Posts Tagged ‘SSD’

Reaching for SANity with the IBM MQ Appliance

October 24, 2017

Appliance reaching

We all like to keep our word, and even though a Statement of Direction is not always a promise, at IBM we like to try and deliver what we set out, as long as the market and need hasn’t changed.

And for customers who have been following the MQ Appliance since the start, back in March 2015, you might recall that there was a Statement of Direction about adding support for the fibre channel cards that were part of the MQ Appliance hardware to allow connectivity to a SAN for customers who wanted that. And we are delighted that as part of the MQ V9.0.4 update on the MQ Appliance, that IBM is delivered on that Statement of Direction.

MQ V9.0.4 is a very significant update for MQ and most of the updates that are not specific to the MQ Appliance are described in my other blog entry here. However, in addition to those features there are specific MQ Appliance updates, and a dedicated announcement letter which I will try to explain in more detail here.

Firstly, the SAN support. One of the reasons the MQ Appliance has been so popular with customers is the built-in storage (2x 3.2TB SSDs in a RAID 1 configuration) and the simple HA configuration allowing a pair of MQ Appliances to replicate messages and log data between them on a QM by QM basis with just a couple of clicks on setup. And a lot of the updates to the MQ Appliance has been to support the wider use of this popular deployment use case.

Appliance SAN options

However, some customers have always wanted us to add support for SANs for some of their use cases. One use case was that 3.2TB of onboard storage might not be enough. If you consider when a consuming application might fail over a long weekend and the queue depths might get very high. One of our customers recently said, “an empty queue is a happy queue” and this is true, but in the case of a protracted failure you want to ensure your queues can hold all the messages. So, if you are concerned about this scenario, then you might want to have the Queues supporting that use case on external storage where there is no effective limit to queue depth.

The other use case, which tends to come up more commonly in customer discussions is around the thought of “consistency groups”. This is when you are trying to recover from a site failure at your disaster recovery location. This is typically a manual task, unlike the automated High Availability configuration. Part of this manual task will be to establish the last known transactions of not just MQ but other parts of the IT infrastructure such as applications and databases. This is likely to be easier if all the data points from all these ‘moving parts’ are stored on a common storage area, and this is replicated consistently between the SAN on one site and the SAN on the disaster recovery site. So this use case is now supported by the MQ Appliance with customer selected Queue Managers able to choose the SAN storage instead of the internal storage and have a MQ Appliance in the disaster recovery site read the replicated messages and data from the SAN there.

A second update to the MQ Appliance specific features is to provide customers with a quicker and simpler way to ensure that Queue Managers on the MQ Appliance have the best allocation of resources on the MQ Appliance. When Queue Managers are initially created the MQ Appliance determines how much space to allow to them. However if workload on a Queue Manager grows faster than other Queue Managers, it is likely that the allocated resources might need to grow to match the likely workload. However, it hasn’t been easy to do that. But with the MQ V9.0.4 update on the MQ Appliance IBM has added dynamic disk allocation, enabling resources for a Queue Manager to be increases even after it was initially created. This will make the ongoing operation and support of production workloads on the MQ Appliance quicker and simpler.

Appliance MFT

Finally, an update off the MQ Appliance will have a positive impact on potential use cases for the MQ Appliance. Virtually every customer still moves large amounts of data around their business as files and file contents. Much of this is unmanaged, insecure and unreliable, and for a number of years IBM has provided a solution: MQ Managed File Transfer. This enables file contents to move from point to point over the MQ network as MQ messages, taking advantage of MQ’s reliability, security and manageability. Part of the MQ Managed File Transfer functions was a logger to track which files were sent over MQ, and this, if used, always needed to run on the same machine as the Queue Manager. This prevented some MQ MFT use cases from being deployed in environments where the MQ Appliance was the only MQ Queue Manager. Now, with MQ V9.0.4 the file logger component of MQ MFT can be deployed remotely from the Queue Manager, meaning it can be used in locations by pointing to a remote MQ Appliance, without the need to install anything on the MQ Appliance, which has always been prohibited. It’s always good to allow customers more flexibility in deployment.

