Smugmug performance question

jfriend

I am frequently involved in pbase vs. smugmug debates (often on dpreview). In general, smugmug has so much going for it that people are usually swayed to pick smugmug over pbase.

The ONLY two issues that occasionally come up in favor of pbase are performance of the site (snappiness of page loads) and inline EXIF display.

While I think the inline EXIF display would be nice as an option for some photo hobbiests (particularly for use in photo forums), I doubt you are losing much business over that point.

But, the performance issue seems harder to combat and, I must admit, it feels like smugmug has really slowed down over the last 6 months - particularly when I'm trying to get through a lot of pages for some type of repetitive maintenance, but also when just viewing.

Do you have any measures of this on your own site? Do you have any metrics versus pbase that I could use when this point comes up and I'm trying to defend/promote smugmug as a better choice? Any plans for improvements in this area?

--John

Nikolai

Here it goes again...
John,

SM may be occasionally slow, but it almost never down.
I don't have enough fingers and toes to count all the cases when I saw red Xs in place of the PB-based pictures.

And if you're speaking from the photographer perspective - I can't emphasize enough the importance of having unlimited space (as opposed to unlimited bandwidth and meager 200Mb)

But this is probably just me:-)
Cheers!

jfriend

Nikolai wrote:

John,

SM may be occasionally slow, but it almost never down.
I don't have enough fingers and toes to count all the cases when I saw red Xs in place of the PB-based pictures.

And if you're speaking from the photographer perspective - I can't emphasize enough the importance of having unlimited space (as opposed to unlimited bandwidth and meager 200Mb)

But this is probably just me:-)
Cheers!

Dont' get me wrong. I already use those arguments and they often work. I'm looking for ammo when those arguments aren't important enough to the buyer and they are saying that SM is much slower than pBase.

--John

onethumb

jfriend wrote:

Dont' get me wrong. I already use those arguments and they often work. I'm looking for ammo when those arguments aren't important enough to the buyer and they are saying that SM is much slower than pBase.

--John

We heard this about two months ago, that PBase was slower, so we carefully graphed the response times from 18 locations around the world, as far away as Helsinki and Paris and as nearby as San Jose and San Diego.

We graphed a random pbase gallery with a random smugmug gallery, every 5 minutes, 24 hours a day, for a week.

We then compared the graphs and found them to be amazingly comparable. We did, however, come out on top. I'll dig up the graph and post it again, but you can see it on a few related threads over on dpreview.

The bottom line is that both sites are fast, smugmug is far more stable, and smugmug has a few slight disadvantages going for it:

- we display a larger image than pbase does on their galleries
- all of our images are bigger byte-wise because we're more sensitive to preserving the photograph
- we optimize our pages for modem users throughput (2X+ on most pages) at the cost of a few hundredths of a second to broadband customers

Despite these disadvantages, we come out marginally faster around the world.

For comparison, we also compare ourselves 24/7 to the top 100 websites in the world from those same 18 locations. We're above average, but fairly close the middle.

We're significantly slower than:

google.com
foxnews.com

We're significantly faster than:

ebay.com
sbc.com
home.americanexpress.com
drudgereport.com
amazon.com
msnbc.msn.com
launch.yahoo.com
imdb.com
usatoday.com
nfl.com
sportsillustrated.com
ucows.com
expedia.com
vividendiuniversal.com
bt.com


PBase is roughly in the same ballpark relative to the above.

Finally, we were significantly slower than usual over the weekend, particularly on Sunday, because Rafael Nadal won the French Open and quickly did 150+ GB of traffic through us. Yes, that's 150GB. Needless to say, the world wasn't prepared for him to win, and neither were we.

Don

jfriend

Thanks for the data - that should help some

onethumb wrote:

We heard this about two months ago, that PBase was slower, so we carefully graphed the response times from 18 locations around the world, as far away as Helsinki and Paris and as nearby as San Jose and San Diego

...

Don

Thanks for the factual data. That should help me make an argument that the two services are pretty similar in speed with maybe a slight edge to smugmug.

