I was in meetings with some SharePoint users recently and reminded again of how, despite the great usability and flexibility of SharePoint in so many regards, it remains a painful experience to interact with it unless your on the same LAN.   Too much data is being moved around in an inefficient manner, and too much of the activity blocks a user from doing anything else while it’s going on.

It doesn’t have to be like this. Replify Syncstor users can link to SharePoint document (and other) libraries with a click of a button, and have that content cached on their PC and made available to them through the familiar Outlook user interface as another folder. Access to read, update or add new documents is instantaneous whether on the LAN,  on a constraint WAN connection, or totally disconnected.  All changes are propagated in both directions between the local cache and the SharePoint library, and activities like check-out and check-in can be handled automatically.

To show just how dramatically this improves the user experience I’ve created a little demo video.  I took the common use case of a user wanting to make a small update to a controlled document.    For a Replify Syncstor user, the entire process takes 20 seconds.  For a regular user, it’s 2 minutes and 40 seconds. That’s an incredible difference, and the poor user doesn’t really have the option to do anything else other than watch while the document is downloaded and opened, and then saved back to SharePoint.  Check it out.

Firstly CLAN:  we already support what we call “static” CLANs; that is the ability to define a group of PC clients as part of a collaborative peer-to-peer network sharing cache content.  For this release we’re adding “dynamic CLANs”, which is the ability for CLAN enabled clients to spontaneously link to peers and share cached data wherever the bandwidth and latency makes that advantegeous. We’ve added full encryption of the stored data to ensure that joining a CLAN doesn’t make your data visible to those without the right to see it.  We’ve also added a CLAN cache such that data pulled from peer clients won’t displace any hot data of your own, and will only move into your own personal CLAN cache if and when you first make use of it.

Secondly Speed: we’ve undertaken an overhaul of our cache indexing and storage mechanisms and currently, strapped to our test bench, it has doubled the throughput. We’re pretty thrilled by that - we haven’t even got the most important new mechanism fully implemented yet, so there are further improvements to come.  We’ve also found more opportunities for the optimizations to be tuned to traffic type and traffic pattern.  More on that some other time.

There are, of course, other refinements and improvements but this is the main payload.

Replify 4.1 will be available early Q1 2012; think of it as something to cheer you up during the cold, wet, dreary months between the Christmas holidays and spring. Unless of course you don’t live in our crappy climate (and why would anyone with a choice).

We’re taking part in the TRAC wan optimization product survey again this year.  http://www.trac-research.com/wan-optimization It’s interesting that the first time we took part in this, and other WOC surveys, it was hard to answer the questions because the concept of a “virtual” WAN Optimization appliance wasn’t really catered for.  Not so any more – all the vendors have following us into virtual appliance territory and so the product analyses reflect that.  Of course CLAN is still hard to describe within the existing framework but we’re confident that within two years these surveys will include questions like “what’s your peer-to-peer capability?”

And btw, the image is a slightly modified Hay Clan crest.  My mother is a Hay, so I’m entitled to use it (…er, but probably not to modify it). The Hays were commoners until they assisted Robert the Bruce in winning a key battle, and as a result he enobled them and awarded them all the lands from where he stood, to where his falcon chose to land.  Fortunately it didn’t just fly around and land on his head or we would have been heard of no more.

In preparation for the release I browsed the competitor’s sites to see what they were up to, and I found a new demonstration video on a major vendor’s site. I’ll not name names, I’m not trying to start a fight. They were trying to show how much they could improve the experience for a SharePoint user, and since that’s a big focus for us too, I took a look.  Simple demo:  4Mbps link with 75ms latency,  21MB Excel document being opened from a SharePoint site.  So consider this: without acceleration the spreadsheet took 64 seconds to open.   Next they show the first pass with WAN optimization enabled – as the download progresses they claim that the link is being accelerated 10-fold to 40Mbps.  If so opening the file would have taken about 6 seconds, but it actually took…  32 seconds.  Ok so they’ve halved the time (so maybe doubled the effective link speed) and that’s more or less what I would expect from a cold pass where compression is doing most of the work. The claim about a 10-fold accelration to 40Mbps is inconsistent with the observed behaviour and frankly ludicrous, but moving on…  Finally, they do a hot pass, and the file opens in… 14 seconds.   Now that was very surprising to me, I was expecting much more improvement than that.