This combination of enhancements, along with the many non-Appliance specific MQ updates in the MQ V9.0.4 release means that there should be a lot of increased opportunities to consider the MQ Appliance as the right deployment option. After all many customers who review it with regards to overall costs find it has the lowest Total Cost of Ownership, as well as outstanding reliability.

Advertisements

Beginning the new, looking back to the old

January 17, 2017

janus2

The month of January is named after the God Janus – who both looked forward to the new year and back to the old one. So it is perhaps time to set ourselves up for what will be no doubt another very busy year for IBM MQ by a quick review of 2016 – looking at what you should have seen, and also finding time to tell you something new, which you are unlikely to be aware of.

So a quick recap first. In June we released a hardware refresh for the IBM MQ Appliance, adding large capacity SSDs and additional 10Gb network ports as described here. And IBM MQ brought out MQ V9.0 with a new option for end-to-end encryption with an order of magnitude performance boost, and CCDTs now accessed through a URI – and this was described here.

There were additional enhancements in November with IBM MQ moving to MQ V9.0.1 – the first Continuous Delivery release, with MFT enhancements and repackaged MFT Agents, availability of the new MQ Console, and the initial delivery of REST API verbs. These were all described here. And the IBM MQ Appliance also moved the MQ V9.0.1 and added additional features like Floating IP support, SNMP and LDAP authentication of admin accounts. This was written up here.

pvu_1

So if we are all ok with that, I had better share the news that you missed at the end of last year. First a word or two about Processor Value Units. This is IBM’s typical capacity based pricing metric for software. Each machine type and processor type has a PVU rating per core. And software products like IBM MQ have a price per PVU. So as a customer you buy a number of PVU entitlements to meet your capacity need and then deploy IBM MQ on the hardware that matches the PVUs you have bought. However this means you need to always count and be sure that the capacity you have provided to IBM MQ is in line with the entitlement you have, and the physical machines you are running on. But more and more these days software is being deployed on environments that are more abstracted from the actual physical machines – and the capacity being allocated, either on premise or in a cloud, is assigned as virtual cores. But with IBM MQ (and other products) priced only by PVUs, there was some confusion in mapping PVUs to virtual cores.

vcpu

On December 6th 2016, IBM MQ addressed this by adding a Virtual Processor Core metric to its pricing. This is only available as a monthly pricing metric but provides a new simple, and possibly more appropriate way of buying capacity for IBM MQ deployed in these virtual environments either on premise or in clouds where IBM MQ is deployed with a number of virtual cores of capacity rather than into a fixed physical machine. This is an additional metric. The PVU metric with both perpetual and monthly pricing is still available, but customers now have an additional option of the Virtual Processor Core pricing. There is no announcement letter for this, but the pricing is already available for IBM MQ and for IBM MQ Advanced, so simply ask your IBM sales rep or business partner about this if you want to know more.

Certain customers who can find it difficult to count PVUs might find this very useful. These might include customers such as retailers or retail banks where IBM MQ can be installed in 1000+ different environments, and for customers like this there are other ways to price for this type of deployment so again ask your IBM rep.

That was the last news and updates from 2016, but there is plenty to come in 2017. And you don’t need to wait for long. Just one week to go and I expect to have something new to share here. Not long to wait.

keepcalm

No waiting in these queues. IBM MQ V9 and the MQ Appliance M2001 delivers fast, reliable and secure message queuing

June 29, 2016

wile_e_coyote

Recent weeks have been pretty busy on this blog, reflecting just how busy the MQ development team has been in bringing out new and updated offerings in MQ V9 and the MQ Appliance M2001 here and here. And of course in our cloud messaging options.

As both of these have been fairly full of new content I thought I would do just a short update to focus on a couple of key benefits which are specifically measurable in these 2 refreshed offerings. After all, a lot of the new and improved features can sometimes be hard to quantify in terms of the benefits they provide, but in each offering this time there are some easy to define benefits.