I've always wondered if you also had a slight disadvantage because of the unique DNS entry for each site, many of which are likely not popular enough to be widely cached - though that should only hit you once one your first page.

--John

jfriend

Why would a pbase and smugmug user say this?
I wonder why a user who has used both pbase and smugmug would say this about smugmug having slower performance (quoted from a dpreview thread)? Do you think they are just wrong? Or does it vary for some users due to their network geography and is significantly slower for some users? Or could they be comparing some different feature?

"I started out on pbase and due to the technical problems they were having last year, decided to try smugmug. I forgot to cancel after 7 days, so I have an account there as well now. While I've been "transitioning" out of pbase and on to smugmug for several months now I've noticed a few things. Of course, I anticipate ending up having most if not all of my pictures on smugmug mostly b/c of the highly convenient bulk uploader and being able to charge easily for prints. However, there is one thing: response time. I'm surprised nobody has mentioned it but pbase is way, way faster when you click on a link in displaying the image than smugmug.

Oh well, as they say smugmug is constantly upgrading. I hope that includes processing/response in addition to features. Otherwise, it is a clear cut winner features wise."


In case you're interested, it comes from this dpreview thread: http://forums.dpreview.com/forums/read.asp?forum=1019&message=13816827.


--John

Baldy

Hi John,

First of all, you've been a great supporter of ours on dpreview and we really appreciate it.

Second, the metrics onethumb is seeing do seem to indicate that smugmug is faster, but this is a persistent observation from some people that I'd really love to understand because there must be something we're missing for some people.

Here are the graphs onethumb was referring to:





Since those charts were made, we added more bandwidth from different suppliers. Since not every supplier is fast to every area, we choose the optimal route -- the supplier who is best to that area.

I've chased a few of these down and on two occassions I found banner bloat in customizing.

One thing that makes me crazy is when I type in some else's home page besides mine, there is this pregnant pause that last 7 seconds or so. Our servers report that they are responding quickly, but Firefox seems to be doing something, possibly. Researching this one.

The other thing is it's possible that Elegant style is perceived to come up slower than traditional, which pbase uses.

Any specifics you have we can chase would be helpful.

jfriend

In search of specifics

Baldy wrote:

Any specifics you have we can chase would be helpful.

I'm in search of specifics next time someone states that objection. I've already asked in the most recent dpreview thread where it was mentioned. My guess is that most users won't have specifics, but it can't hurt to ask them.

I hear this "statement" a lot on dpreview that pbase is faster. People don't seem to be saying that smugmug is intolerably slow, it's just one of the points they seem to give to pbase. Since pbase is often the competition for smugmug on dpreview, I'm just trying to figure out how best to respond and what facts can be used to help them understand if that's really true or not.

--John

jfriend

Ahhh, so it varies by geographic location

Baldy wrote:

Here are the graphs onethumb was referring to:

Hmmm. I can see from these graphs that there is no single answer for who's faster because it varies significantly depending upon location. If you are in Vancouver, smugmug will be significantly faster. If you are in Atlanta, pbase will be significantly faster. This is presumably due to the fact that it looks like you are hosted on the west coast and they are hosted somewhere else (must be further east).

This might be good news because I can use this to defuse the fact that pbase isn't faster everywhere and it might depend more upon your location and ISP than anything else. So, try Smugmug out for yourself and see if it's good for you. We both know that once you get someone to try it out, most people stay. My referrals are running about 9 out of 10 for staying once they sign up for a trial.

--John

jfriend

Any info on smugmug vs. imageevent?
While we're on the topic of competitive info, do you folks have any nice summary of why smugmug beats imageevent.com that would help me when that site is suggested vs. smugmug? I've got the rap down pretty well vs. pbase, but don't really know what to say when someone else mentions imageevent because I don't know them.

--John

Baldy

9 of 10 staying is about right. Our free trial retention is around 92% when everything is working great and it dips a little if we have a glitch. Annual renewal runs around 95%. Most of the ones who don't renew never maintained their albums, got busy, got divorced, died, etc.

