretty quick - that's what SQL does.

 

However, when you then try to relate that data to other data in a database which isn't fully normalised, you end up with a major slowdown as it has to do a heck of a lot of processing when it shouldn't have to. The same issue if tables aren't properly indexed.

 

This is requested by the application, but processed by the server. Some aspects are processed client-side (the word/excel type reports), so client speed applies there.

[vikpaw]

True, but I've borrowed or created my own multi-joined queries recently, even across multiple virtual servers from totally different services, and the performance is still pretty awesome. I know the same data in the app would take an age, and worse still for example if it's my photo db if i call on the photo field even if i don't want to display it.

 

If we can track down the query / stored proc that is being run as part of the benchmark i'd be interested to see performance run in sql management studio from the client machine, i bet it's much quicker. The client is relying on a lot of RAM to do anything and then display it.

 

I don't agree with the earlier statement about leaving SIMS open all day, i recommend closing it every hour, and more frequently if doing heavy editing. If you've imported photos, you've had it. It won't release the memory or the last photo used until you completely shut it down.

 

Not to say all the blame lies at client side, i'm just saying i think it's a big factor. I know different clients around our school suffer differing performance and that's not a server issue.

[matt40k]

I don't agree with the earlier statement about leaving SIMS open all day, i recommend closing it every hour, and more frequently if doing heavy editing.

 

Closing it at the end of the day is wise, closing it every hour means you'll be re-building the cache on the client - so basically you'll be undoing all the hard work Capita have done.

 

The issue you've got is like @localzuk has said it should be fully normalised, so when you write someones new telephone number you only do 1 write. However from a reporting point of view you want it to be dimensional basically because you don't want to have to do loads of joins as they are performance heavy, you just want one table with all the keys in it (Fact) and a few tables off that with the detail (dimensions) and the keys want to be simple int keys. Problem they've got is it's a mismatch, they've tried to please too many people.

[jinnantonnixx]

Closing it at the end of the day is wise, closing it every hour means you'll be re-building the cache on the client - so basically you'll be undoing all the hard work Capita have done.

...

 

Caching of what? The data is considered volatile and re-read with every move. You can show this with Profiler and Wireshark. For example, the pupil browser will run the sta_pix_EditStudentInformation_getstudent ...etc.... stored procedure every single time, even if you repeatedly flip-flop between two pupils. A quick experiment shows no caching is used. Wireshark shows the text (in clear text) coming from the server every time over the network. Perhaps I've misunderstood what you mean.

 

 

The issue you've got is like @localzuk has said it should be fully normalised, so when you write someones new telephone number you only do 1 write. However from a reporting point of view you want it to be dimensional basically because you don't want to have to do loads of joins as they are performance heavy, you just want one table with all the keys in it (Fact) and a few tables off that with the detail (dimensions) and the keys want to be simple int keys. Problem they've got is it's a mismatch, they've tried to please too many people

 

I have to admit they've normalised the data quite a bit since I last cared to look at it. In your example, updating a phone record writes to the sims_telephone table only. This is a simple table linked by person_id - well normalised in anyone's book. A while ago, if memory serves (and I might be quite wrong), there were several telephone fields within the person record, but this has changed. I dare say someone could find horrors, but I think they are improving the underlying structure.

Edited November 13, 2013 by jinnantonnixx

[matt40k]

One would hope that they don't re-query the school name for starters. Not sure I can be bothered to decomple it, even if their is a file call "Cache.dll" and expand on my pretty poor list of "the school name" of things that Capita cache on each login.

[vikpaw]

I'm not convinced by the caching argument, mostly because general usage eats up enough RAM over time to obliterate any positive impact caching might be making.

In practise i've found an hourly or bi-hourly shutdown of the client helps loads, and is worth the 11 sec hit / longer because it appears to make things run faster.

 

I just tell users to close it down whenever they take an eye-rest break which they should do hourly anyway. I'm sure most leave it on all day, but they know when it starts to slow down what they need to do is turn it off and on again! I can't be sure why this works, but it does. That or the presence of an IT Technician in the room makes a huge difference to performance for some unknown reason.

 

I had it on for half a day with no apparent impact just the homepage on screen, and it had used 100 MB RAM. Flicking through a few staff records and registers and it has tripled and not been returned. Maybe it cached the photos, but it was working quickly enough anyway. Getting the student list - registration groups query parameters box, still took 15 secs. Don't know why. Running it was quick though. Loading a second time has now reduced the load time slightly.

 

Running the benchmark addresses report is taking 15 secs still though, to pop into Word.

 

Perhaps someone from Capita could post their benchmarks and we can see if it's of the same order as ours, if so, then it must be running as expected.