As you may have seen in my most recent update, the MQ Appliance M2001 added large capacity SSD storage which enables much faster throughput for persistent messages. These are the messages that get written to storage to ensure they are still available in the case of failure before the message has been successfully deliver to all consumers. At high rates of message throughput, there can be a lot of contention for access to storage with traditional hard drives. With the new MQ Appliance M2001, this potential bottleneck has been removed. You can now read the latest MQ Appliance M2001 performance report here which shows that the performance in those scenarios which saw large volumes of persistent messages sees improvement of up to 3.5 times the previous message rate.

Clearly this represents a significant improvement and given that persistent messages are used in those business critical situations where IBM MQ delivers so much value, it is a hugely important benefit.

 

In MQ V9 there were a number of enhancements but the one I specifically want to call out is, as part of the MQ Advanced package, the enhancement to MQ Advanced Message Security (MQ AMS). The change here was to add a new mode of operation – Confidentiality. This new mode changed the way in which the encryption operations are performed on the message contents (MQ AMS offers policy based encrypted message contents which ensures data at rest is protected in case of a security breach). The goal of this change was to continue to offer a strong level of security for the message contents without too big of an impact on the performance and throughput from the effects of the encryption used.

Now instead of new asymmetric keys being generated for every exchange, the feature can be configured to allow for reusable symmetric keys to be used after the initial generation of an asymmetric key. This still provides a very high level of security, but depending on the reuse count before a new asymmetric key is generated, can drastically cut the performance overhead. The benefits can see more than an order of magnitude increase in throughput. You can see a quick snap shot of some of the early results in Jon Rumsey’s blog here – which includes a small table showing performance improvements exceeding 10x gains. With everyone concerned about security these days, the ability to better protect your information and customer data with little performance impact has to be a good thing.

 

So what are you waiting for? With secure, reliable enterprise messaging for on-premise deployments, cloud deployments or physical appliances, there is no waiting with IBM MQ V9 or IBM MQ Appliance M2001.

no-waiting

[An interesting history of Wile E. Coyote here]

Flash aaaahh – saviour of the universe: IBM MQ Appliance M2001

June 10, 2016

flash_gordon_facebook_cover_by_audrey41lorgeoux-d538dgo

Anyone who is a fan of cheesy sci-fi movies, or soundtracks by Queen will have the words “Flash, Flash I love you but we only have 14 hours to save the earth” running through their head, along with the line in the song that goes “Flash aaaaah, saviour of the universe”. And of course he did save the universe from Ming the Merciless.

But what if I told you Flash could also save your business? Not Flash Gordon of course, but flash storage, in the form of the SSDs that are now a part of the IBM MQ Appliance M2001 which is now generally available (June 10th 2016). We did cover this in an earlier blogpost, but I thought I would take advantage of our initial shipment date to cover just how critical the IBM MQ Appliance, backed by state of the art 3.2TB SSDs can be to your business.

MQ Appliance M2001

Each MQ Appliance M2001 model has 2 of the 3.2TB SSDs in a RAID 1 configuration. This means that every persistent message and all log data is written not just once to the SSD storage, but twice giving you complete redundancy of data. And a key part of the MQ Appliance functionality is the High Availability configuration – essentially nothing more than a simple menu option when creating a queue manager – allowing you to have the MQ queue manager on one MQ Appliance synchronously replicated to another MQ Appliance. This means that any message written to the SSDs on one MQ Appliance is not just copied to the second pair of SSDs but is written under the same unit of work that writes the messages on the first MQ Appliance. This therefore means you have 4 copies of the message stored for both reliability and availability.

Another part of the MQ Appliance update was the ability to do not just synchronous replication for High Availability but also asynchronous replication for Disaster Recovery to another MQ Appliance. Therefore you can point the MQ queue manager at another, typically off-site MQ Appliance and the same message will replicate there, ensuring there are another 2 copies of the message, and providing your business with a highly resilient messaging system designed to ensure optimum reliability and availability of messages.