So out of curiousity this morning I set up a 4Mbps link with 75ms latency and introduced it as a ”hop” in the connection between my laptop and a real SharePoint site in the Cloud. I then created a 25MB Excel spreadsheet using absolutely real data from financial forecasts, employee data, pricing information etc – so no cutting and pasting to create an unrealistically “de-duplicatable” file.  I put the file into a SharePoint doc library and began testing. Without any WAN acceleration the file took 111 seconds to open.  The first cold pass took 56 seconds to open the file – so as I would have predicted, (and just like the competitor) a cold pass halves the time or doubles the effective speed.  Now a hot pass: the file opened in 19 seconds.  So with Replify Accelerator, the hot pass is not twice as fast as the cold pass, it’s three times faster.  Or put another way,  for the competitor the difference between no acceleration and a hot pass is about x4.5,  and with Replify Accelerator it’s almost x6.   Or in tabular form:

I knew we were good, but I didn’t realize we are that much better than the competition.   We’re not standing still either, the 4.1 release is going to deliver a significant further speed increase as a result of some research we’ve been doing with Queen’s University over the last six months.

Here’s what we’ve got:

1) MAPI over HTTPS – we’ve supported vanilla MAPI for some time, but now we support the flavor of MAPI you’ll be using with Office365, and maybe already use within your own organization if you like to make it easy for people to access email outside the firewall.

2) Auto-configuration of the VA connection – virtual appliances will now try the various connection possibilities for themselves to find out what works and then save that configuration. To work well in any environment, our accelerator client must understand the nuances of the network between the user and our virtual appliance.  The Accelerator client will now determine automatically any proxy or firewalls getting in the way of its connection to the virtual appliance, and determine the most appropriate route for connection.  This avoids additional configuration during installation.

3) Resource management – one of the challenges of providing WAN Optimization as a “virtual” appliance, is that we can’t be sure whether a customer is sizing the system realistically, or just loading it up with more users and applications and hoping for the best. With 4.0 the system will monitor consumption of cpu, memory, connections, and when it hits thresholds (which can be modified by an informed user) it will reject further users and send them directly to the application without any optimization

4) Health dashboard – each VA now has a handy dashboard which allows you track a slew of resource measures in detail and allocate more resources when necessary

5) Lots of point improvements to performance and usability

And of course CLAN is available for selected customers

New users and trial customers will get the release via the download links right away.  Existing user’s clients will not flag the  availability of the new release right away. That’s because one of the changes we’ve made requires that the VA be upgraded before the clients in certain use cases, and we don’t want anyone to lose connectivity.

And now for something completely different. I pulled out the user manuals for our Riello UPS  recently and was intrigued by an envelope that fell out.  Inside was a little plastic Riello logo along with an installation guide  (see below) explaining how to stick it to the device correctly depending on whether the UPS was horizontally or vertically mounted in the rack.  I was initially incredulous and then laughed like a loon. 

Is this an elaborate practical joke to be applauded, or lunacy from some totally humorless Brand manager at Riello? I can’t make up my mind whether this is the silliest thing I’ve ever seen, or my satire detection has completely failed.

Given that the Riello logo looks like a question mark  compounds the “what the *?#$ ?” feeling.

We’re in release planning for the next release of Accelerator – due out in August. We’ll update you on feature content once it’s locked down.

BTW Apparantly Steamboat Willy isn’t an occupational disease after all, but I’m not sure about Boxcar Willy – I suspect it might be.

By the time this blog goes live, 3.6 will be available for trial, and as a free upgrade for all customers currently in-support. Your Replify clients will start popping up little messages to tell you that the new client is available but if you want to check out the contents before you proceed with an upgrade, take a look at the release notes.