PBase is in North Carolina, so they're faster in that area. We're in Silicon Valley. You noticed in the chart that PBase was down in Dallas and we sucked in Helsinki.

Probably with our new 100 mbps lines we don't suck so bad in Helsinki, but I'm not sure.

Dunno that much about Imageevent. They were in business for a year or more before us and I think their uptime and speed is pretty good. PBase has some unique things like their community and style sheets. We have unique things like customizing, selling prints, and customer service, but I can't 'think of what's unique about Fotki, Imageevent, Picture Trail and the like.

We're sometimes asked by the press why we're different than sites like Kodak Easyshare, so we have this page.

jfriend

Some specifics from a pbase and smugmug user

Baldy wrote:

Any specifics you have we can chase would be helpful.

Here's a specific observation reported by a user who has galleries at both pbase and smugmug:

"Hmm, clicking on any link or thumbnail on smugmug results in a noticeable delay of a couple seconds, while pbase usually returns in less than a second. Strange that you (and presumably others) aren't experiencing the same thing. Could it be something to do with my ISP (Comcast)?

--
http://www.pbase.com/ruze
http://ruze.smugmug.com"

I must admit that when I try out his two sites, the smugmug site feels slower than the pbase site. They are different layout styles so it's a little hard to compare apples to apples. He's in New Jersey which should probably favor pbase when he views the two, but I'm in the SF Bay Area which should favor smugmug when I view the two.

Try it yourself. There does seem like a pregnant pause on smugmug between mouse click and the next image loading that doesn't feel noticable on pbase. I'm not trying to beat you up on this, just sharing what some other users are saying and trying to help figure out why they are saying it.

--John

Baldy

No worries about beating us up, we just want to know.

Strange, but on this example I found smugmug to be faster. And it's a home page, which you would think would take awhile.

jfriend

Try going from one image to the next

Baldy wrote:

No worries about beating us up, we just want to know.

Strange, but on this example I found smugmug to be faster. And it's a home page, which you would think would take awhile.

The one where I thought smugmug was slower was once you'd opened a gallery on each site and you were navigating from one image to the next - the operation you do the most. I didn't really compare opening the home page of each.

--John

Baldy

I'm wondering if you're using Elegant gallery style for us and traditional for PBase? If you were to click the large link, so there's just one (big) image on the page, then keep hitting the forward arrow, does it seem slower than doing the same operation at PBase?

One thing I notice is our images appear all at once after a short pause, whereas PBase images start immediately but take some time to paint downwards. It may give the perecption of immediate response even if the load times are equal.

And Elegant style requires the fetching of a lot of thumbs from your cache, something we don't control and I'm wondering if on some machines that might be slow.

jfriend

Baldy wrote:

I'm wondering if you're using Elegant gallery style for us and traditional for PBase? If you were to click the large link, so there's just one (big) image on the page, then keep hitting the forward arrow, does it seem slower than doing the same operation at PBase?

One thing I notice is our images appear all at once after a short pause, whereas PBase images start immediately but take some time to paint downwards. It may give the perecption of immediate response even if the load times are equal.

And Elegant style requires the fetching of a lot of thumbs from your cache, something we don't control and I'm wondering if on some machines that might be slow.

Yes, this comparison was Elegant for smugmug vs. traditional for pbase. It seems like most smugmug galleries seem to be set to Elegant (I guess it's the default) and it seems to me like a lot of pbase galleries are set to traditional (I. myself, like Elegant better for navigating).

If the page doesn't render until all the images are there in the Elegant style, that could certainly explain a load time perception thing. In Firefox with a cleared cache, I see a new Elegant style page in smugmug load incrementally (some images painting before others). It is a bit of a shame in navigating within the same Elegant page that the whole page has to get loaded rather than just the one image that you are trying to change. When I load a fresh elegant style page, I see that the thumbs all paint before the main image. When picking a new thumb within the same elegant page that's a bummer. The thumbs probably all repaint (invisibly) while the one thing you are waiting to see comes in last. The perceived speed would be a lot faster if that could be reversed somehow so the large image painted first and then the thumbs came in.

--john