After all, think about how important your messages are to your business. In effect, they are your business. Your messages are your business transactions, your new orders, your customer address details, your stock levels and distribution information. Lose your messages and you lose everything.

With the latest SSD technology inside the MQ Appliance you are calling on Flash to save your business – and with the MQ Appliance M2001, Flash saves the day again.

 

[Flash Gordon image title image above is from  http://orig07.deviantart.net/fc1d/f/2012/163/5/d/flash_gordon_facebook_cover_by_audrey41lorgeoux-d538dgo.jpg]

Going faster by not moving – IBM Appliance M2001

April 19, 2016

totoise-rocket-patch

Go faster. Faster. Move it! Or actually don’t move it. There are times when to go faster you need to stop moving. We are all familiar with the parable of the tortoise and the hare – where slow and steady wins the race. But what about not moving at all? Sometimes that makes you go much faster. And in the case of the latest update to the IBM MQ Appliance that is exactly what we are doing. Hopefully you already know about the MQ Appliance, which IBM releases early in 2015, and have continued to enhance since its release. You can read my original entry here, and the update at the end of last year here.

 

But today, April 19th, IBM is announcing another update to the MQ Appliance which not only provides additional functional enhancement, by allowing queue managers to both synchronously replicate for HA and also asynchronously replicate for DR, and adds support for the AMQP based MQ Light API, but also sees a small but important hardware update, making this a slightly refreshed model – the MQ Appliance M2001.

HA+DR

There are 2 key hardware changes in this model update. To help support the simultaneous HA and DR function, which would use both existing 10Gb network cards, the existing 2 port connection is being replaced with a 4 port connection, providing 4 of these 10GB network ports, enabling 2 to be used for HA and DR and ensuring 2 can be used by applications connecting to the appliance, as well as the existing 1Gb ports.

M2001

The second hardware change is the replacement of the existing pair of 1.2TB hard disk drives (HDD) with a pair of 3.2TB solid state drives (SSD). As well as the benefit of the greatly increased storage capacity, the major benefit of using SSDs is the increase in performance for persistent message throughput. The MQ Appliance is a highly capable system which can process a lot of MQ messages. However, when using persistent messaging, which needs to be written to disk, it is critical that the storage can keep pace with the high rate of workload being handled by the system and at times with heavy workloads the spinning disk simply couldn’t move fast enough. IBM has selected the latest generation of SSDs to provide large capacity, high performance for both reading and writing data at high rates, and also this latest generation of SSDs, even if the MQ Appliance is used heavily all day, every day, should last for the 5 year supported lifespan of the MQ Appliance. Therefore, this provides the payoff from our ‘tortoise and hare’ parable – with no moving parts in the SSDs, they can be a lot faster than spinning disks. Expect to see updated performance figures for the new MQ Appliance M2001 around the time of its availability (June 10th 2016), but early figures suggest for some workloads performance improvements of up to 3 times have been seen.

 

There continue to be 2 editions of the MQ Appliance – the M2001A, providing full access to all the processor cores, and the M2001B, which provides access only to a subset of the cores – with an upgrade available from the B to the A system if needed. For customers who may have already purchased the MQ Appliance M2000, please talk to your IBM sales rep to see whether your appliance can take advantage of an upgrade of the HDDs and network card if available.

 

With the improved HA and DR functions, the increased storage capacity and the greatly increased performance, IBM believes this enhanced MQ Appliance makes even more sense to be used as the heart of your IBM MQ deployment, or as a highly available pair of appliances that can be deployed anywhere you need MQ capability. And for customers who may be running older versions of MQ which were recently subject to an announcement of End of Support – as can be seen here – then the latest version of the IBM MQ Appliance can represent a very good deployment option which is then far simpler to deploy as well as to maintain.

 

By moving from spinning disks, to SSDs with no moving parts, you really can go faster by standing still.