As well as testing until our eyes-bulged, we’ve been taking time out to brief analysts on our strategy – that’s always a little nerve-racking because if the analysts don’t “get it” right away, that can either mean you’re ahead of the game (which can be good), or that you’re heading in the wrong direction (which is never good). We’ve come away from those sessions very encouraged – the analysts all “got it” and could see the value of the “hardware-free WAN optimization” and mainly differed in how quickly they thought the market would embrace it. I think that puts the onus back on Replify to get into CLAN customer trials at the earliest opportunity and make our point. Actions speak louder than words here.  We wrapped-up our analyst piches with the following manifesto, which sums up our proposition nicely: 

The trends of mobility, collaboration and cloud drive us towards the need for a new user-centric approach

Personal WAN optimization is already valuable but undervalued

…with CLAN it becomes fully the equal of existing solutions but without the cost and complexity

Add offline working and you have the most complete solution for remote and mobile workers, and consumers of cloud services

This is a fairly large release of Accelerator – no really big individual features (saving that for CLAN next quarter) but many small additions and improvements.

In no particular order:

- Improvements to our handling of CIFS which give a performance boost for scenerios such as viewing folders with large number of files.

- Support of WFP (Windows Filtering Platform) in Windows 7. This replaces the DNE packet interception that we used to use and has several benefits; improved throughput,  no need to restart after installation, and no tripping over inconsiderate DNE based products like CheckPoint VPN.

- Firewall friendly mode – requires only two ports to be opened in the firewall (or one if you’re not using SSL).  You can use port 80 if you like.

- Full service configurability – you can choose to intercept any protocol on any port, and choose which optimizations you wish to apply e.g. compression only, de-duplication, the whole shooting match.

- Replify Backup Accelerator – a new *personality* for the Accelerator so that if you’ve bought a license key purely for back-up acceleration, the product presents you with a very simple set of configuration screens and self-configures for DoubleTake, EqualLogic or any other similar product.

- A slew [Irish Gaelic sluagh, multitude, from Old Irish slúag.] of performance and usability improvements.

- There might be a couple of bug fixes in there too.  Ok, I admit it, there are several bug fixes in there.

There’s also a cunning little hidden feature which gives a boost to the PC client performance if you don’t mind using more memory. Enabled on request

I feel they may have underestimated the breadth of our capability a little; I guess all the vendors will feel that way to a degree, but I think they described our strategy clearly and I was delighted  that we were awarded “Best Mobility Strategy” based on our leadership in the remote and mobile user space.  That is indeed our focus; it’s a significant and rapidly growing proportion of enterprise knowledge workers and the main players pay relatively little attention to it, in some cases offering much reduced functionality compared to their site-level WOCs.  When we release CLAN later this year, we’ll take the mobility play to a whole new level and close the gap between site and individual acceleration results.

Russian language version

Replify will be formally unveiling its CLAN technology at Internet World in May this year.  The event, held at Earl’s Court in London, hosts over 300 solution providers and 12,000 visitors, combining five shows in one event. We’ve chosen this event because of its exciting combination of themes including: cloud, content management, and e-business.

Replify’s CLAN feature delivers serverless WAN optimization to co-located users, with the levels of acceleration previously only associated with shared site appliances.   CLAN allows groups of users to dynamically bind and share their personal caches, to create a Virtual shared cache, with de-duplication for all common accessed content.  No other product today has this feature, allowing enterprises to remove further cost and complexity from their IT infrastructure while delivering the WAN optimization they need.

Come along and see how our small footprint  PC client delivers true usability for remote applications and content, and provides LAN-like performance for collaborating workers, without the need to deploy anything into a site. 

CLAN is going to change the whole paradigm for WAN optimization – for site-to-site,   site-to-cloud,  user-to-site and user-to-cloud.   We believe that within a couple of years few organizations will even consider deploying expensive to purchase,  and expensive to support,  tin boxes to deliver the application acceleration that distributed working and cloud services demand.

WAN optimization is conventionally based on four techniques working together:

1) TCP/IP optimization – this varies from vendor to vendor, but includes things like agressive window scaling,  re-ordering of lost packets.

2) Protocol optimization – many application protocols are chatty and inefficient because they assume LAN-like latency and bandwidth.  Spoofing, pre-fetching, bulking-up transfers, stripping or modifying protocol headers etc. can greatly improve the user’s experience but typically have to implemented on a protocol-by-protocol basis.

3) Block de-duplication through caching.

4) Compression – just what it says. On average delivers a 50% reduction in data volume for many types of content.

Now techniques 1,2 and 4 are just as valuable to an individual user as they would be to a group of users whose activity is proxied through a shared device. So far so good – nothing here to suggest that an individual WAN Accelerator wouldn’t be a great thing.

The really powerful benefits of WAN optimization, however, come from (3) – de-duplication.  Those that dismiss individual WAN optimization tend do so on the basis that de-duplication only adds value for groups of users, because each user benefits from the activity of other users. In other words, because someone else’s PC has downloaded a file, or browsed a web site,  you will get the benefit of doing the same things from your PC.  For sure there are some use cases where this is largely true: software updates, viewing of web pages perhaps, viewing the latest funny video on YouTube – each user does this only once, but with many users in a site, de-duplication drastically reduces the data transfer. 

But we’ve moved on from the read-only Web 1.0 world into a much more collaborative and interactive Web 2.0 situation.

Consider these scenarios:

a) I’m part of a team collaborating on a number of documents (or designs, drawings, schematics, videos, graphics etc). We’re sending versions of the content back and forth between us as we refine it.

b) I’m receiving content by email which I need to add to a remote records repository, or to upload to SharePoint, or to email to many members of my team.

c) I’m a school pupil using a shared PC.  I’m visiting the same learning applications, browsing the same web sites,  completing the same on-line assessments, and viewing the same videos as all my class-mates.

d) I’m a social network guy who updates his face book page three times a day and visits many other friend’s pages several times a day. I also shop on Amazon, look at the TV listings,  hit the BBC news site and check my stock prices.

The key element of the first two scenarios is that the same “stuff” goes back and forth between individuals with small changes, and the key element of the last two scenarios is that the same computer is accessing the same, or slightly modified, content repeatedly. So in all of these common situations, de-duplication at the device level makes an enormous difference.  And we know this to be true because users tell us. In particular, remote users, who have to access corporate applications, web sites and document repositories on the road, tell us that without Replify they couldn’t do their job.

Now Jeff points out that Windows 7 includes some improvements and he is of course correct.  The new office document formats are already compressed, the newer version of the file access protocol, CIFS, is much better than the previous version, and Branch Cache handles some shared access to office documents (but is rather limited). Interestingly these are in the areas that reduce the value of a shared WAN acceleration device just as much as an individual WAN Acceleration client, so I can agree with him about this, while still not feeling it in any way undermines a belief in the value of individual WAN optimization.

There is another reason perhaps why individual acceleration has been under-valued, and that is that the available products have been pale shadows of the site-level appliances, with reduced funtionality (perhaps only providing compression, or only accelerating in one direction).  Not so with Replify Accelerator,  our PC client is exactly the same code that runs in our virtual appliance and it does all the things that a world-class WOC does.

This not a matter of opinion or dogma and the truth can be easily demonstrated.  We have a virtual appliance hosted on Amazon’s cloud platform, along with some sample file shares, and SharePoint content.   It takes only a few minutes to download and install our PC client, see the benefits of accessing the content on both cold and hot passes, and understand how much value this would add to your collaboration and case-based workflow.  If you want to try this, just email us and we’ll get you set up.

The story doesn’t end here though – later this year Replify will be releasing an enhancement to Replify Accelerator called CLAN (Co-operative LAN).  With CLAN, individual clients will share their caches with others on the same LAN, and so you get the best of both worlds: personal acceleration when alone plus a virtual shared cache when working in the office.  So you can have an accelerated branch office without any server infrastructure. I think that’s pretty cool.